Measured

Name: Fox Red Labs benchmark results
Creator: Fox Red Labs
License: https://creativecommons.org/licenses/by/4.0/

One bench, one person, medians of three runs.
Every number has a method, a date and a CSV row. Your numbers will differ.

21 boards measured on one set of tests, side by side.

How the boards compare

The three tests that separate them most. Sort by any column; click a board, or pick it here, to follow it through every chart below.

Highlight

	CoreMark per MHz, higher is better	ns, lower is better	µs, lower is better, log bars
CC35X1Cortex-M33	4.49	not run	not run
LP-EM-CC2755P10Cortex-M33	4.44	219 ns	10.5 µs
LP-AM13E230Cortex-M33	4.36	not run	1.33 µs
LP-CC1354P10-1Cortex-M33	4.17	479 ns	21.4 µs
LP-AM263PCortex-R5F	3.74	488 ns	not run
LP-MSPM33C321ACortex-M33	3.40	200 ns	5.03 µs
FRDM-MCXA153 (NXP)Cortex-M33 · NXP	2.94	177 ns	not run
Raspberry Pi Pico 2 HCortex-M33 · Raspberry Pi	2.92	173 ns	not run
ESP32-C6-DevKitC-1-N8 (Espressif)RISC-V · Espressif	2.73	1069 ns	not run
LP-MSPM0L1306Cortex-M0+	2.45	781 ns	448 µs
LP-EM-CC2340R5Cortex-M0+	2.31	583 ns	337 µs
LP-MSPM0G3507Cortex-M0+	2.30	362 ns	88.5 µs
LP-MSPM0G5187Cortex-M0+	2.29	450 ns	181 µs
LP-MSPM0G3519Cortex-M0+	2.28	388 ns	89.3 µs
NUCLEO-U083RC (ST)Cortex-M0+ · ST	1.96	458 ns	299 µs
LAUNCHXL-F2800137C28x	1.18	250 ns	2.58 µs
LAUNCHXL-F28P65XC28x	1.06	150 ns	not run
LP-MSPM0C1104Cortex-M0+	not run	958 ns	not run

Bars scale to the best board in each column (control loop on a log scale, as in its chart). Boards with none of these three results are left out. Outlined bars are other vendors' boards; a muted value is an unofficial build, not ranked. Hover a value for the clock and build; each chart below has the rows behind it.

Two things to know before you start. At -O3, where GCC is built with -funroll-loops, TI Arm Clang scores 26% higher than GCC on the MSPM0 G3507. In float code, each chip on its own math unit, the LAUNCHXL-F2800137 runs the control loop 34 times faster than the MSPM0 G3507.

Compiler and OS choice, LP-MSPM0G3507

CoreMark per MHz. At -O3, TI Arm Clang scores 26% higher than GCC with loop unrolling. At -O2 the order flips: GCC is 15% ahead. Adding Zephyr costs GCC 7% at -O2.

Control loop, C2000 against MSPM0

ADC trigger to new PWM value in float code, each chip on its own math unit. The C2000 took about 1/34 of the time.

CPU speed

CoreMark per MHz by board

The Cortex-M33 boards score about 1.7 times the Cortex-M0+ boards per clock.

CoreMark per MHz: how much general code a core gets done each clock tick. The same source runs on every board at the clock shown in its hover, the board's top clock unless the chart note says otherwise. TI's compiler builds the TI boards and GCC the others, with no OS unless marked.

TI boardsother vendors' boardsyour board

LP-EM-CC35X1

4.49

LP-EM-CC2755P10

4.44

LP-AM13E230

4.36

LP-CC1354P10-1

4.17

LP-AM263P

3.74

LP-MSPM33C321A

3.40

FRDM-MCXA153 (NXP)

2.94

Raspberry Pi Pico 2 H

2.92

ESP32-C6-DevKitC-1-N8 (Espressif)

2.73

LP-MSPM0L1306

2.45

LP-EM-CC2340R5

2.31

LP-MSPM0G3507

2.30

LP-MSPM0G5187

2.29

LP-MSPM0G3519

2.28

NUCLEO-U083RC (ST)

1.96

LAUNCHXL-F2800137unofficial build, unranked

1.18

LAUNCHXL-F28P65Xunofficial build, unranked

1.06

Hover a score for the core, clock and build. GCC built FRDM-MCXA153 (NXP), Raspberry Pi Pico 2 H, ESP32-C6-DevKitC-1-N8 (Espressif), NUCLEO-U083RC (ST); TI's compilers built the rest. The C28x build has three source changes for its 16-bit byte, so its score is shown for scale only. Each board ran at the clock in its hover; NUCLEO-U083RC ran at 48 MHz, its maximum is 56 MHz.

All rows

CoreMark iter/s, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X	LP-AM263P	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
armgnu14-bare-o2										267.596 MHz
armgnu14-bare-o3										275.396 MHz
armgnu14-bare-o3-48mhz										137.648 MHz, 1 run
armgnu14-bare-os										219.396 MHz
c28patch-ti-c2000-o2-flash								97.1120 MHz, 1 run
c28patch-ti-c2000-o2-ram								114.1120 MHz, 1 run
c28patch-ti-c2000-o4-flash								107.3120 MHz, 1 run
c28patch-ti-c2000-o4-ram								124.4120 MHz, 1 run
c28patch-ti-c2000-o4s0-flash								106.5120 MHz, 1 run
c28patch-ti-c2000-o4s0-ram								123.4120 MHz, 1 run
c28patch-ti-c2000-o4s5-flash								141120 MHz, 1 run									211.8200 MHz, 1 run
c28patch-ti-c2000-o4s5-ram								155.3120 MHz, 1 run									258.9200 MHz, 1 run
gcc-bare-o2	140.580 MHz, 1 run									263.596 MHz
gcc-bare-o2-48mhz									92.148 MHz
gcc-bare-o2-m33											419.9150 MHz
gcc-bare-o2-m33-gcc12											407.9150 MHz
gcc-bare-o2-riscv											503.1150 MHz	443.8160 MHz
gcc-bare-o3		54.632 MHz								282.296 MHz
gcc-bare-o3-48mhz										141.248 MHz
gcc-bare-o3-m33											434.1150 MHz
gcc-bare-o3-m33-gcc12											437.6150 MHz
gcc-bare-o3-m33-gcc12-ram											437.6150 MHz
gcc-bare-o3-m33-ram											434.1150 MHz
gcc-bare-o3-riscv											495.2150 MHz	436.9160 MHz
gcc-bare-o3-riscv-ram											493.8150 MHz
gcc-bare-o3-unroll	145.480 MHz, 1 run
gcc-bare-o3-unroll-48mhz									97.348 MHz
gcc-bare-o3-unroll-reset									8.224 MHz
gcc-bare-o3unroll		58.632 MHz
gcc-bare-o3x-m33											511.6150 MHz
gcc-bare-o3x-riscv											585.8150 MHz
gcc-bare-os	99.680 MHz, 1 run									233.696 MHz
gcc-bare-os-48mhz									71.548 MHz
gcc-bare-os-m33											339.3150 MHz
gcc-bare-os-m33-gcc12											364.7150 MHz
gcc-bare-os-riscv											379.9150 MHz	341.2160 MHz
gcc-linux-o2-1t																			10,441 to 10,4622000 MHz
gcc-linux-o2-1t-oursrc																			10,466.3 to 10,467.42000 MHz
gcc-linux-o2-mcpu-1t																			10,450.4 to 10,471.22000 MHz
gcc-linux-o3-1t																			9,834.6 to 9,839.9
gcc-linux-o3-2t																			19,579.9 to 19,583.4
gcc-linux-o3-mcpu-1t																			10,333.8 to 10,338.12000 MHz
gcc-linux-o3-mcpu-unroll-1t																			10,781.7 to 10,783.42000 MHz
gcc-vendor-o3												451.1160 MHz
gcc-zephyr-o2	130.280 MHz, 1 run	53.732 MHz, 1 run		12680 MHz, 1 run		74.148 MHz, 1 run			88.148 MHz, 1 run	251.496 MHz, 1 run
gcc-zephyr-o2unroll	139.980 MHz, 1 run	59.732 MHz, 1 run		135.380 MHz, 1 run		84.148 MHz, 1 run
gcc-zephyr-o3	136.780 MHz, 1 run	56.732 MHz, 1 run		130.280 MHz, 1 run		79.548 MHz, 1 run			92.248 MHz, 1 run
gcc-zephyr-os	9580 MHz, 1 run	4232 MHz, 1 run		9280 MHz, 1 run		58.448 MHz, 1 run			70.748 MHz, 1 run	223.596 MHz, 1 run
gcc12.2-zephyr-core1-o2-ddr																				4,007 to 4,007.51250 MHz
gcc14.2-static-o2-1t-kvm-guest-alpine																				3,926 to 4,0151250 MHz
gcc14.2-static-o2-1t-kvm-host																				4,077 to 4,1531250 MHz
gcc15.3-linux-o2-1t																				4,255.61250 MHz, 1 run
gcc15.3-linux-o2-1t-recheck																				4,254.41250 MHz, 1 run
gcc15.3-linux-o2-1t-zephyr-running																				3,895.61250 MHz, 1 run
gcc15.3-linux-o2-2t																				8,428.21250 MHz, 1 run
r5f-ticlang-nortos
starmclang-bare-o3-48mhz									87.948 MHz
starmclang-bare-omax-48mhz									115.148 MHz
stgcc14-bare-o3-48mhz									93.948 MHz
stgcc14-bare-o3-unroll-48mhz									98.848 MHz
ti-port-o2														612.5200 MHz
ti-port-o3														871.6200 MHz
ti-port-o3-ocsram																		1,464.5400 MHz
ti-port-o3-ocsram-60k-beside-loop-core1																		1,463400 MHz
ti-port-o3-ocsram-60k-beside-loop-core2																		1,460.2400 MHz
ti-port-o3-ocsram-60k-beside-loop-core3																		1,461.4400 MHz
ti-port-o3-ocsram-60k-cores1to3-alone-core1																		1,463400 MHz
ti-port-o3-ocsram-60k-cores1to3-alone-core2																		1,460.2400 MHz
ti-port-o3-ocsram-60k-cores1to3-alone-core3																		1,461.4400 MHz
ti-port-o3-ocsram-all4-core0																		1,462.1400 MHz
ti-port-o3-ocsram-all4-core1																		1,464.1400 MHz
ti-port-o3-ocsram-all4-core2																		1,463.7400 MHz
ti-port-o3-ocsram-all4-core3																		1,464.1400 MHz
ti-port-o3-ocsram-all4-sum																		5,854400 MHz
ti-port-o3-ocsram-core0-alone-freertos																		1,463.3 to 1,463.4400 MHz
ti-port-o3-ocsram-core1-alone																		1,464.1400 MHz
ti-port-o3-ocsram-core2-alone																		1,463.7 to 1,463.8400 MHz
ti-port-o3-ocsram-core3-alone																		1,463.5400 MHz
ti-port-o3-ocsram-lockstep																		1,464.5400 MHz
ti-port-o3-xip																		1,394.5 to 1,394.6400 MHz
ti-port-o3-xip-l2-64k																		1,464.5400 MHz
ti-port-os														476.7200 MHz
ti-port-oz														438.3200 MHz
ticlang-bare-o2	122.680 MHz, 1 run	54.332 MHz, 1 run		122.780 MHz	122.680 MHz								376.9160 MHz	612.7200 MHz
ticlang-bare-o2-32mhz				54.832 MHz	54.832 MHz								96.932 MHz
ticlang-bare-o2-ocsram																		1,026400 MHz
ticlang-bare-o2-tcm																		1,039.8400 MHz
ticlang-bare-o3	183.680 MHz	78.332 MHz, 1 run	Not for this chip 4,330 short	182.780 MHz	182.880 MHz	111.148 MHz, 1 run	199.948 MHz						544160 MHz	872.4200 MHz	426.396 MHz
ticlang-bare-o3-100mhz														442.3100 MHz, 1 run
ticlang-bare-o3-32mhz	78.832 MHz, 1 run			78.832 MHz	78.832 MHz								140.332 MHz
ticlang-bare-o3-4mhz		10.54 MHz, 1 run
ticlang-bare-o3-50mhz														221.150 MHz, 1 run
ticlang-bare-o3-ocsram																		1,464.2400 MHz
ticlang-bare-o3-tcm																		1,497.8400 MHz
ticlang-bare-o3-tcm-core1-alone
ticlang-bare-o3-tcm-core2-alone
ticlang-bare-o3-tcm-core3-alone
ticlang-bare-o3-tcm-lockstep
ticlang-bare-o3-xip
ticlang-bare-o3-xip-sp
ticlang-bare-os	114.580 MHz, 1 run	47.432 MHz, 1 run		112.980 MHz	11080 MHz								323.9160 MHz	477.1200 MHz
ticlang-bare-os-32mhz				49.432 MHz	49.532 MHz								76.432 MHz
ticlang-bare-os-400b			Not for this chip
ticlang-bare-os-ocsram																		877.5400 MHz
ticlang-bare-os-tcm																		870.4400 MHz
ticlang-bare-oz	102.280 MHz, 1 run	41.632 MHz, 1 run		100.880 MHz	98.580 MHz	57.548 MHz, 1 run	94.748 MHz						285.2160 MHz	438.1200 MHz	21196 MHz
ticlang-bare-oz-32mhz				44.232 MHz	44.332 MHz								69.232 MHz
ticlang-bare-oz-4mhz		5.984 MHz, 1 run
ticlang-bare-oz-ocsram																		850400 MHz
ticlang-bare-oz-tcm																		854.2400 MHz
ticlang-freertos-o3																718.9160 MHz
ticlang-freertos-oz																354.1160 MHz
unmodified-ticlang-c2000-o4								Not for this chip

Sources

am13e230/bench/coremark/main.c, am13e230/bench/coremark/ti_printf.c, am13e230/tools/build_lowclk.sh, am62l-evm/README.md, cc1354p10/baremetal/README.md, cc2340r5/baremetal/README.md, mspm0l1306/zephyr/README.md

CoreMark per MHz iter/s/MHz, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X	LP-AM263P	TMDS62LEVM (AM62L EVM)
armgnu14-bare-o2										2.7996 MHz
armgnu14-bare-o3										2.8796 MHz
armgnu14-bare-o3-48mhz										2.8748 MHz, 1 run
armgnu14-bare-os										2.2896 MHz
c28patch-ti-c2000-o2-flash								0.809120 MHz, 1 run
c28patch-ti-c2000-o2-ram								0.951120 MHz, 1 run
c28patch-ti-c2000-o4-flash								0.894120 MHz, 1 run
c28patch-ti-c2000-o4-ram								1.04120 MHz, 1 run
c28patch-ti-c2000-o4s0-flash								0.888120 MHz, 1 run
c28patch-ti-c2000-o4s0-ram								1.03120 MHz, 1 run
c28patch-ti-c2000-o4s5-flash								1.18120 MHz, 1 run									1.06200 MHz, 1 run
c28patch-ti-c2000-o4s5-ram								1.29120 MHz, 1 run									1.29200 MHz, 1 run
gcc-bare-o2	1.7680 MHz, 1 run									2.7596 MHz
gcc-bare-o2-48mhz									1.9248 MHz
gcc-bare-o2-m33											2.8150 MHz, 1 run
gcc-bare-o2-m33-gcc12											2.72150 MHz, 1 run
gcc-bare-o2-riscv											3.35150 MHz, 1 run	2.77160 MHz
gcc-bare-o3		1.7132 MHz								2.9496 MHz
gcc-bare-o3-48mhz										2.9448 MHz
gcc-bare-o3-m33											2.89150 MHz, 1 run
gcc-bare-o3-m33-gcc12											2.92150 MHz, 1 run
gcc-bare-o3-m33-gcc12-ram											2.92150 MHz, 1 run
gcc-bare-o3-m33-ram											2.89150 MHz, 1 run
gcc-bare-o3-riscv											3.3150 MHz, 1 run	2.73160 MHz
gcc-bare-o3-riscv-ram											3.29150 MHz, 1 run
gcc-bare-o3-unroll	1.8280 MHz, 1 run
gcc-bare-o3-unroll-48mhz									2.0348 MHz
gcc-bare-o3-unroll-reset									2.054 MHz
gcc-bare-o3unroll		1.8332 MHz
gcc-bare-o3x-m33											3.41150 MHz, 1 run
gcc-bare-o3x-riscv											3.91150 MHz, 1 run
gcc-bare-os	1.2480 MHz, 1 run									2.4396 MHz
gcc-bare-os-48mhz									1.4948 MHz
gcc-bare-os-m33											2.26150 MHz, 1 run
gcc-bare-os-m33-gcc12											2.43150 MHz, 1 run
gcc-bare-os-riscv											2.53150 MHz, 1 run	2.13160 MHz
gcc-vendor-o3												2.82160 MHz
gcc-zephyr-o2	1.6380 MHz, 1 run	1.6832 MHz, 1 run		1.5780 MHz, 1 run		1.5448 MHz, 1 run			1.8448 MHz, 1 run	2.6296 MHz, 1 run
gcc-zephyr-o2unroll	1.7580 MHz, 1 run	1.8732 MHz, 1 run		1.6980 MHz, 1 run		1.7548 MHz, 1 run
gcc-zephyr-o3	1.7180 MHz, 1 run	1.7732 MHz, 1 run		1.6380 MHz, 1 run		1.6648 MHz, 1 run			1.9248 MHz, 1 run
gcc-zephyr-os	1.1980 MHz, 1 run	1.3132 MHz, 1 run		1.1580 MHz, 1 run		1.2248 MHz, 1 run			1.4748 MHz, 1 run	2.3396 MHz, 1 run
gcc12.2-zephyr-core1-o2-ddr																			3.211250 MHz, 1 run
gcc15.3-linux-o2-1t																			3.41250 MHz, 1 run
starmclang-bare-o3-48mhz									1.8348 MHz
starmclang-bare-omax-48mhz									2.448 MHz
stgcc14-bare-o3-48mhz									1.9648 MHz
stgcc14-bare-o3-unroll-48mhz									2.0648 MHz
ti-port-o3-ocsram																		3.66400 MHz
ticlang-bare-o2	1.5380 MHz, 1 run	1.732 MHz, 1 run		1.5380 MHz	1.5380 MHz								2.36160 MHz
ticlang-bare-o2-32mhz				1.7132 MHz	1.7132 MHz								3.0332 MHz
ticlang-bare-o3	2.380 MHz	2.4532 MHz, 1 run	Not for this chip	2.2880 MHz	2.2980 MHz	2.3148 MHz, 1 run	4.1748 MHz						3.4160 MHz	4.36200 MHz	4.4496 MHz
ticlang-bare-o3-100mhz														4.42100 MHz, 1 run
ticlang-bare-o3-32mhz	2.4632 MHz, 1 run			2.4632 MHz	2.4632 MHz								4.3832 MHz
ticlang-bare-o3-4mhz		2.634 MHz, 1 run
ticlang-bare-o3-ocsram																		3.66400 MHz
ticlang-bare-o3-tcm																		3.74400 MHz
ticlang-bare-os	1.4380 MHz, 1 run	1.4832 MHz, 1 run		1.4180 MHz	1.3880 MHz								2.02160 MHz
ticlang-bare-os-32mhz				1.5432 MHz	1.5532 MHz								2.3932 MHz
ticlang-bare-oz	1.2880 MHz, 1 run	1.332 MHz, 1 run		1.2680 MHz	1.2380 MHz	1.248 MHz, 1 run	1.9748 MHz						1.78160 MHz		2.296 MHz
ticlang-bare-oz-32mhz				1.3832 MHz	1.3932 MHz								2.1632 MHz
ticlang-bare-oz-4mhz		1.54 MHz, 1 run
ticlang-freertos-o3																4.49160 MHz
ticlang-freertos-oz																2.21160 MHz

Sources

am13e230/bench/coremark/main.c, am13e230/tools/build_lowclk.sh, am62l-evm/README.md, cc1354p10/baremetal/README.md, cc2340r5/baremetal/README.md, mspm0l1306/zephyr/README.md

Dhrystone 2.1 DMIPS/MHz, higher is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X	LP-AM263P	SK-TDA4VM
armgnu14-bare-o3									0.8496 MHz, 1 run
gcc-bare-o3									0.83596 MHz, 1 run
gcc-bare-o3-noinline-m33										1.32150 MHz
gcc-bare-o3-noinline-m33-gcc12										1.32150 MHz
gcc-bare-o3-noinline-riscv										1.56150 MHz
gcc-linux-o2-noinline-mcpu																	4.61 to 4.622000 MHz
gcc-linux-o3-inline																	5.26 to 5.682000 MHz
gcc-linux-o3-noinline																	4.832000 MHz
gcc-linux-o3-noinline-mcpu																	4.682000 MHz
gcc-linux-os-noinline																	3.09 to 3.12000 MHz
gcc12-os								0.47948 MHz
linux
ti-c2000-o3-inline-ram							0.477120 MHz, 1 run
ti-c2000-o3-noinline-flash							0.362120 MHz, 1 run								0.279200 MHz, 1 run
ti-c2000-o3-noinline-ram							0.442120 MHz, 1 run								0.444200 MHz, 1 run
ti-example-o3																1.94400 MHz, 1 run
ticlang-bare-o3-inline					0.86648 MHz	1.4948 MHz					1.08160 MHz	1.34200 MHz, 1 run	1.6296 MHz			1.88400 MHz
ticlang-bare-o3-noinline			0.65280 MHz	0.65180 MHz	0.60248 MHz	1.248 MHz					0.876160 MHz	1.05200 MHz, 1 run	1.3296 MHz			1.5400 MHz
ticlang-bare-o3-noinline-32mhz											1.2232 MHz
ticlang-bare-os-noinline											0.87160 MHz	1200 MHz, 1 run				1.45400 MHz
ticlang-bare-os-noinline-32mhz											1.2132 MHz
ticlang-bare-oz-noinline					0.27548 MHz	0.47848 MHz							0.5696 MHz
ticlang-freertos-o3-inline														1.65160 MHz
ticlang-freertos-o3-noinline														1.34160 MHz
ticlang-freertos-os-noinline														1.32160 MHz
(one way)	0.73132 MHz, 1 run	Not for this chip						0.67248 MHz

Sources

am62l-evm/README.md, cc35x1/dhrystone/build.sh

Embench programs that fit and pass of 19, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1
gcc-linux-a72
gcc-o2-m33											19150 MHz
gcc-o2-riscv											19150 MHz
gcc-os-m33											19150 MHz
gcc-os-riscv											19150 MHz
ti-c2000-o2-flash								15120 MHz, 1 run
ticlang-freertos-o2															Not for this chip
ticlang-freertos-o2-retry
ticlang-freertos-os															Not for this chip
ticlang-o2			724 MHz, 1 run			1848 MHz	1948 MHz					19160 MHz	19200 MHz	1996 MHz, 1 run
ticlang-os			624 MHz, 1 run			1848 MHz	1948 MHz					19160 MHz	19200 MHz	1996 MHz, 1 run
(one way)	1880 MHz, 1 run	1332 MHz, 1 run	924 MHz	1880 MHz	1880 MHz				1948 MHz	1996 MHz, 1 run

Sources

am13e230/data/logs/tb/, mspm0l1306/third-party/README.md

Reaction time

Interrupt latency, in time

Best case, every board here reaches its handler in under 1 µs except the ESP32-C6-DevKitC-1-N8 (Espressif).

Time from a pin or timer signal to the first line of your handler, no OS. Cycle counts flatter slow clocks, so the bars show time, calculated from measured cycles and the clock.

TI boardsother vendors' boardsyour board

LAUNCHXL-F28P65Xcode in RAM

150 ns

Raspberry Pi Pico 2 H

173 ns

FRDM-MCXA153 (NXP)

177 ns

LP-MSPM33C321A

200 ns

LP-EM-CC2755P10

219 ns

LAUNCHXL-F2800137

250 ns

LP-MSPM0G3507

362 ns

LP-MSPM0G3519

388 ns

LP-MSPM0G5187

450 ns

NUCLEO-U083RC (ST)

458 ns

LP-CC1354P10-1

479 ns

LP-AM263P

488 ns

LP-EM-CC2340R5

583 ns

LP-MSPM0L1306

781 ns

LP-MSPM0C1104

958 ns

ESP32-C6-DevKitC-1-N8 (Espressif)under FreeRTOS

1069 ns

Times are calculated from measured cycles and the clock; hover a time for both. No OS, code in flash unless marked (LAUNCHXL-F28P65X ran from RAM). Zephyr adds cycles on every board that ran it; see the rows below.

All rows

Interrupt pend to first line of handler cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-EM-CC2755P10	LAUNCHXL-F28P65X	LP-AM263P
bare	29 to 3680 MHz, 1 run	25 to 2832 MHz, 1 run	23 to 2524 MHz	31 to 3480 MHz	36 to 4080 MHz	28 to 2948 MHz, 1 run	23 to 2948 MHz	30 to 32120 MHz	22 to 2648 MHz	17 to 2996 MHz, 1 run			32160 MHz	21 to 3596 MHz		195 to 227400 MHz
bare-32mhz													2032 MHz
bare-external-wire4
bare-fiq																144 to 205400 MHz
bare-flash															36200 MHz
bare-m33											26150 MHz
bare-m33-gcc12											26150 MHz
bare-ram															30200 MHz
bare-riscv											45150 MHz
freertos		38 to 3932 MHz, 1 run				39 to 20248 MHz	2748 MHz					171 to 175160 MHz
pend-sw7
zephyr	163 to 18180 MHz, 1 run	119 to 13732 MHz, 1 run		157 to 18580 MHz, 1 run		48 to 6048 MHz, 1 run			108 to 12948 MHz, 1 run

Sources

cc1354p10/baremetal/README.md, cc2340r5/baremetal/README.md, mspm0l1306/freertos/README.md, mspm0l1306/zephyr/README.md

Control loops

Control loop: ADC trigger to new PWM value

From 1.33 µs to 448 µs across the boards, in float code: hardware math where the chip has it, software math where it does not, each bar says which. The Cortex-M0+ boards run floating point in software, so none gets under 88.5 µs even with a math accelerator for sine and cosine. TI's fixed-point IQmath library brings the MSPM0 G3507 to 12.3 µs.

One field-oriented control step, timed from the ADC finishing to the new PWM value.

hardware math unitsoftware mathyour board

0.1 µs1 µs10 µs100 µs1000 µs

LP-AM13E230

1.33 µs

LAUNCHXL-F2800137

2.58 µs

LP-MSPM33C321A

5.03 µs

LP-EM-CC2755P10

10.5 µs

LP-CC1354P10-1

21.4 µs

LP-MSPM0G3507

88.5 µs

LP-MSPM0G3519

89.3 µs

LP-MSPM0G5187

181 µs

NUCLEO-U083RC (ST)

299 µs

LP-EM-CC2340R5

337 µs

LP-MSPM0L1306

448 µs

One run per board: code in flash, timed from the ADC end-of-conversion interrupt, where the board has that run; otherwise its fastest plain run. Early-trigger, RAM-code, lockstep and IQmath runs are left out; IQmath is in the rows below. AM263P ran only with code in RAM, so it is not drawn. Log scale: each gridline is 10 times slower. Hover a time for the clock and build.

All rows

PWM or timer trigger to first line of ISR us, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P
ctrl-systick-o2
early-flash							0.402 to 0.433120 MHz
early-flash-ramfunc							0.368 to 0.417120 MHz
early-ram							0.392 to 0.408120 MHz
eoc-flash		1.86 to 1.9732 MHz			4.21 to 4.3148 MHz	5.84 to 15.748 MHz	0.555 to 0.592120 MHz	1.6948 MHz	0.538 to 0.55160 MHz		3.77 to 3.7996 MHz
eoc-flash-o3					4.22 to 4.3148 MHz	5.84 to 15.548 MHz		1.6948 MHz	0.713 to 0.725160 MHz
eoc-flash-os					4.2 to 4.2548 MHz
eoc-flash-ramfunc							0.535 to 0.583120 MHz
eoc-ram								1.58 to 1.648 MHz
hw-tmu-eoc-flash										0.365 to 0.38200 MHz, 1 run
hw-tmu-eoc-ocsram													0.632400 MHz
hw-tmu-eoc-ocsram-200k-alone													0.637 to 0.638400 MHz
hw-tmu-eoc-ocsram-200k-beside-bytecopy3													0.638400 MHz
hw-tmu-eoc-ocsram-200k-beside-coremark3													0.638400 MHz
hw-tmu-eoc-ocsram-200k-beside-nearcopy3													0.637 to 0.638400 MHz
hw-tmu-eoc-ocsram-200k-beside-wordcopy3													0.638400 MHz
hw-tmu-eoc-ocsram-lockstep													0.631 to 0.632400 MHz
hw-tmu-eoc-tcm-200k-alone													0.592400 MHz
hw-tmu-eoc-tcm-200k-beside-nearcopy3													0.593400 MHz
o3				2.69 to 2.7180 MHz
sdk-eoc-ocsram													0.636 to 0.637400 MHz
sdk-eoc-ocsram-lockstep													0.637400 MHz
sw-eoc-flash										0.397 to 0.41200 MHz, 1 run
sw-eoc-ocsram													0.654 to 0.655400 MHz
sw-eoc-ocsram-lockstep													0.661 to 0.663400 MHz
ticlang-freertos-o2												Not for this chip
ticlang-freertos-o3												Not for this chip
(one way)	0.86 to 0.9580 MHz, 1 run	1.94 to 1.9732 MHz, 1 run	1.11 to 1.2180 MHz	2.7 to 2.7480 MHz			0.55 to 0.58120 MHz

Sources

Run records are in the notebook, not this copy.

Trigger, ADC, control step, new PWM value (mean) us, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P
ctrl-systick-o2
hw	88.5 to 90.480 MHz, 1 run	n/a	89.3 to 9180 MHz	n/a			2.4120 MHz	n/a
hw-early-flash							2.45 to 2.49120 MHz
hw-early-flash-ramfunc							2.15 to 2.2120 MHz
hw-early-ram							2.21 to 2.23120 MHz
hw-eoc-flash							2.58 to 2.62120 MHz
hw-eoc-flash-ramfunc							2.32 to 2.37120 MHz
hw-tmu-eoc-flash										1.33 to 1.69200 MHz, 1 run
hw-tmu-eoc-ocsram													1.24400 MHz
hw-tmu-eoc-ocsram-200k-alone													1.23400 MHz
hw-tmu-eoc-ocsram-200k-beside-bytecopy3													1.23400 MHz
hw-tmu-eoc-ocsram-200k-beside-coremark3													1.23400 MHz
hw-tmu-eoc-ocsram-200k-beside-nearcopy3													1.23400 MHz
hw-tmu-eoc-ocsram-200k-beside-wordcopy3													1.23400 MHz
hw-tmu-eoc-ocsram-lockstep													1.24400 MHz
hw-tmu-eoc-tcm-200k-alone													1.19400 MHz
hw-tmu-eoc-tcm-200k-beside-nearcopy3													1.19400 MHz
iq-hw	12.3 to 12.680 MHz		12.4 to 12.780 MHz
iq-sw	24.3 to 24.680 MHz	55.5 to 55.632 MHz, 1 run	24.4 to 24.680 MHz	26.3 to 26.480 MHz
microstep-sw										1.23 to 1.77200 MHz, 1 run
microstep-tmu										0.691 to 1.09200 MHz, 1 run
sdk-eoc-ocsram													1.45400 MHz
sdk-eoc-ocsram-lockstep													1.46400 MHz
sw	171.9 to 177.780 MHz, 1 run	401.6 to 413.132 MHz, 1 run	179.8 to 182.480 MHz	181.1 to 183.480 MHz			3.7120 MHz
sw-early-ram							3.46 to 3.51120 MHz
sw-eoc-flash		448.1 to 54032 MHz			336.6 to 403.448 MHz	21.4 to 24.848 MHz	4.1 to 4.18120 MHz	298.8 to 319.648 MHz	5.17 to 5.74160 MHz	4.74 to 4.79200 MHz, 1 run	10.5 to 1396 MHz
sw-eoc-flash-o3					338.4 to 408.148 MHz	21.4 to 24.348 MHz		299.3 to 319.848 MHz	5.03 to 5.47160 MHz		10.5 to 12.996 MHz
sw-eoc-flash-os					336.6 to 403.448 MHz						10.5 to 12.996 MHz
sw-eoc-ocsram													2.06400 MHz
sw-eoc-ocsram-lockstep													2.07400 MHz
sw-eoc-ram								272.3 to 289.148 MHz
sw-o3				180.8 to 183.580 MHz
ticlang-freertos-o2												Not for this chip
ticlang-freertos-o3												Not for this chip

Sources

am13e230/stepper-uln2003/microstep/

One field-oriented control step cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
ctrl-systick-o2
fpu-flash										308.4150 MHz
fpu-flash-gcc12										314.2150 MHz
fpu-flash-o3										308.4150 MHz
fpu-flash-o3-gcc12										314.2150 MHz
fpu-flash-os										306.1150 MHz
fpu-flash-os-gcc12										313.6150 MHz
hw	6.4k80 MHz, 1 run	n/a	6.5k80 MHz				194.1120 MHz	n/a	n/a
hw-flash							220.6120 MHz
hw-flash-ramfunc							195.3120 MHz
hw-tmu-flash												117200 MHz
hw-tmu-ocsram														160.3 to 833400 MHz
iq-hw	918.280 MHz
iq-sw	1.9k80 MHz	1.7k32 MHz, 1 run
iq-sw-ram		1.7k32 MHz
q15-rtcesl									44896 MHz, 1 run
sdk-ocsram														247.4 to 263400 MHz
sw	14k80 MHz, 1 run	13k32 MHz, 1 run	14k80 MHz	14k80 MHz			348.9120 MHz
sw-flash		13k32 MHz			14k48 MHz	671.448 MHz	400.2120 MHz	12k48 MHz	5.4k96 MHz, 1 run		531.1160 MHz, 1 run	655.6200 MHz	593.796 MHz
sw-flash-32mhz											489.132 MHz, 1 run
sw-flash-o3		13k32 MHz			14k48 MHz	671.348 MHz		12k48 MHz	5.3k96 MHz, 1 run				593.796 MHz
sw-flash-o3-riscv										4.5k150 MHz
sw-flash-o3-stgcc14								13k48 MHz
sw-flash-os		13k32 MHz			14k48 MHz								600.796 MHz
sw-flash-os-riscv										4.5k150 MHz
sw-flash-riscv										4.5k150 MHz
sw-flash-stgcc14								13k48 MHz
sw-ocsram														412 to 789400 MHz
sw-ram		13k32 MHz						11k48 MHz
sw-ram-stgcc14								12k48 MHz
ti-aci-chain												555200 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Digital-power PID loop, trigger to duty write (mean) cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A
float-flash	1.8k to 1.9k80 MHz, 1 run	1.4k to 1.8k32 MHz	1.8k to 1.9k80 MHz	2k to 2.1k48 MHz	396.3 to 42048 MHz	194 to 198120 MHz, 1 run	235.3 to 278160 MHz
float-flash-o3					379.8 to 40148 MHz		230.2 to 258160 MHz
q15-flash	280 to 30080 MHz, 1 run	227.7 to 24832 MHz	301.3 to 32180 MHz	410 to 43348 MHz	382.1 to 40948 MHz	258.4 to 263120 MHz, 1 run	184.3 to 209160 MHz
q15-flash-o3					369.9 to 39448 MHz		174.2 to 194160 MHz

Sources

cc2755p10/RESULTS-cc2755p10.md

Loop rate with 20 percent of the period free (computed, superseded) kHz, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137
float-flash	34 calc.	14.4 calc.	17.8 calc.		460.8 calc.
q15-flash	205 calc.	99.3 calc.	85 calc.		351.9 calc.

Sources

Run records are in the notebook, not this copy.

Control loop worst case, trigger to new PWM value, as a percent of a 100 kHz period %, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
ctrl-systick-o2
hw	903.180 MHz		909.680 MHz				24.2120 MHz
hw-early-flash							24.9120 MHz
hw-early-flash-ramfunc							22120 MHz
hw-early-ram							22.2120 MHz
hw-eoc-flash							26.2120 MHz
hw-eoc-flash-ramfunc							23.7120 MHz
hw-o3	912.580 MHz		901.480 MHz
hw-tmu-eoc-flash										16.9200 MHz, 1 run
hw-tmu-eoc-ocsram												10.7400 MHz
hw-tmu-eoc-ocsram-200k-alone												10.5400 MHz
hw-tmu-eoc-ocsram-200k-beside-bytecopy3												10.5400 MHz
hw-tmu-eoc-ocsram-200k-beside-coremark3												10.5400 MHz
hw-tmu-eoc-ocsram-200k-beside-nearcopy3												13.1 to 13.5400 MHz
hw-tmu-eoc-ocsram-200k-beside-wordcopy3												10.5400 MHz
hw-tmu-eoc-ocsram-lockstep												10.7400 MHz
hw-tmu-eoc-tcm-200k-alone												6.3400 MHz
hw-tmu-eoc-tcm-200k-beside-nearcopy3												6.3400 MHz
iq-hw	125.580 MHz		125.980 MHz
iq-sw	245.680 MHz	555.932 MHz, 1 run	24680 MHz	263.880 MHz
sdk-eoc-ocsram												11.8400 MHz
sdk-eoc-ocsram-lockstep												11.8400 MHz
sw	2,318.980 MHz	4,131.232 MHz, 1 run	1,819.980 MHz	2,330.580 MHz			37.1120 MHz
sw-early-ram							35.1120 MHz
sw-eoc-flash		5,399.732 MHz			4,033.548 MHz	248.148 MHz	41.8120 MHz	3,192.548 MHz	57.4160 MHz	47.9200 MHz, 1 run	12996 MHz
sw-eoc-flash-o3					4,081.548 MHz	243.148 MHz		3,197.748 MHz	54.8160 MHz		129.496 MHz
sw-eoc-flash-os					4,033.548 MHz						12996 MHz
sw-eoc-ocsram												13.7 to 14400 MHz
sw-eoc-ocsram-lockstep												14400 MHz
sw-eoc-ram								2,890.448 MHz
sw-o3	2,322.180 MHz		1,774.980 MHz	2,324.880 MHz

Sources

Run records are in the notebook, not this copy.

Control loop jitter, worst minus best trigger to new PWM value us, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
ctrl-systick-o2
hw	7.680 MHz		7.1680 MHz				0.092120 MHz
hw-early-flash							0.108120 MHz
hw-early-flash-ramfunc							0.083120 MHz
hw-early-ram							0.067120 MHz
hw-eoc-flash							0.108120 MHz
hw-eoc-flash-ramfunc							0.083120 MHz
hw-o3	7.6780 MHz		7.4680 MHz
hw-tmu-eoc-flash										0.395200 MHz, 1 run
hw-tmu-eoc-ocsram												0.93400 MHz
hw-tmu-eoc-ocsram-200k-alone												0.905400 MHz
hw-tmu-eoc-ocsram-200k-beside-bytecopy3												0.905400 MHz
hw-tmu-eoc-ocsram-200k-beside-coremark3												0.905400 MHz
hw-tmu-eoc-ocsram-200k-beside-nearcopy3												1.41 to 1.5400 MHz
hw-tmu-eoc-ocsram-200k-beside-wordcopy3												0.905400 MHz
hw-tmu-eoc-ocsram-lockstep												0.93400 MHz
hw-tmu-eoc-tcm-200k-alone												0.09 to 0.095400 MHz
hw-tmu-eoc-tcm-200k-beside-nearcopy3												0.095 to 0.1400 MHz
iq-hw	0.480 MHz		0.3580 MHz
iq-sw	0.31280 MHz	0.34432 MHz, 1 run	0.26380 MHz	0.18880 MHz
sdk-eoc-ocsram												0.91400 MHz
sdk-eoc-ocsram-lockstep												0.91400 MHz
sw	157.680 MHz	246.232 MHz, 1 run	106.780 MHz	157.280 MHz			0.458120 MHz
sw-early-ram							0.483120 MHz
sw-eoc-flash		370.932 MHz			272.148 MHz	3.7548 MHz	0.55120 MHz	176.248 MHz	2.79160 MHz	2.52200 MHz, 1 run	3.9496 MHz
sw-eoc-flash-o3					277.148 MHz	3.2548 MHz		177.348 MHz	2.74160 MHz		3.7596 MHz
sw-eoc-flash-os					271.648 MHz						3.8596 MHz
sw-eoc-ocsram												0.89 to 0.95400 MHz
sw-eoc-ocsram-lockstep												0.95 to 0.96400 MHz
sw-eoc-ram								160.348 MHz
sw-o3	157.880 MHz		10280 MHz	156.380 MHz

Sources

Run records are in the notebook, not this copy.

Highest control-loop trigger rate with no missed period in 1,000 kHz, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-EM-CC2755P10
ctrl-systick-o2
hw	10.9 to 1180 MHz		10.880 MHz				336.1120 MHz
hw-early-flash							297120 MHz
hw-early-flash-ramfunc							336.1120 MHz
hw-early-ram							334.3120 MHz
hw-eoc-flash							301.5120 MHz
hw-eoc-flash-ramfunc							336.1120 MHz
hw-o3	10.980 MHz		10.980 MHz
iq-hw	70.880 MHz		70.380 MHz
iq-sw	38.180 MHz	16.932 MHz	38 to 38.380 MHz	37.280 MHz
sw	4.26 to 4.3180 MHz	1.83 to 1.8532 MHz	4.2980 MHz	4.2880 MHz			231.7120 MHz
sw-early-ram							231.7120 MHz
sw-eoc-flash					2.42 to 2.4548 MHz	46.248 MHz	209.4120 MHz	178160 MHz	90.2 to 91.296 MHz
sw-eoc-flash-gcc14
sw-eoc-flash-o3					2.43 to 2.4548 MHz	45.848 MHz			93.896 MHz
sw-eoc-flash-os					2.48 to 2.548 MHz				90.9 to 91.196 MHz
sw-o3	4.26 to 4.380 MHz	1.8332 MHz	4.24 to 4.2880 MHz	4.26 to 4.2980 MHz

Sources

Run records are in the notebook, not this copy.

Control-loop CPU load, interrupt entry to exit, as a percent of a 100 kHz period %, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-EM-CC2755P10
ctrl-systick-o2
hw	894.2 to 928.180 MHz		900.2 to 934.580 MHz				26.4 to 33.3120 MHz
hw-early-flash							29.9 to 38.1120 MHz
hw-early-flash-ramfunc							27.1 to 34.3120 MHz
hw-early-ram							26.5 to 33.4120 MHz
hw-eoc-flash							29.7 to 37.7120 MHz
hw-eoc-flash-ramfunc							27 to 34.2120 MHz
hw-o3	901.5 to 934.680 MHz		892.4 to 927.580 MHz
iq-hw	132.1 to 151.180 MHz		133.3 to 153.880 MHz
iq-sw	251.8 to 270.580 MHz	573.5 to 611.932 MHz	252.6 to 271.680 MHz	254.8 to 272.680 MHz
sw	1,949.8 to 2,337.580 MHz	4,504.8 to 5,457.832 MHz	1,952.3 to 2,343.280 MHz	1,951.1 to 2,339.980 MHz			39.6 to 46.8120 MHz
sw-early-ram							39.5 to 46.8120 MHz
sw-eoc-flash					3,380.3 to 4,09448 MHz	180.4 to 211.548 MHz	43.7 to 52.1120 MHz	59.5 to 65.3160 MHz	76.6 to 99.396 MHz
sw-eoc-flash-gcc14
sw-eoc-flash-o3					3,374.548 MHz	180 to 203.848 MHz			76.5 to 99.196 MHz
sw-eoc-flash-os					3,326.8 to 4,009.248 MHz				76.5 to 97.796 MHz
sw-o3	1,951.5 to 2,343.480 MHz	4,517.6 to 5,470.932 MHz	1,962.2 to 2,351.180 MHz	1,931.6 to 2,336.480 MHz

Sources

Run records are in the notebook, not this copy.

Digital-power loop worst case, trigger to duty write, as a percent of a 100 kHz period %, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A
float-flash	235.580 MHz	554.432 MHz	236.880 MHz	437.348 MHz	87.548 MHz	16.5120 MHz, 1 run	17.4160 MHz
float-flash-o3					83.548 MHz		16.1160 MHz
q15-flash	39.480 MHz	77.532 MHz	40.180 MHz	88.548 MHz	85.248 MHz	21.9120 MHz, 1 run	13.1160 MHz
q15-flash-o3					8248 MHz		12.1160 MHz

Sources

cc2755p10/RESULTS-cc2755p10.md

Digital-power loop jitter, worst minus best trigger to duty write cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LP-MSPM33C321A
float-flash	24780 MHz	51132 MHz	25880 MHz	26848 MHz	3048 MHz	0.269160 MHz
float-flash-o3					2448 MHz	0.175160 MHz
q15-flash	2380 MHz	2132 MHz	2180 MHz	2348 MHz	3148 MHz	0.156160 MHz
q15-flash-o3					2748 MHz	0.125160 MHz

Sources

cc2755p10/RESULTS-cc2755p10.md

Highest digital-power loop rate with no missed period in 1,000 kHz, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A
float-flash	39.7 to 4080 MHz	17.332 MHz	39.680 MHz	21.8 to 2248 MHz	117.148 MHz	439.6120 MHz	645.2160 MHz
float-flash-o3				22.2 to 22.448 MHz	121.248 MHz		655.7160 MHz
float-ram						500120 MHz
q15-flash	190.980 MHz	92.832 MHz	190.980 MHz	97.448 MHz	120.648 MHz	367120 MHz	680.9160 MHz
q15-flash-o3				113.548 MHz	123.148 MHz		720.7160 MHz
q15-ram						431.6120 MHz

Sources

cc2755p10/RESULTS-cc2755p10.md

Digital-power loop CPU load, interrupt entry to exit, as a percent of a 100 kHz period %, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A
float-flash	237.2 to 26180 MHz	471.2 to 606.232 MHz	236.8 to 260.980 MHz	430.5 to 477.748 MHz	48.5 to 70.648 MHz	19.7 to 26.9120 MHz	18.7 to 25.6160 MHz
float-flash-o3				427.9 to 472.748 MHz	47.2 to 7048 MHz		18.2 to 24.9160 MHz
float-ram						17.3 to 23.7120 MHz
q15-flash	41.4 to 6180 MHz	85.5 to 127.832 MHz	41.5 to 61.480 MHz	99.6 to 121.948 MHz	42.9 to 64.448 MHz	24.2 to 31.5120 MHz	13.3 to 20.2160 MHz
q15-flash-o3				90.8 to 109.848 MHz	41.6 to 62.148 MHz		12.7 to 19.6160 MHz
q15-ram						20.5 to 26.8120 MHz

Sources

cc2755p10/RESULTS-cc2755p10.md

Brushless motor speed at a duty, six-step from Hall sensors rpm, higher is better

Variant	LP-AM13E230
no-meter-retry-6x-10pct	0200 MHz, 1 run
rewired-6x-10pct-start	0200 MHz, 1 run
sine-hall-20pct-fwd	531 to 536200 MHz
sine-hall-30pct-fwd	894 to 907200 MHz
sine-hall-30pct-rev	892200 MHz, 1 run
six-step-hall-10pct-fwd	405 to 411200 MHz
six-step-hall-10pct-rev	376 to 391200 MHz
six-step-hall-20pct-fwd	924 to 929200 MHz
six-step-hall-20pct-rev	891 to 907200 MHz
six-step-hall-30pct-fwd	1,561200 MHz
six-step-hall-30pct-rev	1,446 to 1,457200 MHz
six-step-hall-40pct-fwd	2,312 to 2,339200 MHz
six-step-hall-40pct-rev	2,091200 MHz
six-step-hall-50pct-fwd	3,067200 MHz
six-step-hall-50pct-rev	2,728 to 2,805200 MHz
six-step-hall-60pct-fwd	3,632 to 3,700200 MHz
six-step-hall-60pct-rev	3,442200 MHz
speed-bench-firmware-trapezoid-retry	0200 MHz, 1 run

Sources

am13e230/bldc-drv8376/README.md

Hall sensor misalignment: largest deviation of a sector's width from one sixth of an electrical turn %, lower is better

Variant	LP-AM13E230
six-step-hall-30pct-fwd	2200 MHz
six-step-hall-30pct-rev	1.9200 MHz

Sources

am13e230/bldc-drv8376/README.md

Radio RAM

Radio stack RAM on the LP-EM-CC2340R5

6 of 11 builds fit, including every Bluetooth-only and Zigbee one. 5 do not link at 36 kB of RAM (Bluetooth + TI 15.4 at once; Thread end device, NCP build; Thread end device, CLI build; Thread router, CLI build; Thread router, NCP build). They link at 64 kB on the CC2340R53; none has run there.

Bar = RAM the stack needs. The chip has 36 kB.

RAM the stack needs, from the linker mapcalculated: what a build that did not link would need

Zigbee update client

36 kB, the chip's RAM

10.7 kB left

Thread radio only (RCP)

8.8 kB left

Zigbee end device

6.2 kB left

Bluetooth, TI's data stream

5.3 kB left

Bluetooth, TI's basic example

5.1 kB left

Zigbee router, logging on

799 B left

Bluetooth + TI 15.4 at once

2.1 kB over

Thread end device, NCP build

7.5 kB over

Thread end device, CLI build

10.0 kB over

Thread router, CLI build

16.0 kB over

Thread router, NCP build

20.7 kB over

Scale 0 to 58 kB; 1 kB = 1,024 bytes. These are TI's configurations as built here. No trimmed build was made, so none is a proven chip limit.

All rows

RAM left after linking a stack B, higher is better

Variant	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
ble-basic-ble	5.1 kB				81 kB
ble-basic-ble-occ-8k	8191 run
ble-basic-ble-text	5.1 kB
ble-broadcaster		Not for this chip
ble-central-b2b	2.6 kB
ble-data-stream	5.3 kB				81.2 kB
ble-desk-oled	3.4 kB1 run
ble-esphome-proxy				262.1 kB160 MHz, 1 run
ble-hci-controller					79.9 kB
ble-hid-remote	5.1 kB1 run
ble-nimble-peripheral				347.8 kB160 MHz, 1 run
ble-nordic-lbs-nus	4 kB1 run
ble-oad-dual-image	4.5 kB				80.5 kB
ble-proxy-zigbee-ed				247.4 kB160 MHz, 1 run
ble-simple-peripheral
ble-tooltag	1.2 kB1 run				76.8 kB1 run
ble-zephyr-ti-fork	1.7 kB
dmm-ble-ti154	Not for this chip 2.1 kB short				57.4 kB
esphome-minimal-wifi				303.3 kB160 MHz, 1 run
esphome-thread				335 kB160 MHz, 1 run
esphome-zigbee-ed				358.9 kB160 MHz, 1 run
none			n/a
proprietary-rftx	22.4 kB
thread-cli-ftd	Not for this chip 16 kB short	163.3 kB			66.5 kB1 run
thread-cli-mtd	Not for this chip 10 kB short	169.4 kB			72.5 kB1 run
thread-ftd				329 kB160 MHz, 1 run
thread-ncp-ftd	Not for this chip 20.7 kB short
thread-ncp-mtd	Not for this chip 7.5 kB short
thread-radio-cli	10.2 kB	191.4 kB			93.1 kB1 run
thread-rcp	8.8 kB	190.1 kB			91.7 kB1 run
ti154-collector	Not for this chip
ti154-sensor	1.6 kB1 run				127.2 kB1 run
wifi-esphome-api						74 kB to 86.7 kB160 MHz, 1 run
wifi-ota-example-plus-voice-fe						72.2 kB calc.
wisun-border-router
wisun-coap-node
wisun-router	Not for this chip
zigbee-ble-dmm-zr-light		222.2 kB
zigbee-coordinator
zigbee-end-device	6.2 kB				132.1 kB
zigbee-ota-client	10.7 kB				136.4 kB
zigbee-ota-client-router	5.3 kB
zigbee-router	799			360.1 kB160 MHz, 1 run	126.5 kB
zigbee-router-esphome	Not for this chip 1.3 kB short				124.4 kB1 run
zigbee-router-esphome-cc23xx	6831 run
zigbee-router-esphome-cc27xx					124.4 kB1 run
zigbee-router-ti-example					126.7 kB1 run
zigbee-router-ti-nolog	5.9 kB1 run				131.7 kB1 run
zigbee-zed-esphome	4 kB1 run				129.8 kB1 run
zigbee-zed-esphome-cc23xx	4 kB1 run
zigbee-zed-sleepy-temp		210.5 kB
zigbee-zed-ti-example		210.6 kB
zigbee-zed-ti-nolog	11.3 kB1 run				137.1 kB1 run

Sources

cc2340r5/thread/README.md, docs/experiments/2026-10-13-cc2340r5-esphome-fit.md, docs/experiments/2026-10-13-cc2755p10-esphome-fit.md, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, docs/experiments/2026-10-13-esphome-cc35xx-platform.md, esp32c6/proxy/README.md, esp32c6/proxy/frl-ble-proxy-zigbee.yaml, esp32c6/proxy/frl-esphome-minimal.yaml, esp32c6/proxy/frl-esphome-thread.yaml, esp32c6/proxy/frl-esphome-zigbee.yaml, esp32c6/zigbee/README.md

Flash used by a stack and role, from the linker map B, lower is better

Variant	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
ble-basic-ble	187.4 kB				205.3 kB
ble-basic-ble-occ-8k	208.6 kB1 run
ble-basic-ble-text	187.4 kB
ble-broadcaster		Not for this chip
ble-central-b2b	194 kB
ble-data-stream	185.9 kB				203.9 kB
ble-desk-oled	190.5 kB1 run
ble-esphome-proxy				1.4 MB160 MHz, 1 run
ble-hci-controller					189.4 kB
ble-hid-remote	186.5 kB1 run
ble-nimble-peripheral				640.3 kB160 MHz, 1 run
ble-nordic-lbs-nus	187.3 kB1 run
ble-oad-dual-image	192 kB				209.5 kB
ble-proxy-zigbee-ed				1.6 MB160 MHz, 1 run
ble-simple-peripheral		166 kB
ble-tooltag	193.5 kB1 run				211.9 kB1 run
dmm-ble-ti154					253.8 kB
esphome-minimal-wifi				816.2 kB160 MHz, 1 run
esphome-thread				748.4 kB160 MHz, 1 run
esphome-zigbee-ed				491.8 kB160 MHz, 1 run
none			n/a
proprietary-rftx	32.5 kB	32.2 kB
thread-cli-ftd	417.5 kB	397.9 kB			432.2 kB1 run
thread-cli-mtd	353.3 kB	335.4 kB			370.1 kB1 run
thread-ftd				1.2 MB160 MHz, 1 run
thread-ncp-ftd	404.4 kB
thread-ncp-mtd	342.5 kB
thread-radio-cli	96.1 kB	75.4 kB			118.2 kB1 run
thread-rcp	107.1 kB	86.6 kB			128.7 kB1 run
ti154-collector	Not for this chip	117 kB
ti154-sensor	90.4 kB1 run	111 kB			90.9 kB1 run
wifi-ble-prov-mqtt-chained						1.3 MB160 MHz, 1 run
wifi-esphome-api						1.2 MB160 MHz, 1 run
wisun-border-router		398.3 kB
wisun-coap-node		338.3 kB
wisun-router	Not for this chip	342.3 kB
zigbee-ble-dmm-zr-light		458.9 kB
zigbee-coordinator		219.9 kB
zigbee-end-device	215.2 kB				205.7 kB
zigbee-ota-client	216.8 kB				205.9 kB
zigbee-ota-client-router	274.7 kB
zigbee-router	275.7 kB	223.3 kB		512.9 kB160 MHz, 1 run	274.2 kB
zigbee-router-esphome	277.7 kB1 run				263.1 kB1 run
zigbee-router-esphome-cc23xx	277.8 kB1 run
zigbee-router-esphome-cc27xx					262.7 kB1 run
zigbee-router-ti-example					259.6 kB1 run
zigbee-router-ti-nolog	268.8 kB1 run				253.1 kB1 run
zigbee-zed-esphome	219.8 kB1 run				211.1 kB1 run
zigbee-zed-esphome-cc23xx	219.9 kB1 run
zigbee-zed-sleepy-temp		179.4 kB
zigbee-zed-ti-example		178.9 kB
zigbee-zed-ti-nolog	210.9 kB1 run				201.4 kB1 run
(one way)

Sources

cc3200-launchxl/README.md, docs/experiments/2026-10-13-cc2340r5-esphome-fit.md, docs/experiments/2026-10-13-cc2755p10-esphome-fit.md, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, docs/experiments/2026-10-13-esphome-cc35xx-platform.md, esp32c6/proxy/README.md, esp32c6/proxy/frl-ble-proxy-zigbee.yaml, esp32c6/proxy/frl-esphome-minimal.yaml, esp32c6/proxy/frl-esphome-thread.yaml, esp32c6/proxy/frl-esphome-zigbee.yaml, esp32c6/zigbee/README.md, ports/wifi-provisioning-ble/README.md

Time on air for one packet ms, lower is better

Variant	LP-MSPM0G5187	LAUNCHXL-F2800137
cc1407p-2-gfsk-38-4-kbps-200b-host-view	45.580 MHz
cc1407p-2-gfsk-50-kbps-high-performance-200b-host-view	35.280 MHz
cc1407p-2-gfsk-50-kbps-low-power-200b-host-view	35.280 MHz
cc1407p-2-gfsk-9-6-kbps-200b-host-view	178.480 MHz
cc1407p-general-purpose-gfsk-200b-host-view	35.180 MHz
cc1407p-ofdm-lr-16-67-ksps-dsss-6-200b-host-view	1,216.780 MHz
cc1407p-ofdm-lr-66-67-ksps-dsss-2-200b-host-view	107.480 MHz
cc1407p-ofdm-lr-8-33-ksps-dsss-6-200b-host-view	2,432.280 MHz
cc1407p-sun-o-qpsk-rate-mode-3-50-kbps-200b-host-view	56.580 MHz
cc1407p-wi-sun-fsk-mode-2b-200b-host-view	18.780 MHz
cc1407p-wi-sun-fsk-mode-4b-200b-host-view	10.180 MHz
cc1407p-wi-sun-ofdm-opt1-mcs6-200b-host-view	3.180 MHz
cc1407p-wi-sun-ofdm-opt2-mcs3-200b-host-view	6.9480 MHz
cc1407p-wi-sun-ofdm-opt2-mcs4-200b-host-view	5.5180 MHz
cc1407p-wi-sun-ofdm-opt2-mcs5-200b-host-view	4.9180 MHz
cc1407p-wi-sun-ofdm-opt2-mcs6-200b-host-view	4.1780 MHz
cc1407p-wi-sun-ofdm-opt3-mcs4-200b-host-view	8.3180 MHz
cc1407p-wi-sun-ofdm-opt3-mcs5-200b-host-view	6.9980 MHz
cc1407p-wi-sun-ofdm-opt3-mcs6-200b-host-view	5.5680 MHz
cc1407p-wi-sun-ofdm-opt4-mcs4-200b-host-view	13.880 MHz
cc1407p-wi-sun-ofdm-opt4-mcs5-200b-host-view	11.180 MHz
cc1407p-wi-sun-ofdm-opt4-mcs6-200b-host-view	8.3680 MHz
cc1407p-xlr-20-ksps-dsss-2-200b-host-view	344.380 MHz
none		n/a
(one way)

Sources

cc3200-launchxl/README.md

Charge per radio event at 3.0 V uC, lower is better

Variant	LP-EM-CC2340R5	LP-CC1354P10-1	CC3200-LAUNCHXL	LAUNCHXL-F2800137
none				n/a
(one way)

Sources

cc3200-launchxl/README.md

Average current at one event per 1000 ms, 3.0 V uA, lower is better

Variant	LP-EM-CC2340R5	LP-CC1354P10-1	CC3200-LAUNCHXL	LAUNCHXL-F2800137
none				n/a
(one way)

Sources

cc3200-launchxl/README.md

Payload throughput board to board kbps, higher is better

Variant	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-EM-CC35X1
ble-oad-236b	61.548 MHz, 1 run
none		n/a
wifi-tcp			12,111160 MHz, 1 run
wifi-udp			15,973160 MHz, 1 run
(one way)

Sources

cc2340r5/oad/logs/oad-run1.json, cc3200-launchxl/README.md, cc35x1/RESULTS-cc35x1.md

Distance at which packet error rate stays under 1% m, higher is better

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	CC3200-LAUNCHXL	LAUNCHXL-F2800137
cc1407p-gfsk50k-desk-ha
none					n/a
(one way)

Sources

cc3200-launchxl/README.md

Boot to joined network s, lower is better

Variant	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
ble-proxy-zigbee-ed				Not for this chip
esphome-minimal-wifi				Not for this chip
esphome-thread				1.2160 MHz
esphome-zigbee-ed				6.4160 MHz, 1 run
matter-thermostat-commission-alexa
matter-thermostat-commission-apple-home						1160 MHz, 1 run
matter-thermostat-commission-google-home
matter-thermostat-commission-home-assistant
matter-wifi-commission						5160 MHz, 1 run
none			n/a
thread-child				4.55 to 4.56160 MHz
thread-host-rcp-apple-tv		3.5 to 4.748 MHz
thread-host-rcp-restart		3.4 to 3.748 MHz
thread-leader-alone		6 to 101 run
thread-leader-fresh-image		81 run
thread-router-to-cc1354p10-leader	15.51 run
wifi-ble-provisioning						7.13160 MHz
wifi-ble-provisioning-mqtt
wifi-ble-provisioning-wrong-password
wifi-esphome-guest						5.3 to 16.9160 MHz
wifi-stored-credentials-reset
wifi-unknown-network-ble-fallback						Not for this chip
wifi-wpa2-2.4g
wifi-wpa2-5g						5.74160 MHz
wifi-xiaozhi-boot						6.5 to 6.7160 MHz
zigbee-end-device	151 run
zigbee-router	371 run			32.5160 MHz, 1 run
zigbee-router-esphome					71 run
zigbee-router-esphome-cc23xx	61 run
zigbee-router-esphome-cc27xx					5.7 to 6.3
zigbee-router-repeat	Not for this chip
zigbee-router-to-cc1354p10-coordinator	6.21 run
zigbee-zed-esphome-cc23xx	7 to 19
zigbee-zed-zha		648 MHz, 1 run
zigbee-zr-dmm-zha		1 to 6.648 MHz, 1 run

Sources

board-to-board/thread-two-radios-2026-10-06/README.md, board-to-board/zigbee-cc1354-cc2340/README.md, cc1354p10/thread/README.md, cc2340r5/zigbee/README.md, cc35x1/RESULTS-cc35x1.md, cc35x1/matter/COMPATIBILITY.md, cc35x1/matter/README.md, docs/experiments/2026-10-10-cc35x1-matter-thermostat.md, docs/experiments/2026-10-13-cc2755p10-esphome-fit.md, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, docs/experiments/2026-10-13-esphome-cc35xx-platform.md, esp32c6/proxy/frl-ble-proxy-zigbee.yaml, esp32c6/proxy/frl-esphome-minimal.yaml, esp32c6/proxy/frl-esphome-thread.yaml, esp32c6/zigbee/report_bench.py, ports/wifi-provisioning-ble/README.md

Command to state change, median of 20 ms, lower is better

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
ble-tooltag-event							1,415
ble-tooltag-event-e2e						1,641
cc1407p-tag-out-of-range-alarm-ha	3,010 (1 run)80 MHz
esphome-api-loopback							120160 MHz
esphome-home-assistant
matter-onoff-toggle							250 to 830160 MHz
matter-presence-radar-to-alexa
matter-presence-radar-to-apple-home
matter-presence-radar-to-google-home
matter-presence-radar-to-home-assistant
matter-thermostat-setpoint-alexa
matter-thermostat-setpoint-apple-home
matter-thermostat-setpoint-google-home
matter-thermostat-setpoint-home-assistant
mqtt-sensors-request-reply							167.4 to 1,206.2160 MHz
none				n/a
thermostat-uart-to-mqtt
zigbee-home-assistant		95 to 292 (1 run)
zigbee-home-assistant-esphome						87 (1 run)
zigbee-home-assistant-esphome-cc23xx		136 (1 run)
zigbee-home-assistant-esphome-cc23xx-router		97 (1 run)
zigbee-home-assistant-esphome-cc27xx						92.5 (1 run)
zigbee-home-assistant-repaired		5,000 to 7,000 (1 run)
zigbee-light-toggle					98.5 to 167160 MHz
zigbee-zha-read-sleepy			1,574 to 2,79948 MHz

Sources

cc2340r5/zigbee/README.md, cc35x1/matter/COMPATIBILITY.md, cc35x1/matter/README.md, cc35x1/mqtt-test/README.md, docs/experiments/2026-10-10-thermostat-mcu-only.md, docs/experiments/2026-10-13-cc2755p10-esphome-fit.md, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, docs/experiments/2026-10-13-esphome-cc35xx-platform.md, site/data/radio-demos.json

Commands that landed %, higher is better

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
ble-dmm-light-write			10048 MHz
ble-tooltag-event						100	100
ble-tooltag-two-tags							100
cc1407p-23-phy-profiles	10080 MHz
cc1407p-gfsk50k-ping-from-cc1407p	10080 MHz
cc1407p-parent-message-to-handset	10080 MHz
cc1407p-spi-bare-no-radio
cc1407p-spi-bare-no-radio-2026-10-24
cc1407p-wisun-fsk-1a-ping-from-cc1407p	10080 MHz
cc1407p-wisun-fsk-1b-ping-from-cc1407p	10080 MHz
cc1407p-wisun-fsk-2a-ping-from-cc1407p	4580 MHz
cc1407p-wisun-fsk-2b-ping-from-cc1407p	8580 MHz
cc1407p-wisun-fsk-3-ping-from-cc1407p	10080 MHz
cc1407p-wisun-fsk-4a-ping-from-cc1407p	2080 MHz
cc1407p-wisun-fsk-4b-ping-from-cc1407p	10080 MHz
cc1407p-wisun-fsk-5-ping-from-cc1407p	5580 MHz
esphome-api-loopback-24h							99.9160 MHz
mqtt-sensors-request-reply							100160 MHz
none				n/a
thread-ping-cc1354p10		100
thread-ping-cc2340r5			100
zigbee-home-assistant		100 (1 run)
zigbee-home-assistant-esphome-cc23xx		93
zigbee-home-assistant-esphome-cc23xx-router		100
zigbee-home-assistant-esphome-cc27xx						100
zigbee-home-assistant-repaired		100 (1 run)
zigbee-light-toggle					100160 MHz
zigbee-zha-read-sleepy			10048 MHz

Sources

board-to-board/thread-two-radios-2026-10-06/README.md, cc2340r5/zigbee/README.md, cc35x1/mqtt-test/README.md, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, mspm0g5187/radio/README.md, mspm0l1306/radio-host/README.md, site/data/radio-demos.json

Flash to advertising s, lower is better

Variant	LP-EM-CC2340R5	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
default-100ms-interval	1.41 run			0.91 run
hid-remote-from-dslite-start	11.91 run
interval-20ms	0.561 run
nimble-legacy-30ms			0.452 to 0.889160 MHz
none		n/a
nordic-lbs-nus-100ms-interval	0.17 to 0.69
(one way)					51 run

Sources

cc2755p10/ble/onair.py

BLE echo round trip, 20 bytes ms, lower is better

Variant	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-EM-CC2755P10
desk-oled-mac-bleak	45.9 to 134.548 MHz
none		n/a
nordic-uart-14-byte-lines	24.2
nordic-uart-19-byte-lines-100	23.9 to 40.348 MHz
(one way)	45		29.2

Sources

data/benchmark-suite.json, docs/experiments/2026-10-30-cc2340r5-desk-display.md

BLE write with response, round trip on the Mac clock (the board acts in its write callback before it answers) ms, lower is better

Variant	LP-EM-CC2340R5
nordic-lbs-led-1-byte	24.3
nordic-lbs-led-gpio-readback	29.9 to 59.648 MHz

Sources

Run records are in the notebook, not this copy.

BLE connect and discover services s, lower is better

Variant	LP-EM-CC2340R5	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1
data-stream-mac-bleak	0.551 run			0.55
nimble-bleprph			0.615 to 0.748160 MHz
none		n/a
nordic-lbs-nus-mac-bleak	Not for this chip
(one way)					0.521 run

Sources

cc35x1/README.md

Signal strength at the test spot dBm, higher is better

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8
ble-mac-1m					-47 to -41
cc1407p-foxhunt-beacon--12dbm-lowpa	-6280 MHz
cc1407p-foxhunt-beacon--20dbm-lowpa	-6580 MHz
cc1407p-foxhunt-beacon--30dbm-lowpa	-7880 MHz
cc1407p-foxhunt-beacon--6dbm-lowpa	-5680 MHz
cc1407p-foxhunt-beacon-0dbm-lowpa	-4680 MHz
cc1407p-foxhunt-beacon-10dbm-lowpa	-3580 MHz
cc1407p-foxhunt-beacon-14dbm-highpa	-3480 MHz
cc1407p-foxhunt-beacon-4dbm-lowpa	-4180 MHz
cc1407p-gfsk50k-915-to-cc1354p10	-14 to -1580 MHz
none				n/a
thread-peer-cc1354p10		-301 run
thread-peer-cc2340r5			-291 run
wifi-esphome-proxy					-391 run
zigbee-zed-parent

Sources

board-to-board/thread-two-radios-2026-10-06/README.md, cc1354p10/zigbee-sleepy-sensor/README.md, esp32c6/proxy/README.md, mspm0g5187/radio/README.md, site/data/radio-demos.json

Board to board round trip ms, lower is better

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC35X1
ble-from-cc2340r5			195.4 to 199.8
ble-to-cc1354p10		195.4 to 199.8
cc1407p-gfsk50k-30b	17.980 MHz
cc1407p-keyfinder-ring-found-gfsk50k	2180 MHz
cc1407p-wisun-fsk-1a-30b	19.280 MHz
cc1407p-wisun-fsk-1b-30b	19.280 MHz
cc1407p-wisun-fsk-2a-30b	11.780 MHz
cc1407p-wisun-fsk-2b-30b	11.680 MHz
cc1407p-wisun-fsk-3-30b	9.280 MHz
cc1407p-wisun-fsk-4a-30b	8.280 MHz
cc1407p-wisun-fsk-4b-30b	8.180 MHz
cc1407p-wisun-fsk-5-30b	7.180 MHz
none				n/a
thread-appletv-leader					32 to 144160 MHz
thread-host-mac-via-apple-tv			28.5 to 70.948 MHz
thread-host-to-parent			26 to 2948 MHz
thread-to-cc1354p10		44 to 103
thread-to-cc2340r5			42 to 47
wifi-ping-mac						17160 MHz, 1 run

Sources

board-to-board/ble-cc2340r5-central/README.md, board-to-board/thread-two-radios-2026-10-06/README.md, cc35x1/RESULTS-cc35x1.md, site/data/radio-demos.json

Over-the-air update completed %, higher is better

Variant	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-EM-CC35X1
ble-mac	10048 MHz, 1 run
none		n/a
wifi-ota-http			100160 MHz, 1 run
zigbee-home-assistant	Not for this chip
(one way)

Sources

cc1354p10/zigbee-sleepy-sensor/README.md, cc2340r5/oad/README.md, cc35x1/RESULTS-cc35x1.md

Packet loss over one hour %, lower is better

Variant	LP-MSPM0G5187	LP-CC1354P10-1	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC35X1
cc1407p-gfsk50k-915	080 MHz
cc1407p-gfsk50k-915-350pkt	080 MHz
cc1407p-gfsk50k-915-6h	080 MHz
cc1407p-gfsk50k-from-cc1354p10-20pkt	080 MHz
cc1407p-gfsk50k-spi-sweep-80pkt	080 MHz
cc1407p-ofdm-ofdmlr-xlr
cc1407p-sun-oqpsk-rm3-to-cc1354p10	Not for this chip
cc1407p-voice-adpcm32k-wisun2b-122frames	080 MHz
cc1407p-wisun-fsk-1a-from-cc1354p10-20pkt	080 MHz
cc1407p-wisun-fsk-1b-from-cc1354p10-20pkt	080 MHz
cc1407p-wisun-fsk-2a-from-cc1354p10-20pkt	080 MHz
cc1407p-wisun-fsk-2b-from-cc1354p10-20pkt	080 MHz
cc1407p-wisun-fsk-2b-to-cc1354p10-260pkt	080 MHz
cc1407p-wisun-fsk-3-from-cc1354p10-20pkt	080 MHz
cc1407p-wisun-fsk-4a-from-cc1354p10-20pkt	480 MHz
cc1407p-wisun-fsk-4b-from-cc1354p10-20pkt	080 MHz
cc1407p-wisun-fsk-4b-to-cc1354p10-260pkt	080 MHz
cc1407p-wisun-fsk-5-from-cc1354p10-20pkt	1480 MHz
cc1407p-xlr-to-cc1354p10-slr2500bps2gfsk	Not for this chip
cc1407p-xlr-to-cc1354p10-slr5kbps2gfsk	Not for this chip
mqtt					0160 MHz
none			n/a
thread-appletv-leader				0160 MHz, 1 run
thread-host-rcp-apple-tv		0.4448 MHz
zigbee-zed-zha
(one way)

Sources

cc1354p10/zigbee-sleepy-sensor/README.md, cc35x1/mqtt-test/README.md, mspm0g5187/radio/README.md, site/data/radio-demos.json

Stack offered and runs status

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	CC3200-LAUNCHXL	LAUNCHXL-F2800137	ESP32-C6-DevKitC-1-N8	LP-EM-CC2755P10	LP-EM-CC35X1	IWRL6432BOOST
ble		Works	Works				Works	Works
ble-proxy-zigbee-ed						Not for this chip
ble-zephyr-ti-fork		Not for this chip
ble-zephyr-upstream		Not for this chip
dmm		Not for this chip	Works				Works
matlab-support-package
matter		Not for this chip						yes
mmwave-l-sdk-host									Not for this chip
none					n/a
ota-ble		Works48 MHz
ota-zigbee		Not for this chip
proprietary
radar-toolbox
sub1ghz
sub1ghz-bare-host-hal
sub1ghz-cc1407p	Works80 MHz
sub1ghz-zephyr-cc140xp-samples
thread		Not for this chip	Works			yes160 MHz
thread-ftd							Works
thread-rcp		Works	Works
thread-two-radios			Works
visualizer-mac
wifi				Not for this chip				Works
zigbee		Works	Works			yes160 MHz, 1 run
zigbee-thread						Not for this chip
zigbee-zha-end-device			Works

Sources

board-to-board/thread-two-radios-2026-10-06/README.md, board-to-board/zigbee-cc1354-cc2340/README.md, cc1354p10/README.md, cc1354p10/dmm-zigbee-ble/README.md, cc1354p10/thread/README.md, cc1354p10/zigbee-sleepy-sensor/README.md, cc2340r5/FAQ.md, cc2340r5/matrix/README.md, cc2340r5/oad/logs/2026-11-02/oad-run4.json, cc2340r5/thread/README.md, cc2340r5/zephyr-ti/README.md, cc2340r5/zigbee/README.md, cc2755p10/README.md, cc2755p10/host-cc1407p/README.md, cc3200-launchxl/README.md, cc35x1/README.md, docs/experiments/2026-10-24-cc2755p10-cc1407p-host-on-silicon.md, esp32c6/proxy/frl-ble-proxy-zigbee.yaml, esp32c6/zigbee/README.md, esp32c6/zigbee/zb-thread-merge/main.patch, iwrl6432boost/README.md, mspm0g5187/radio/README.md

Setpoint or state change still works with the internet cable pulled yes/no

Variant	LP-EM-CC35X1
alexa
apple-home
google-home
home-assistant

Sources

cc35x1/matter/COMPATIBILITY.md

Controllers that hold one device at once and all still control it controllers, higher is better

Variant	LP-EM-CC35X1
(one way)

Sources

cc35x1/matter/COMPATIBILITY.md

Setup again after a reflash and factory reset, per platform s, lower is better

Variant	LP-EM-CC35X1
alexa
apple-home
google-home
home-assistant

Sources

cc35x1/matter/COMPATIBILITY.md

Button press on a Bluetooth LE HID device to the host's action (a media key changing the player), median of 20 ms, lower is better

Variant	LP-EM-CC2340R5
macos-music

Sources

ports/ble-hid-remote/README.md

Input change on the board to the notification on the central, on the central's clock ms, lower is better

Variant	LP-EM-CC2340R5
nordic-lbs-button-uart-stand-in	23.8 to 29.848 MHz

Sources

Run records are in the notebook, not this copy.

Bluetooth LE HID device paired and bonded by the host's own settings, no app: time from the host's pairing request to bonded s, lower is better

Variant	LP-EM-CC2340R5
macos-just-works	1211 run

Sources

Run records are in the notebook, not this copy.

Advertisement callbacks per second a Mac scanner (bleak) reports for the gamepad, 30 s windows 1/s

Variant	LP-EM-CC2340R5	ESP32-C6-DevKitC-1-N8
arduino-nimble-library		4.1 to 9.5
core-ticlang-ble-hogp	2 to 3.248 MHz

Sources

Run records are in the notebook, not this copy.

Time for the Mac (bleak) to connect to the gamepad, median of 4 connects ms, lower is better

Variant	LP-EM-CC2340R5	ESP32-C6-DevKitC-1-N8
arduino-nimble-library		11 to 12
core-ticlang-ble-hogp	31,341 to 31,39148 MHz

Sources

Run records are in the notebook, not this copy.

Edge AI

A hardware AI engine against the CPU, same model

The MSPM0 G5187's NPU answers 2 to 22 times faster than its own CPU. The SK-TDA4VM's C7x is 13 to 135 times faster than its A72 cores on the models shown.

The same model built for the accelerator and for the CPU, timed per answer.

Open the Edge AI explorer.

acceleratorCPU, same modelyour board

0.1 ms1 ms10 ms100 ms1000 ms

Room occupancy: tiny on the NPU

1.29 ms

Room occupancy: tiny on the CPU

4.94 ms

Room occupancy: small on the NPU

1.45 ms

Room occupancy: small on the CPU

12 ms

Room occupancy: medium on the NPU

2.12 ms

Room occupancy: medium on the CPU

45.6 ms

Room occupancy: large on the NPU

69.4 ms

Room occupancy: large on the CPU

159 ms

Time per answer on the MSPM0 G5187 at 80 MHz, the same model built for the NPU and for the CPU by the same compiler.

acceleratorCPU, same modelyour board

1 ms10 ms100 ms1000 ms10000 ms

Image classification (RegNetX 200MF), C7x

2.89 ms

Image classification (RegNetX 200MF), A72

37.4 ms

Object detection, small (lite), C7x

4.66 ms

Object detection, small (lite), A72

193 ms

Object detection (SSD-Lite MobileNetV2), C7x

6.1 ms

Object detection (SSD-Lite MobileNetV2), A72

438 ms

Segmentation (DeepLabV3-Lite), C7x

7.3 ms

Segmentation (DeepLabV3-Lite), A72

714 ms

Face detection, C7x

11.2 ms

Face detection, A72

1504 ms

Object detection, medium, C7x

11.7 ms

Object detection, medium, A72

1523 ms

SK-TDA4VM: the six fastest of 8 zoo models that ran on both; the rest are in the Linux block.

All rows

Room-occupancy model, one inference ms, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X	SK-TDA4VM
npu	n/a	n/a
Room occupancy: small, GCC 14 -O3										3.5196 MHz, 1 run
Room occupancy: small, GCC 14 -Os										10.496 MHz, 1 run
Room occupancy: small, GCC -O3, on the A72 cores																		0.04472000 MHz
Room occupancy: small, plain C -O2, code in flash								5.11 to 5.12120 MHz				2.75 to 2.76160 MHz
Room occupancy: small, plain C -O2, code in RAM								4.31 to 4.32120 MHz
Room occupancy: small, plain C -O3	9.0580 MHz, 1 run	21.9 to 2232 MHz, 1 run	Not for this chip	9.0480 MHz, 1 run	9.05 to 9.180 MHz, 1 run	16.648 MHz	6.2248 MHz			3.696 MHz, 1 run		2.17160 MHz	1.86160 MHz, 1 run	1.25200 MHz, 1 run	2.696 MHz	1.54160 MHz
Room occupancy: small, plain C -O3, at 32 MHz													7.7932 MHz, 1 run
Room occupancy: small, plain C -O3, M33 core											1.89150 MHz
Room occupancy: small, plain C -O3, M33 core, GCC 12											1.93150 MHz
Room occupancy: small, plain C -O3, code in RAM, M33 core											1.9150 MHz
Room occupancy: small, plain C -O3, code in RAM, RISC-V core											2.09150 MHz
Room occupancy: small, plain C -O3, RISC-V core											2.1150 MHz
Room occupancy: small, plain C -O4, code in flash								4.83 to 4.85120 MHz
Room occupancy: small, plain C -O4, code in RAM								4.11120 MHz
Room occupancy: small, plain C -O4, speed setting 0, code in flash								4.93 to 4.94120 MHz
Room occupancy: small, plain C -O4, speed setting 0, code in RAM								4.12 to 4.13120 MHz
Room occupancy: small, plain C -O4, speed setting 5, code in flash								4.81 to 4.82120 MHz
Room occupancy: small, plain C -O4, speed setting 5, code in RAM								4.09120 MHz									2.37200 MHz
Room occupancy: small, plain C -Os	10.980 MHz, 1 run	28.332 MHz, 1 run	Not for this chip	10.980 MHz, 1 run	10.980 MHz, 1 run	2048 MHz	9.8148 MHz			10.496 MHz, 1 run		3.82 to 3.83160 MHz	2.62160 MHz, 1 run	1.86200 MHz, 1 run	3.8696 MHz	2.32160 MHz
Room occupancy: small, plain C -Os, M33 core											4.36150 MHz
Room occupancy: small, plain C -Os, M33 core, GCC 12											4.35150 MHz
Room occupancy: small, plain C -Os, RISC-V core											3.79150 MHz
Room occupancy: small, GCC 12 -O3									15.3 to 15.548 MHz, 1 run
Room occupancy: small, GCC 12 -Os									28 to 28.248 MHz, 1 run
Room occupancy: small, TI neural-network compiler, on the CDE coprocessor														Not for this chip
Room occupancy: small, TI neural-network compiler, on the CPU	1280 MHz, 1 run			1280 MHz, 1 run	1280 MHz	20.448 MHz		2.6120 MHz						1.38200 MHz
Room occupancy: small, TI neural-network compiler, on the CPU, at 32 MHz					26.6 to 26.732 MHz
Room occupancy: small, TI neural-network compiler, on the CPU, code in RAM								2.97120 MHz
Room occupancy: small, TI neural-network compiler, on the CPU, source built for the M33							6.0548 MHz								2.7396 MHz
Room occupancy: small, TI neural-network compiler, on the NPU					1.45 to 1.4980 MHz									0.205200 MHz
Room occupancy: small, TI neural-network compiler, on the NPU, at 32 MHz					3.23 to 3.3332 MHz
Room occupancy: small, ST's GCC 14 -O3									15.3 to 15.448 MHz, 1 run
Room occupancy: large, GCC 14 -O3										45.596 MHz, 1 run
Room occupancy: large, GCC 14 -Os										16396 MHz, 1 run
Room occupancy: large, GCC -O3, on the A72 cores																		0.6862000 MHz
Room occupancy: large, plain C -O2, code in flash								67.4120 MHz				40.1160 MHz
Room occupancy: large, plain C -O3	129.480 MHz, 1 run	Not for this chip	Not for this chip	129.480 MHz, 1 run	129.480 MHz, 1 run	228.948 MHz	96.448 MHz			45.896 MHz, 1 run		27.1 to 27.2160 MHz	27.2160 MHz, 1 run	18.7200 MHz, 1 run	3996 MHz	24.4160 MHz
Room occupancy: large, plain C -O3, at 32 MHz													116.732 MHz, 1 run
Room occupancy: large, plain C -O3, M33 core											27150 MHz
Room occupancy: large, plain C -O3, M33 core, GCC 12											27.6150 MHz
Room occupancy: large, plain C -O3, code in RAM, M33 core											25.9150 MHz
Room occupancy: large, plain C -O3, code in RAM, RISC-V core											27.3150 MHz
Room occupancy: large, plain C -O3, RISC-V core											29150 MHz
Room occupancy: large, plain C -O4, code in flash								65.4 to 65.5120 MHz
Room occupancy: large, plain C -O4, code in RAM								Not for this chip
Room occupancy: large, plain C -O4, speed setting 0, code in flash								66.6 to 66.7120 MHz
Room occupancy: large, plain C -O4, speed setting 5, code in flash								65.3 to 65.4120 MHz									47.9200 MHz
Room occupancy: large, plain C -Os	199.480 MHz, 1 run	Not for this chip	Not for this chip	145.380 MHz, 1 run	145.3 to 145.480 MHz, 1 run	30848 MHz	155.248 MHz			16396 MHz, 1 run		57.1 to 57.2160 MHz	39.7160 MHz, 1 run	28.6200 MHz, 1 run	59.696 MHz	36.3160 MHz
Room occupancy: large, plain C -Os, M33 core											71.6150 MHz
Room occupancy: large, plain C -Os, M33 core, GCC 12											71.6150 MHz
Room occupancy: large, plain C -Os, RISC-V core											61.9150 MHz
Room occupancy: large, GCC 12 -O3									185.2 to 185.448 MHz, 1 run
Room occupancy: large, GCC 12 -Os									414.6 to 414.748 MHz, 1 run
Room occupancy: large, TI neural-network compiler, on the CDE coprocessor														Not for this chip
Room occupancy: large, TI neural-network compiler, on the CPU	159.180 MHz, 1 run			159.280 MHz, 1 run	159.280 MHz	267.848 MHz		30.8120 MHz						22.9200 MHz
Room occupancy: large, TI neural-network compiler, on the CPU, at 32 MHz					357.4 to 357.532 MHz
Room occupancy: large, TI neural-network compiler, on the CPU, code in RAM								Not for this chip
Room occupancy: large, TI neural-network compiler, on the CPU, source built for the M33							9148 MHz								40.696 MHz
Room occupancy: large, TI neural-network compiler, on the NPU					69.480 MHz									9.53200 MHz
Room occupancy: large, TI neural-network compiler, on the NPU, at 32 MHz					161 to 161.132 MHz
Room occupancy: large, ST's GCC 14 -O3									185.7 to 185.848 MHz, 1 run
Room occupancy: tiny, GCC 14 -O3										1.6896 MHz, 1 run
Room occupancy: tiny, GCC 14 -Os										3.0796 MHz, 1 run
Room occupancy: tiny, GCC -O3, on the A72 cores																		0.01742000 MHz
Room occupancy: tiny, plain C -O2, code in flash								1.98120 MHz				1.23 to 1.25160 MHz
Room occupancy: tiny, plain C -O2, code in RAM								1.72 to 1.73120 MHz
Room occupancy: tiny, plain C -O3	480 MHz, 1 run	9.3 to 9.632 MHz, 1 run	Not for this chip	3.9180 MHz, 1 run	3.92 to 4.0280 MHz, 1 run	7.0348 MHz	2.3348 MHz			1.6696 MHz, 1 run		1.12 to 1.14160 MHz	0.749160 MHz, 1 run	0.489200 MHz, 1 run	1.0196 MHz	0.602160 MHz
Room occupancy: tiny, plain C -O3, at 32 MHz													3.0432 MHz, 1 run
Room occupancy: tiny, plain C -O3, M33 core											0.75150 MHz
Room occupancy: tiny, plain C -O3, M33 core, GCC 12											0.76150 MHz
Room occupancy: tiny, plain C -O3, code in RAM, M33 core											0.759150 MHz
Room occupancy: tiny, plain C -O3, code in RAM, RISC-V core											0.994150 MHz
Room occupancy: tiny, plain C -O3, RISC-V core											0.988150 MHz
Room occupancy: tiny, plain C -O4, code in flash								1.82 to 1.83120 MHz
Room occupancy: tiny, plain C -O4, code in RAM								1.6120 MHz
Room occupancy: tiny, plain C -O4, speed setting 0, code in flash								1.86 to 1.87120 MHz
Room occupancy: tiny, plain C -O4, speed setting 0, code in RAM								1.6120 MHz
Room occupancy: tiny, plain C -O4, speed setting 5, code in flash								1.8 to 1.81120 MHz
Room occupancy: tiny, plain C -O4, speed setting 5, code in RAM								1.58120 MHz									0.893 to 0.896200 MHz
Room occupancy: tiny, plain C -Os	4.6880 MHz, 1 run	12 to 12.232 MHz, 1 run	Not for this chip	4.6780 MHz, 1 run	4.68 to 4.7680 MHz, 1 run	8.3748 MHz	3.5448 MHz			3.0796 MHz, 1 run		1.62 to 1.64160 MHz	0.965160 MHz, 1 run	0.681200 MHz, 1 run	1.4196 MHz	0.84160 MHz
Room occupancy: tiny, plain C -Os, M33 core											1.38150 MHz
Room occupancy: tiny, plain C -Os, M33 core, GCC 12											1.38150 MHz
Room occupancy: tiny, plain C -Os, RISC-V core											1.48150 MHz
Room occupancy: tiny, GCC 12 -O3									8.16 to 8.3948 MHz, 1 run
Room occupancy: tiny, GCC 12 -Os									11 to 11.248 MHz, 1 run
Room occupancy: tiny, TI neural-network compiler, on the CDE coprocessor														Not for this chip
Room occupancy: tiny, TI neural-network compiler, on the CPU	4.9580 MHz, 1 run			4.9380 MHz, 1 run	4.94 to 4.9780 MHz	7.9648 MHz		1.12 to 1.13120 MHz						0.499200 MHz
Room occupancy: tiny, TI neural-network compiler, on the CPU, at 32 MHz					10.932 MHz
Room occupancy: tiny, TI neural-network compiler, on the CPU, code in RAM								1.11 to 1.12120 MHz
Room occupancy: tiny, TI neural-network compiler, on the CPU, source built for the M33							2.1348 MHz								0.95296 MHz
Room occupancy: tiny, TI neural-network compiler, on the NPU					1.29 to 1.3280 MHz									0.107200 MHz
Room occupancy: tiny, TI neural-network compiler, on the NPU, at 32 MHz					2.88 to 2.9432 MHz
Room occupancy: tiny, TI neural-network compiler, on the NPU, container														0.106200 MHz
Room occupancy: tiny, ST's GCC 14 -O3									8.05 to 8.2648 MHz, 1 run
Room occupancy: medium, GCC 14 -O3										12.696 MHz, 1 run
Room occupancy: medium, GCC 14 -Os										42.996 MHz, 1 run
Room occupancy: medium, GCC -O3, on the A72 cores																		0.1932000 MHz
Room occupancy: medium, plain C -O2, code in flash								18.2 to 18.3120 MHz				10.1160 MHz
Room occupancy: medium, plain C -O3	35.380 MHz, 1 run	83.7 to 83.832 MHz, 1 run	Not for this chip	35.380 MHz, 1 run	35.380 MHz, 1 run	63.248 MHz	25.848 MHz			12.696 MHz, 1 run		7.14 to 7.16160 MHz	7.38160 MHz, 1 run	5.05200 MHz, 1 run	10.596 MHz	6.25160 MHz
Room occupancy: medium, plain C -O3, at 32 MHz													31.532 MHz, 1 run
Room occupancy: medium, plain C -O3, M33 core											6.82150 MHz
Room occupancy: medium, plain C -O3, M33 core, GCC 12											7150 MHz
Room occupancy: medium, plain C -O3, code in RAM, M33 core											6.99150 MHz
Room occupancy: medium, plain C -O3, code in RAM, RISC-V core											7.58150 MHz
Room occupancy: medium, plain C -O3, RISC-V core											7.5150 MHz
Room occupancy: medium, plain C -O4, code in flash								17.6120 MHz
Room occupancy: medium, plain C -O4, code in RAM								Not for this chip
Room occupancy: medium, plain C -O4, speed setting 0, code in flash								17.9120 MHz
Room occupancy: medium, plain C -O4, speed setting 5, code in flash								17.5120 MHz
Room occupancy: medium, plain C -O4, speed setting 5, code in RAM																	8.38 to 8.39200 MHz
Room occupancy: medium, plain C -Os	39.980 MHz, 1 run	99.9 to 10032 MHz, 1 run	Not for this chip	39.980 MHz, 1 run	39.9 to 4080 MHz, 1 run	83.248 MHz	4148 MHz			42.996 MHz, 1 run		15.1 to 15.2160 MHz	10.6160 MHz, 1 run	7.61200 MHz, 1 run	15.896 MHz	9.45160 MHz
Room occupancy: medium, plain C -Os, M33 core											18150 MHz
Room occupancy: medium, plain C -Os, M33 core, GCC 12											18150 MHz
Room occupancy: medium, plain C -Os, RISC-V core											15.2150 MHz
Room occupancy: medium, plain C -Os, with Bluetooth running						71.948 MHz
Room occupancy: medium, GCC 12 -O3									51.6 to 51.848 MHz, 1 run
Room occupancy: medium, GCC 12 -Os									109.9 to 11048 MHz, 1 run
Room occupancy: medium, low-power core, network only												213.8 to 213.9160 MHz
Room occupancy: medium, TI neural-network compiler, on the CDE coprocessor														Not for this chip
Room occupancy: medium, TI neural-network compiler, on the CPU	45.680 MHz, 1 run			45.680 MHz, 1 run	45.6 to 45.780 MHz	77.148 MHz		9.05 to 9.06120 MHz						5.54200 MHz
Room occupancy: medium, TI neural-network compiler, on the CPU, at 32 MHz					101 to 101.132 MHz
Room occupancy: medium, TI neural-network compiler, on the CPU, code in RAM								Not for this chip
Room occupancy: medium, TI neural-network compiler, on the CPU, source built for the M33							24.448 MHz								10.896 MHz
Room occupancy: medium, TI neural-network compiler, on the NPU					2.12 to 2.1580 MHz									0.63200 MHz
Room occupancy: medium, TI neural-network compiler, on the NPU, at 32 MHz					4.77 to 4.8432 MHz
Room occupancy: medium, ST's GCC 14 -O3									52 to 52.148 MHz, 1 run
occ_float_2k-ort-int8-infloat-nnc														1.56200 MHz
occ_float_2k-ort-int8-nnc														1.61200 MHz
occ_float_2k-ti-ptq-nnc-npu														0.205200 MHz
occ_float_30k-ort-int8-infloat-nnc														21.3200 MHz
occ_float_30k-ort-int8-nnc														21.4200 MHz
occ_float_30k-ti-ptq-nnc-npu														9.46200 MHz
occ_float_500-ort-int8-infloat-nnc														0.645200 MHz
occ_float_500-ort-int8-nnc														0.655200 MHz
occ_float_500-ti-ptq-nnc-npu														0.107200 MHz
occ_float_8k-ort-int8-infloat-nnc														6.44200 MHz
occ_float_8k-ort-int8-nnc														6.45200 MHz
occ_float_8k-ti-ptq-nnc-npu														0.63200 MHz

Sources

cc35x1/occ/occ_run.c

Room-occupancy model fits and passes 20 of 20 yes/no

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X
Room occupancy: small	Works80 MHz, 1 run	Works32 MHz, 1 run	Not for this chip	Works80 MHz, 1 run	Works80 MHz	yes48 MHz	yes48 MHz		Works48 MHz, 1 run	Works96 MHz, 1 run	Works160 MHz	yes160 MHz, 1 run	yes200 MHz, 1 run	yes96 MHz, 1 run	yes160 MHz
Room occupancy: small, plain C -O4, code in flash								yes120 MHz, 1 run
Room occupancy: small, plain C -O4, code in RAM								yes120 MHz, 1 run
Room occupancy: small, plain C -O4, speed setting 5, code in RAM																yes200 MHz, 1 run
Room occupancy: small, TI neural-network compiler, on the CPU								yes120 MHz
Room occupancy: small, TI neural-network compiler, on the CPU, code in RAM								yes120 MHz
Room occupancy: large	Works80 MHz, 1 run	Not for this chip	Not for this chip	Works80 MHz, 1 run	Works80 MHz	yes48 MHz	yes48 MHz		Works48 MHz, 1 run	Works96 MHz, 1 run	Works160 MHz	yes160 MHz, 1 run	yes200 MHz, 1 run	yes96 MHz, 1 run	yes160 MHz
Room occupancy: large, plain C -O2, code in RAM								Not for this chip
Room occupancy: large, plain C -O4, code in flash								yes120 MHz, 1 run
Room occupancy: large, plain C -O4, code in RAM								Not for this chip
Room occupancy: large, plain C -O4, speed setting 0, code in RAM								Not for this chip
Room occupancy: large, plain C -O4, speed setting 5, code in flash																yes200 MHz, 1 run
Room occupancy: large, plain C -O4, speed setting 5, code in RAM								Not for this chip								Not for this chip
Room occupancy: large, TI neural-network compiler, on the CPU								yes120 MHz
Room occupancy: large, TI neural-network compiler, on the CPU, code in RAM								Not for this chip
Room occupancy: large, under Zephyr										Works96 MHz, 1 run
Room occupancy: tiny	Works80 MHz, 1 run	Works32 MHz, 1 run	Not for this chip	Works80 MHz, 1 run	Works80 MHz	yes48 MHz	yes48 MHz		Works48 MHz, 1 run	Works96 MHz, 1 run	Works160 MHz	yes160 MHz, 1 run	yes200 MHz, 1 run	yes96 MHz, 1 run	yes160 MHz
Room occupancy: tiny, plain C -O4, code in flash								yes120 MHz, 1 run
Room occupancy: tiny, plain C -O4, code in RAM								yes120 MHz, 1 run
Room occupancy: tiny, plain C -O4, speed setting 5, code in RAM																yes200 MHz, 1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU								yes120 MHz
Room occupancy: tiny, TI neural-network compiler, on the CPU, code in RAM								yes120 MHz
Room occupancy: medium	Works80 MHz, 1 run	Works32 MHz, 1 run	Not for this chip	Works80 MHz, 1 run	Works80 MHz	yes48 MHz	yes48 MHz		Works48 MHz, 1 run	Works96 MHz, 1 run	Works160 MHz	yes160 MHz, 1 run	yes200 MHz, 1 run	yes96 MHz, 1 run	yes160 MHz
Room occupancy: medium, plain C -O2, code in RAM								Not for this chip
Room occupancy: medium, plain C -O4, code in flash								yes120 MHz, 1 run
Room occupancy: medium, plain C -O4, code in RAM								Not for this chip
Room occupancy: medium, plain C -O4, speed setting 0, code in RAM								Not for this chip
Room occupancy: medium, plain C -O4, speed setting 5, code in RAM								Not for this chip								yes200 MHz, 1 run
Room occupancy: medium, plain C -Os, with Bluetooth running						yes48 MHz
Room occupancy: medium, low-power core, whole program											Not for this chip
Room occupancy: medium, TI neural-network compiler, on the CPU								yes120 MHz
Room occupancy: medium, TI neural-network compiler, on the CPU, code in RAM								Not for this chip

Sources

cc35x1/occ/occ_run.c, docs/experiments/2026-10-11-frdm-mcxa153-model-under-zephyr.md

Room-occupancy model: weights plus model code (plus the NNC runtime it links), from the linker map B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-AM13E230	LP-EM-CC2755P10	LAUNCHXL-F28P65X
character_recognition				66.4 kB1 run
character_recognition_npu				71.4 kB1 run
ecg_anomaly_detection_ai				63.5 kB1 run
ecg_anomaly_detection_ai_npu				86.9 kB1 run
fall_detection_ai				12.6 kB1 run
fall_detection_ai_npu				13.8 kB1 run
motor_fan_blade_fault_ai				1.4 kB1 run
motor_fan_blade_fault_ai_npu				2.1 kB1 run
Room occupancy: small, plain C -O2, code in flash							5.3 kB1 run
Room occupancy: small, plain C -O2, code in RAM							5.3 kB1 run
Room occupancy: small, plain C -O3	9.5 kB1 run	9.5 kB1 run	9.5 kB1 run	9.5 kB1 run	9.5 kB1 run	10.1 kB1 run		10.1 kB1 run	10.1 kB1 run
Room occupancy: small, plain C -O4, code in flash							4.9 kB1 run
Room occupancy: small, plain C -O4, code in RAM							4.9 kB1 run
Room occupancy: small, plain C -O4, speed setting 0, code in flash							4.9 kB1 run
Room occupancy: small, plain C -O4, speed setting 0, code in RAM							4.9 kB1 run
Room occupancy: small, plain C -O4, speed setting 5, code in flash							5.4 kB1 run
Room occupancy: small, plain C -O4, speed setting 5, code in RAM							5.4 kB1 run			5.4 kB1 run
Room occupancy: small, plain C -Os	3.1 kB1 run	3.1 kB1 run	3.1 kB1 run	3.1 kB1 run	3.1 kB1 run	3.1 kB1 run		3.1 kB1 run	3.1 kB1 run
Room occupancy: small, TI neural-network compiler, on the CPU	5.3 kB1 run		5.3 kB1 run	5.3 kB1 run	5.3 kB1 run		10.4 kB1 run	12.5 kB1 run
Room occupancy: small, TI neural-network compiler, on the CPU, code in RAM							10.4 kB1 run
Room occupancy: small, TI neural-network compiler, on the CPU, source built for the M33						12 kB1 run			12 kB1 run
Room occupancy: small, TI neural-network compiler, on the NPU				6.5 kB1 run				9 kB1 run
Room occupancy: large, plain C -O2, code in flash							67.1 kB1 run
Room occupancy: large, plain C -O3	40.7 kB1 run		40.7 kB1 run	40.7 kB1 run	40.7 kB1 run	40.8 kB1 run		40.8 kB1 run	40.8 kB1 run
Room occupancy: large, plain C -O4, code in flash							66.8 kB1 run
Room occupancy: large, plain C -O4, speed setting 0, code in flash							66.8 kB1 run
Room occupancy: large, plain C -O4, speed setting 5, code in flash							67.2 kB1 run			67.2 kB1 run
Room occupancy: large, plain C -Os	34.6 kB1 run		34.6 kB1 run	34.6 kB1 run	34.6 kB1 run	34.6 kB1 run		34.6 kB1 run	34.6 kB1 run
Room occupancy: large, TI neural-network compiler, on the CPU	37.9 kB1 run		37.9 kB1 run	37.9 kB1 run	37.9 kB1 run		73.4 kB1 run	45.1 kB1 run
Room occupancy: large, TI neural-network compiler, on the CPU, source built for the M33						43.7 kB1 run			43.7 kB1 run
Room occupancy: large, TI neural-network compiler, on the NPU				59.8 kB1 run				62.5 kB1 run
Room occupancy: tiny, plain C -O2, code in flash							2.8 kB1 run
Room occupancy: tiny, plain C -O2, code in RAM							2.8 kB1 run
Room occupancy: tiny, plain C -O3	6.3 kB1 run	6.3 kB1 run	6.3 kB1 run	6.3 kB1 run	6.3 kB1 run	6.7 kB1 run		6.7 kB1 run	6.7 kB1 run
Room occupancy: tiny, plain C -O4, code in flash							2.5 kB1 run
Room occupancy: tiny, plain C -O4, code in RAM							2.5 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 0, code in flash							2.5 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 0, code in RAM							2.5 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 5, code in flash							2.9 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 5, code in RAM							2.9 kB1 run			2.9 kB1 run
Room occupancy: tiny, plain C -Os	1.7 kB1 run	1.7 kB1 run	1.7 kB1 run	1.7 kB1 run	1.7 kB1 run	1.7 kB1 run		1.7 kB1 run	1.7 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU	2.6 kB1 run		2.6 kB1 run	2.6 kB1 run	2.6 kB1 run		5 kB1 run	7 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU, code in RAM							5 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU, source built for the M33						6.5 kB1 run			6.5 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the NPU				4.7 kB1 run				7.1 kB1 run
Room occupancy: medium, plain C -O2, code in flash							18.8 kB1 run
Room occupancy: medium, plain C -O3	16.6 kB1 run	16.6 kB1 run	16.6 kB1 run	16.6 kB1 run	16.6 kB1 run	16.6 kB1 run		16.6 kB1 run	16.6 kB1 run
Room occupancy: medium, plain C -O4, code in flash							18.5 kB1 run
Room occupancy: medium, plain C -O4, speed setting 0, code in flash							18.5 kB1 run
Room occupancy: medium, plain C -O4, speed setting 5, code in flash							18.9 kB1 run
Room occupancy: medium, plain C -O4, speed setting 5, code in RAM										18.9 kB1 run
Room occupancy: medium, plain C -Os	10.1 kB1 run	10.1 kB1 run	10.1 kB1 run	10.1 kB1 run	10.1 kB1 run	10 kB1 run		10 kB1 run	10 kB1 run
Room occupancy: medium, TI neural-network compiler, on the CPU	12.6 kB1 run		12.6 kB1 run	12.6 kB1 run	12.6 kB1 run		24.6 kB1 run	20.5 kB1 run
Room occupancy: medium, TI neural-network compiler, on the CPU, source built for the M33						19.2 kB1 run			19.2 kB1 run
Room occupancy: medium, TI neural-network compiler, on the NPU				14 kB1 run				16.6 kB1 run

Sources

data/ai-model-sizes.csv, docs/experiments/2026-10-09-mspm0g5187-npu-vs-cpu.md

Room-occupancy model: scratch RAM (activation buffers), from the linker map B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-AM13E230	LP-EM-CC2755P10	LAUNCHXL-F28P65X
character_recognition				8 kB1 run
character_recognition_npu				10.7 kB1 run
ecg_anomaly_detection_ai				18 kB1 run
ecg_anomaly_detection_ai_npu				12.9 kB1 run
fall_detection_ai				9601 run
fall_detection_ai_npu				7321 run
motor_fan_blade_fault_ai				1 kB1 run
motor_fan_blade_fault_ai_npu				8721 run
Room occupancy: small, plain C -O2, code in flash							1.2 kB1 run
Room occupancy: small, plain C -O2, code in RAM							1.2 kB1 run
Room occupancy: small, plain C -O3	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run		1.1 kB1 run	1.1 kB1 run
Room occupancy: small, plain C -O4, code in flash							1.2 kB1 run
Room occupancy: small, plain C -O4, code in RAM							1.2 kB1 run
Room occupancy: small, plain C -O4, speed setting 0, code in flash							1.2 kB1 run
Room occupancy: small, plain C -O4, speed setting 0, code in RAM							1.2 kB1 run
Room occupancy: small, plain C -O4, speed setting 5, code in flash							1.2 kB1 run
Room occupancy: small, plain C -O4, speed setting 5, code in RAM							1.2 kB1 run			1.2 kB1 run
Room occupancy: small, plain C -Os	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run		1.1 kB1 run	1.1 kB1 run
Room occupancy: small, TI neural-network compiler, on the CPU	1.5 kB1 run		1.5 kB1 run	1.5 kB1 run	1.5 kB1 run		3.1 kB1 run	1.5 kB1 run
Room occupancy: small, TI neural-network compiler, on the CPU, code in RAM							3.1 kB1 run
Room occupancy: small, TI neural-network compiler, on the CPU, source built for the M33						1.5 kB1 run			1.5 kB1 run
Room occupancy: small, TI neural-network compiler, on the NPU				7001 run				7001 run
Room occupancy: large, plain C -O2, code in flash							4.2 kB1 run
Room occupancy: large, plain C -O3	4.1 kB1 run		4.1 kB1 run	4.1 kB1 run	4.1 kB1 run	4.1 kB1 run		4.1 kB1 run	4.1 kB1 run
Room occupancy: large, plain C -O4, code in flash							4.2 kB1 run
Room occupancy: large, plain C -O4, speed setting 0, code in flash							4.2 kB1 run
Room occupancy: large, plain C -O4, speed setting 5, code in flash							4.2 kB1 run			4.2 kB1 run
Room occupancy: large, plain C -Os	4.1 kB1 run		4.1 kB1 run	4.1 kB1 run	4.1 kB1 run	4.1 kB1 run		4.1 kB1 run	4.1 kB1 run
Room occupancy: large, TI neural-network compiler, on the CPU	7 kB1 run		7 kB1 run	7 kB1 run	7 kB1 run		12.2 kB1 run	7 kB1 run
Room occupancy: large, TI neural-network compiler, on the CPU, source built for the M33						7 kB1 run			7 kB1 run
Room occupancy: large, TI neural-network compiler, on the NPU				4.6 kB1 run				4.6 kB1 run
Room occupancy: tiny, plain C -O2, code in flash							1.2 kB1 run
Room occupancy: tiny, plain C -O2, code in RAM							1.2 kB1 run
Room occupancy: tiny, plain C -O3	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run		1.1 kB1 run	1.1 kB1 run
Room occupancy: tiny, plain C -O4, code in flash							1.2 kB1 run
Room occupancy: tiny, plain C -O4, code in RAM							1.2 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 0, code in flash							1.2 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 0, code in RAM							1.2 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 5, code in flash							1.2 kB1 run
Room occupancy: tiny, plain C -O4, speed setting 5, code in RAM							1.2 kB1 run			1.2 kB1 run
Room occupancy: tiny, plain C -Os	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run	1.1 kB1 run		1.1 kB1 run	1.1 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU	1.3 kB1 run		1.3 kB1 run	1.3 kB1 run	1.3 kB1 run		1.8 kB1 run	1.3 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU, code in RAM							1.8 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the CPU, source built for the M33						1.3 kB1 run			1.3 kB1 run
Room occupancy: tiny, TI neural-network compiler, on the NPU				5481 run				5481 run
Room occupancy: medium, plain C -O2, code in flash							2.2 kB1 run
Room occupancy: medium, plain C -O3	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run		2.1 kB1 run	2.1 kB1 run
Room occupancy: medium, plain C -O4, code in flash							2.2 kB1 run
Room occupancy: medium, plain C -O4, speed setting 0, code in flash							2.2 kB1 run
Room occupancy: medium, plain C -O4, speed setting 5, code in flash							2.2 kB1 run
Room occupancy: medium, plain C -O4, speed setting 5, code in RAM										2.2 kB1 run
Room occupancy: medium, plain C -Os	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run	2.1 kB1 run		2.1 kB1 run	2.1 kB1 run
Room occupancy: medium, TI neural-network compiler, on the CPU	3.5 kB1 run		3.5 kB1 run	3.5 kB1 run	3.5 kB1 run		6.1 kB1 run	3.5 kB1 run
Room occupancy: medium, TI neural-network compiler, on the CPU, source built for the M33						3.5 kB1 run			3.5 kB1 run
Room occupancy: medium, TI neural-network compiler, on the NPU				1.4 kB1 run				1.4 kB1 run

Sources

data/ai-model-sizes.csv, docs/experiments/2026-10-09-mspm0g5187-npu-vs-cpu.md

BitNetMCU handwritten digit, one inference cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-EM-CC2755P10
12k-gcc12-o3-96mhz								245k96 MHz, 1 run
12k-ticlang-o3-32mhz	430k32 MHz	413k32 MHz	430k32 MHz	430k32 MHz
1k-gcc-o3-150mhz-m33									51k150 MHz
1k-gcc-o3-150mhz-riscv									57k150 MHz
1k-gcc12-o3-96mhz								44k96 MHz, 1 run
1k-ticlang-o3-160mhz										36k160 MHz
1k-ticlang-o3-32mhz	89k32 MHz	80k32 MHz	89k32 MHz	89k32 MHz
1k-ticlang-o3-48mhz					100k48 MHz	45k48 MHz
1k-ticlang-o3-96mhz											36k96 MHz
cnn-16small-gcc12-o3-96mhz								332k96 MHz, 1 run
cnn-64-c2000-o4s5-120mhz-flash							2.9M120 MHz, 1 run
cnn-64-c2000-o4s5-120mhz-ram							2.55M120 MHz, 1 run
cnn16-small-ticlang-o3-32mhz	661k32 MHz	650k32 MHz	661k32 MHz	661k32 MHz
cnn64-gcc-o3-150mhz-m33									1.44M150 MHz
cnn64-gcc-o3-150mhz-riscv									1.26M150 MHz
cnn64-gcc12-o3-96mhz								1.3M96 MHz, 1 run
cnn64-ticlang-o3-160mhz										1.5M160 MHz
cnn64-ticlang-o3-32mhz	2.64M32 MHz	2.65M32 MHz	2.64M32 MHz	2.64M32 MHz
cnn64-ticlang-o3-32mhz-sramloops	2.95M32 MHz
cnn64-ticlang-o3-48mhz					2.76M48 MHz	1.13M48 MHz
cnn64-ticlang-o3-80mhz	2.75M80 MHz		2.75M80 MHz	2.75M80 MHz
cnn64-ticlang-o3-96mhz											1.09M96 MHz
(one way)

Sources

ports/bitnetmcu/README.md

Vendor AI example, one inference ms, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
arc_fault						11.1120 MHz, 1 run	41.532 MHz, 1 run	0.648200 MHz, 1 run		2.93400 MHz
arc_fault-xip										3.11400 MHz
arc_fault-xip-l2-64k										2.93400 MHz
bitnetmcu-1k			1024 MHz, 1 run
bitnetmcu-cnn16			Not for this chip
blower_imbalance						6.25120 MHz, 1 run	2332 MHz, 1 run	0.401200 MHz, 1 run		1.72400 MHz
blower_imbalance-xip										1.79400 MHz
blower_imbalance-xip-l2-64k										1.72400 MHz
character_recognition					91.980 MHz		18.1160 MHz, 1 run
character_recognition_npu					10.180 MHz
deploy-and-live-preview-examples
ecg_anomaly_detection_ai					496.580 MHz
ecg_anomaly_detection_ai_npu					96.680 MHz
fall_detection_ai					8.5780 MHz
fall_detection_ai_npu					1.4480 MHz
fan_blade_anomalydetection						23.6120 MHz, 1 run		5.43200 MHz, 1 run		2.95400 MHz
fan_blade_anomalydetection-xip										4.96400 MHz
fan_blade_anomalydetection-xip-l2-64k										3.01400 MHz
forecasting_pmsm_rotor_temp						0.619120 MHz, 1 run		0.172200 MHz, 1 run		0.0623400 MHz
forecasting_pmsm_rotor_temp-xip										0.0721400 MHz
forecasting_pmsm_rotor_temp-xip-l2-64k										0.0623400 MHz
gearbox_fault_diagnosis_ai					37.132 MHz
gearbox_fault_diagnosis_ai_npu					9.6532 MHz
generic_timeseries_anomalydetection						37.6120 MHz, 1 run	67.132 MHz, 1 run	10.6200 MHz, 1 run		4.72400 MHz
generic_timeseries_anomalydetection-xip										6.52400 MHz
generic_timeseries_anomalydetection-xip-l2-64k										4.74400 MHz
generic_timeseries_classification							41.532 MHz, 1 run	0.653200 MHz, 1 run		3.39400 MHz
generic_timeseries_classification-xip										3.49400 MHz
generic_timeseries_classification-xip-l2-64k										3.39400 MHz
generic_timeseries_forecasting						8.09120 MHz, 1 run	1032 MHz, 1 run	1.92200 MHz, 1 run		0.914400 MHz
generic_timeseries_forecasting-xip										7.17400 MHz
generic_timeseries_forecasting-xip-l2-64k										6.29400 MHz
generic_timeseries_regression						0.216120 MHz, 1 run	0.48932 MHz, 1 run	0.0887200 MHz, 1 run		0.704400 MHz
generic_timeseries_regression-xip										1.5400 MHz
generic_timeseries_regression-xip-l2-64k										0.712400 MHz
hello_world_ai	125.2 to 126.732 MHz, 1 run			126.232 MHz	126.332 MHz
hello_world_ai_npu					1332 MHz
hvac_indoor_temp_forecast						1.04120 MHz, 1 run		0.315200 MHz, 1 run		0.117400 MHz
hvac_indoor_temp_forecast-xip										0.147400 MHz
hvac_indoor_temp_forecast-xip-l2-64k										0.117400 MHz
jerk_detection
motor_bearing_fault_anomaly_detection_ai	13.180 MHz, 1 run
motor_bearing_fault_detection_ai	175.7 to 176.180 MHz, 1 run				106.880 MHz
motor_bearing_fault_detection_ai_npu					1.2880 MHz
motor_fan_blade_fault_ai					1.7680 MHz
motor_fan_blade_fault_ai_npu					0.18280 MHz
motor_fault						6.26120 MHz, 1 run	23.432 MHz, 1 run	0.412200 MHz, 1 run		0.96400 MHz
motor_fault-xip										0.96400 MHz
motor_fault-xip-l2-64k										0.96400 MHz
pir_detection_ai							65.332 MHz, 1 run
torque_measurement						5.6120 MHz, 1 run	16.732 MHz, 1 run	2.59200 MHz, 1 run		1.2400 MHz
torque_measurement-xip										1.64400 MHz
torque_measurement-xip-l2-64k										1.2400 MHz
wakeword-frame-o0					27.580 MHz, 1 run
wakeword-frame-o2					24.180 MHz, 1 run
wakeword-lpc-frame-o2					8.580 MHz, 1 run
wakeword-npu-run-o2					11.280 MHz, 1 run
washing_machine_load_weighing						24.9120 MHz, 1 run		31200 MHz, 1 run		15.2400 MHz
washing_machine_load_weighing-xip										16.3400 MHz
washing_machine_load_weighing-xip-l2-64k										15.2400 MHz
waveform_classifier_ai					54.3 to 54.480 MHz		12.4160 MHz, 1 run
waveform_classifier_ai_npu					5.86 to 6.0280 MHz
(one way)		Not for this chip							Not for this chip

Sources

am13e230/tools/build_ai.py, cc2755p10/README.md, data/benchmark-suite.json, docs/experiments/2026-10-08-mspm0c1104-1kb-ram.md

Presence model: windows classed right in a timed room protocol (empty, person still, person moving, empty; first 20 s of each phase not scored) %, higher is better

Variant	LP-EM-CC35X1
presence-lab-moving-variance
presence-lab-pca
presence-lab-ti-model
wifi-csi-all	Not for this chip
wifi-csi-empty	0.7160 MHz, 1 run
wifi-csi-empty-guest-33min	99.5160 MHz, 1 run
wifi-csi-empty-overnight	100160 MHz, 1 run
wifi-csi-moving	Not for this chip
wifi-csi-still
wifi-presence-and-2x2-both	84.2160 MHz, 1 run
wifi-presence-and-2x3	85.1160 MHz, 1 run
wifi-presence-empty-pass-2x2-both	0160 MHz, 1 run
wifi-presence-empty-pass-2x3	0160 MHz, 1 run
wifi-presence-or-2x2-both	34.4160 MHz, 1 run
wifi-presence-or-2x3	34.6160 MHz, 1 run
wifi-presence-radar-2x2-both	84.2160 MHz, 1 run
wifi-presence-radar-2x3	85.1160 MHz, 1 run
wifi-presence-radar-confirms-wifi-2x2-both	78160 MHz, 1 run
wifi-presence-radar-confirms-wifi-2x3	78.3160 MHz, 1 run
wifi-presence-wifi-2x2-both	34.4160 MHz, 1 run
wifi-presence-wifi-2x3	34.6160 MHz, 1 run

Sources

docs/experiments/2026-10-09-cc35x1-wifi-csi-presence.md

Presence model: host asks for the stream to the first result s, lower is better

Variant	LP-EM-CC35X1
matter-csi-join-to-first-window	6160 MHz, 1 run
matter-csi-window-period	2.08 to 2.2160 MHz, 1 run
wifi-csi	7.4160 MHz, 1 run

Sources

docs/experiments/2026-10-09-cc35x1-wifi-csi-presence.md, docs/experiments/2026-10-12-cc35x1-matter-plus-wifi-presence.md

Deep-learning model: one inference, median wall time ms, lower is better

Variant	SK-TDA4VM
deeplabv3-mobv2-ade20k-mlperf-512-tflitert-tidl-c7x	12.1
Segmentation (DeepLabV3-Lite), ONNX Runtime, on the A72 cores	714.1
Segmentation (DeepLabV3-Lite), ONNX Runtime, on the C7x	7.3
Segmentation (DeepLabV3-Lite, 768x432), ONNX Runtime, on the A72 cores	900.5
Segmentation (DeepLabV3-Lite, 768x432), ONNX Runtime, on the C7x	11.8
mobilenetv1-mlperf-224-tflitert-tidl-c7x	2.17
Image classification (RegNetX 200MF), ONNX Runtime, on the A72 cores	37.4
Image classification (RegNetX 200MF), ONNX Runtime, on the C7x	2.89
Object detection (SSD-Lite MobileNetV2), ONNX Runtime, on the A72 cores	438.5
Object detection (SSD-Lite MobileNetV2), ONNX Runtime, on the C7x	6.1
ssdlite-mobdet-320-tflitert-tidl-c7x	6.4
tvm-cl-3090-mobilenetv2-tv-tvmdlr	2.93 to 3.41
tvm-od-5120-ssdlite-mobdet-dsp-coco-320x320-tvmdlr	13 to 13.3
tvm-ss-5710-deeplabv3lite-mobv2-cocoseg21-512x512-tvmdlr	8.85 to 9.08
Object detection, small (lite), ONNX Runtime, on the A72 cores	193
Object detection, small (lite), ONNX Runtime, on the C7x	4.66
Object detection, medium, ONNX Runtime, on the A72 cores	1,523.1
Object detection, medium, ONNX Runtime, on the C7x	11.7
Face detection, ONNX Runtime, on the A72 cores	1,504
Face detection, ONNX Runtime, on the C7x	11.2
Pose estimation, ONNX Runtime, on the A72 cores	1,767
Pose estimation, ONNX Runtime, on the C7x	13.8

Sources

Run records are in the notebook, not this copy.

Whole vision graph: decode, preprocess, inference, postprocess, sink fps, higher is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
file-pose-encode	17.8 to 18.1
file-pose-encode-video0-30fps	301 run
file-ssd-lite-512-zerocopy-fakesink	91.61 run
file-yolox-nano-416-plain-fakesink	56.2
file-yolox-nano-416-zerocopy-fakesink	110.8
kmscube-1080p60-llvmpipe		301 run
three-pipelines-yolox-c7x	216.41 run
two-pipelines-yolox-plus-ssd-c7x	163.61 run
two-pipelines-yolox-twice-c7x	182.41 run
videotestsrc-ball-720p30-kmssink		1.841 run
videotestsrc-smpte-1080p60-kmssink		5.921 run
videotestsrc-smpte-720p60-kmssink		27.11 run

Sources

am62l-evm/README.md

TI's wake-word model ("OK Kilby"): utterances detected, of the number said (20 unless the conditions say 10), one Mac voice at one level hits, higher is better

Variant	LP-MSPM0G5187
daniel-live-mic-vol70
daniel-replay-m20dbfs	2080 MHz
fred-replay-m20dbfs	2080 MHz
karen-replay-m20dbfs	2080 MHz
kathy-replay-m20dbfs	2080 MHz
ralph-replay-m20dbfs	2080 MHz
samantha-live-mic-vol70
samantha-replay-m10dbfs	1080 MHz
samantha-replay-m20dbfs	2080 MHz
samantha-replay-m30dbfs	1080 MHz
samantha-replay-m3dbfs	1080 MHz
samantha-replay-m40dbfs	1080 MHz

Sources

mspm0g5187/wakeword/run_live.sh

TI's wake-word model: WAKE events in 10 minutes of silence or of other sentences spoken by the Mac per 10 min, lower is better

Variant	LP-MSPM0G5187
live-silence
live-speech
replay-silence	080 MHz, 1 run
replay-speech-near-miss	1880 MHz, 1 run
replay-speech-plain	080 MHz, 1 run

Sources

mspm0g5187/wakeword/run_live.sh

TI's wake-word model: end of the spoken phrase to the board's WAKE line, median s, lower is better

Variant	LP-MSPM0G5187
daniel-replay-m20dbfs	0.0380 MHz
fred-replay-m20dbfs	0.3680 MHz
karen-replay-m20dbfs	0.4980 MHz
kathy-replay-m20dbfs	0.4480 MHz
ralph-replay-m20dbfs	0.3880 MHz
samantha-live-mic-vol70
samantha-replay-m20dbfs	0.1880 MHz

Sources

mspm0g5187/wakeword/run_live.sh

Voice assistant turn: end of the spoken question to the board's first reply audio, median s, lower is better

Variant	LP-EM-CC35X1
xiaozhi-claude-haiku	2.47 to 2.67160 MHz

Sources

Run records are in the notebook, not this copy.

TI's wake-word example as built for the board's own microphone: flash and RAM (text plus data, bss) from the linker size report B, lower is better

Variant	LP-MSPM0G5187
flash-i2s-o2	66.8 kB80 MHz, 1 run
flash-shipped	59.7 kB1 run
ram-i2s-o2	25 kB80 MHz, 1 run
ram-shipped	24.4 kB1 run

Sources

Run records are in the notebook, not this copy.

Samples a second the wake-word example's audio input really delivers (the model was trained at 8,000), counted by the board against the Mac clock Hz, higher is better

Variant	LP-MSPM0G5187
adc-shipped-o2	6,63080 MHz, 1 run
i2s-microphone	7,74980 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

One block of a fixed-weight DSP graph on the TinyEngine NPU or the CPU (4 channels x 256 outputs unless the variant says) ms, lower is better

Variant	LP-MSPM0G5187
bank8-c-hand-8ch	22.780 MHz, 1 run
bank8-nnc-cpu	36.780 MHz, 1 run
bank8-nnc-npu	0.96780 MHz, 1 run
bank8_trip-nnc-cpu	33.380 MHz, 1 run
bank8_trip-nnc-npu	0.67780 MHz, 1 run
bank8_trip2-nnc-cpu	33.380 MHz, 1 run
bank8_trip2-nnc-npu	0.69480 MHz, 1 run
bank8_trip3-nnc-cpu	33.780 MHz, 1 run
bank8_trip3-nnc-npu	0.71380 MHz, 1 run
bank8_trips-nnc-cpu	33.380 MHz, 1 run
bank8_trips-nnc-npu	0.67780 MHz, 1 run
dft64-nnc-cpu	0.8380 MHz, 1 run
dft64-nnc-npu	0.19580 MHz, 1 run
fir64-c-hand-1ch	3.2980 MHz, 1 run
fir64-c-hand-4ch	12.280 MHz, 1 run
fir64-nnc-cpu	16.880 MHz, 1 run
fir64-nnc-npu	0.56680 MHz, 1 run
fir64_dec4-nnc-cpu	4.380 MHz, 1 run
fir64_dec4-nnc-npu	0.18580 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

NPU time against the same graph as NNC CPU code, or against hand-written C x, higher is better

Variant	LP-MSPM0G5187
bank8-npu-vs-c-hand-8ch	23.5 calc.
bank8-npu-vs-nnc-cpu	38 calc.
bank8_trip-npu-vs-nnc-cpu	49.1 calc.
bank8_trip2-npu-vs-nnc-cpu	48 calc.
bank8_trip3-npu-vs-nnc-cpu	47.2 calc.
bank8_trips-npu-vs-nnc-cpu	49.1 calc.
dft64-npu-vs-nnc-cpu	4.3 calc.
fir64-npu-vs-c-hand-4ch	21.6 calc.
fir64-npu-vs-nnc-cpu	29.7 calc.
fir64-npu4ch-vs-c-hand-1ch	5.8 calc.
fir64_dec4-npu-vs-nnc-cpu	23.3 calc.

Sources

Run records are in the notebook, not this copy.

Signal to error ratio of the 8-bit graph against the float reference dB, higher is better

Variant	LP-MSPM0G5187
fir64-ch0-total	34.7 calc.
fir64-ch0-weights-only	44.3 calc.
fir64-ch1-total	34.6 calc.
fir64-ch1-weights-only	53.5 calc.
fir64-ch2-total	35.1 calc.
fir64-ch2-weights-only	42.6 calc.
fir64-ch3-total	35.4 calc.
fir64-ch3-weights-only	47 calc.

Sources

Run records are in the notebook, not this copy.

Images whose output equals the reference count, higher is better

Variant	LP-MSPM0G5187
16-images-vs-reference	1680 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Flash text of the whole image with harness and NPU library B, lower is better

Variant	LP-MSPM0G5187
bank8-nnc-cpu-ram	20.4 kB1 run
bank8-nnc-npu-flash	8.3 kB1 run
bank8-nnc-npu-ram	8.1 kB1 run
bank8_trip-nnc-npu-flash	6.9 kB1 run
bank8_trip-nnc-npu-ram	6.1 kB1 run
dft64-nnc-npu-flash	9.9 kB1 run
dft64-nnc-npu-ram	2.6 kB1 run
fir64-nnc-npu-flash	7.9 kB1 run
fir64-nnc-npu-ram	4.9 kB1 run
fir64_dec4-nnc-npu-flash	6.1 kB1 run
fir64_dec4-nnc-npu-ram	3.4 kB1 run

Sources

mspm0g5187/npu-dsp/build.sh

Energy per block on the NPU against the CPU uJ, lower is better

Variant	LP-MSPM0G5187
(one way)

Sources

mspm0g5187/npu-dsp/README.md

Streaming with the taps-1 overlap carried between blocks works yes/no

Variant	LP-MSPM0G5187
(one way)

Sources

mspm0g5187/npu-dsp/README.md

Share of CPU time left free while the NPU runs a block %, higher is better

Variant	LP-MSPM0G5187
(one way)

Sources

mspm0g5187/npu-dsp/README.md

The same graph on the LP-AM13E230 yes/no

Variant	LP-MSPM0G5187
(one way)

Sources

mspm0g5187/npu-dsp/README.md

Repeat runs without a reflash count, higher is better

Variant	LP-MSPM0G5187
(one way)

Sources

mspm0g5187/npu-dsp/README.md

Instructions per generated token (emulator count) instructions, lower is better

Variant	LP-MSPM33C321A
tinystories260k-int8-gs4-o2-emulator	4.81M to 5.14M1 run

Sources

Run records are in the notebook, not this copy.

Deepest stack use while generating B, lower is better

Variant	LP-MSPM33C321A
tinystories260k-int8-gs4-o2-emulator	1.7 kB1 run

Sources

ports/llama2c/README.md

Generated token ids equal to a reference build (greedy, temperature 0) count, higher is better

Variant	LP-MSPM33C321A
int8-m33-emulator-vs-mac-greedy	1281 run
int8-m33-emulator-vs-mac-seed42	391 run
int8-port-vs-runq-c-mac	128
int8-vs-float32-mac-greedy	1281 run

Sources

ports/llama2c/tools/check_host.sh

Tokens per second of a small language model tokens/s, higher is better

Variant	LP-MSPM33C321A
llama2c-on-edumkii-screen-display-on	19.1160 MHz, 1 run
llama2c-on-edumkii-screen-drawing-excluded	21.1160 MHz, 1 run
tinystories260k-int8-gs4-o2-board	21.2160 MHz
tinystories260k-int8-gs4-o2-estimate-160mhz	25.8 calc.
tinystories260k-weights-in-qspi

Sources

ports/llama2c/README.md, ports/llama2c/run_board.py

Seconds between stories when the model writes to the screen unattended s

Variant	LP-MSPM33C321A
llama2c-on-edumkii-screen	30160 MHz, 1 run

Sources

ports/llama2c/README.md

The S2 button starts a new story on the screen yes/no

Variant	LP-MSPM33C321A
llama2c-on-edumkii-screen

Sources

ports/llama2c/README.md

Model from the training tool's own file, one inference, plain C from tools/byom ms, lower is better

Variant	LP-MSPM0G5187	LP-AM13E230
dtree-c-o3		0.0014 to 0.0022200 MHz
forest-c-o3		0.0243 to 0.0349200 MHz
occ_float_2k-c-o3		1.19200 MHz
occ_float_2k-fast-c-o3		0.846200 MHz
occ_float_2k-fastreg-c-o3		0.938200 MHz
occ_float_30k-c-o3		22.5200 MHz
occ_float_30k-fast-c-o3		13.6200 MHz
occ_float_30k-fastreg-c-o3		14.7200 MHz
occ_float_500-c-o3		0.392200 MHz
occ_float_500-fast-c-o3	29.1 to 33.780 MHz	0.302200 MHz
occ_float_500-fastreg-c-o3		0.344200 MHz
occ_float_8k-c-o3		5.58200 MHz
occ_float_8k-fast-c-o3		3.7200 MHz
occ_float_8k-fastreg-c-o3		3.93200 MHz
occ_int8own_2k-c-o3		1.23200 MHz
occ_int8own_30k-c-o3		20200 MHz
occ_int8own_30k-cmsis-c-o3		8.54200 MHz
occ_int8own_500-c-o3		0.447200 MHz
occ_int8own_500-cmsis-c-o3		0.491200 MHz
occ_int8own_8k-c-o3		5.39200 MHz
occ_int8own_8k-cmsis-c-o3		2.89200 MHz
occ_tflite_float-fast-c-o3		0.304200 MHz
resnet8-float-c-o3		Not for this chip
resnet8-int8own-c-o3		444.6200 MHz
resnet8-int8own-cmsis-c-o3		201.6200 MHz
xgb-c-o3	1.01 to 1.3680 MHz	0.0801 to 0.113200 MHz
xgb-onnx-c-o3		0.0802 to 0.113200 MHz

Sources

Run records are in the notebook, not this copy.

Vectors whose class matches the training tool, model through tools/byom of 20, higher is better

Variant	LP-MSPM0G5187	LP-AM13E230
dtree-c-o3		20200 MHz
forest-c-o3		20200 MHz
occ_float_2k-c-o3		20200 MHz
occ_float_2k-fast-c-o3		20200 MHz
occ_float_2k-fastreg-c-o3		20200 MHz
occ_float_30k-c-o3		20200 MHz
occ_float_30k-fast-c-o3		20200 MHz
occ_float_30k-fastreg-c-o3		20200 MHz
occ_float_500-c-o3		20200 MHz
occ_float_500-fast-c-o3	2080 MHz	20200 MHz
occ_float_500-fastreg-c-o3		20200 MHz
occ_float_8k-c-o3		20200 MHz
occ_float_8k-fast-c-o3		20200 MHz
occ_float_8k-fastreg-c-o3		20200 MHz
occ_int8own_2k-c-o3		19200 MHz
occ_int8own_30k-c-o3		20200 MHz
occ_int8own_30k-cmsis-c-o3		20200 MHz
occ_int8own_500-c-o3		20200 MHz
occ_int8own_500-cmsis-c-o3		20200 MHz
occ_int8own_8k-c-o3		20200 MHz
occ_int8own_8k-cmsis-c-o3		20200 MHz
occ_tflite_float-fast-c-o3		20200 MHz
resnet8-int8own-c-o3		20200 MHz
resnet8-int8own-cmsis-c-o3		20200 MHz
xgb-c-o3	2080 MHz	20200 MHz
xgb-onnx-c-o3		20200 MHz

Sources

Run records are in the notebook, not this copy.

MLPerf Tiny image classification (ResNet-8, CIFAR-10), one 32x32 image ms, lower is better

Variant	LP-AM13E230
resnet8-ti-qat-nnc-cpu	Not for this chip
resnet8-ti-qat-nnc-npu	198.2200 MHz

Sources

Run records are in the notebook, not this copy.

MLPerf Tiny ResNet-8: board images whose class matches the host run of the same int8 file of 20, higher is better

Variant	LP-AM13E230
resnet8-ti-qat-nnc-npu	20200 MHz

Sources

Run records are in the notebook, not this copy.

Every number

Best result per test, the scorecard, what runs where and every table, each folded.

Best result per test

4.44

Most work per clock tick (CoreMark per MHz)

LP-EM-CC2755P10, TI compiler, no OS, lowest clock run

150 ns

Quickest interrupt response

LAUNCHXL-F28P65X, no OS, signal to first line of the handler, code in RAM

1.33 µs

Shortest control loop

LP-AM13E230, ADC trigger to new PWM value

5 Msps

Fastest ADC

LP-MSPM33C321A, DMA

131.5 times

Biggest AES gain from hardware

LP-EM-CC2340R5, TI's AES driver against software AES, 1 kB

5.05 ms

Fastest medium room-occupancy model, microcontroller CPU

LP-AM13E230

0.63 ms

Same model on a microcontroller NPU

LP-AM13E230

The scorecard: one number per board per test

Measured	LP-MSPM0G3507Cortex-M0+, 80 MHz, 32 kB RAM	LP-MSPM0L1306Cortex-M0+, 32 MHz, 4 kB RAM	LP-EM-CC2340R5Cortex-M0+, 48 MHz, 36 kB RAM	LP-CC1354P10-1Cortex-M33, 48 MHz, 296 kB RAM	LAUNCHXL-F2800137C28x, 120 MHz, 36 kB RAM	LP-AM13E230Cortex-M33 with TMU and NPU, 200 MHz, 128 kB RAM	LP-AM263P4x Cortex-R5F with TMU, 400 MHz, 3 MB RAM	LP-EM-CC2755P10Cortex-M33, 96 MHz, 162 kB RAM	LP-EM-CC35X1Cortex-M33, 160 MHz, 1.1 MB RAM	ESP32-C6-DevKitC-1-N8RISC-V, 160 MHz, 512 kB RAM	LAUNCHXL-F28P65XC28x, 200 MHz, 248 kB RAM	FRDM-MCXA153Cortex-M33, 96 MHz, 32 kB RAM	IWRL6432BOOSTCortex-M4F, 160 MHz, 1 MB RAM	LP-MSPM0C1104Cortex-M0+, 24 MHz, 1 kB RAM	LP-MSPM0G3519Cortex-M0+, 80 MHz, 128 kB RAM	LP-MSPM0G5187Cortex-M0+ with NPU, 80 MHz, 32 kB RAM	LP-MSPM33C321ACortex-M33, 160 MHz, 256 kB RAM	NUCLEO-U083RCCortex-M0+, 56 MHz, 40 kB RAM	Raspberry Pi Pico 2 HCortex-M33, 150 MHz, 520 kB RAM	SK-TDA4VM2x Cortex-A72 + C7x/MMA + 2x C66x + 6x R5F, 2000 MHz, 4 GB RAM
Speed
CoreMark per MHz, TI compiler, no OS, lowest clock runhigher is better; the lowest clock run, so flash wait states cost the least (clock under the number); the C2000 build is unofficial, not ranked	2.46 Build details ticlang-bare-o3-32mhzat 32 MHz	2.63 Build details ticlang-bare-o3-4mhzat 4 MHz	2.31worst Build details ticlang-bare-o3	4.17 Build details ticlang-bare-o3	1.18 Build details c28patch-ti-c2000-o4s5-flash unofficial build, not ranked	4.42 Build details ticlang-bare-o3-100mhzat 100 MHz		4.44best Build details ticlang-bare-o3			1.06 Build details c28patch-ti-c2000-o4s5-flash unofficial build, not ranked			Not for this chip	2.46 Build details ticlang-bare-o3-32mhzat 32 MHz	2.46 Build details ticlang-bare-o3-32mhzat 32 MHz	4.38 Build details ticlang-bare-o3-32mhzat 32 MHz
CoreMark per MHz, TI compiler, no OS, top clockhigher is better; each board at the clock its run used, usually its top clock, where flash wait states cost cycles	2.3 Build details ticlang-bare-o3	2.45 Build details ticlang-bare-o3	2.31 Build details ticlang-bare-o3	4.17 Build details ticlang-bare-o3	1.18 Build details c28patch-ti-c2000-o4s5-flash unofficial build, not ranked	4.36 Build details ticlang-bare-o3		4.44best Build details ticlang-bare-o3			1.06 Build details c28patch-ti-c2000-o4s5-flash unofficial build, not ranked			Not for this chip	2.28worst Build details ticlang-bare-o3	2.29 Build details ticlang-bare-o3	3.4 Build details ticlang-bare-o3
CoreMark per MHz, GCC under Zephyrhigher is better	1.75 Build details gcc-zephyr-o2unroll	1.87best Build details gcc-zephyr-o2unroll	1.75 Build details gcc-zephyr-o2unroll												1.69worst Build details gcc-zephyr-o2unroll
Dhrystone, DMIPS per MHzhigher is better		0.731	0.602worst Build details ticlang-bare-o3-noinline	1.2 Build details ticlang-bare-o3-noinline	0.442 Build details ti-c2000-o3-noinline-ram C28x: 16-bit bytes, not ranked	1.05 Build details ticlang-bare-o3-noinline	1.5 Build details ticlang-bare-o3-noinline	1.32 Build details ticlang-bare-o3-noinline	1.34 Build details ticlang-freertos-o3-noinline		0.444 Build details ti-c2000-o3-noinline-ram C28x: 16-bit bytes, not ranked	0.835 Build details gcc-bare-o3		Not for this chip	0.652 Build details ticlang-bare-o3-noinline	0.651 Build details ticlang-bare-o3-noinline	1.22 Build details ticlang-bare-o3-noinline-32mhzat 32 MHz	0.672 Build details at 48 MHz	1.32 Build details gcc-bare-o3-noinline-m33	4.83best Build details gcc-linux-o3-noinline
Embench programs that fit and pass, of 19higher is better	18	13	18 Build details ticlang-o2	19best Build details ticlang-o2	15 Build details ti-c2000-o2-flash	19best Build details ticlang-o2		19best Build details ticlang-o2	Not for this chip			19best		9worst	18	18	19best Build details ticlang-o2	19best Build details at 48 MHz	19best Build details gcc-o2-m33
Reaction
Interrupt latency, no OS, cycleslower is better; signal to the first line of the handler	29 Build details bare	25 Build details bare	28 Build details bare	23 Build details bare	30 Build details bare C28x: different core family, not ranked		195worst Build details bare	21 Build details bare				17best Build details bare		23 Build details bare	31 Build details bare	36 Build details bare	32 Build details bare	22 Build details bareat 48 MHz
Interrupt latency under Zephyr, cycleslower is better	163worst Build details zephyr	119 Build details zephyr	48best Build details zephyr												157 Build details zephyr			108 Build details zephyrat 48 MHz
GPIO toggle by register write, cycleslower is better	1best Build details register	1best Build details register	2.1 Build details register-bare-o3	1.51 Build details register-bare-o3	1.56 Build details register		10 Build details register	3.99 Build details register-bare-o3		38worst Build details register-bare-o3	1.56 Build details register	1.12 Build details register		2 Build details register-bare-o3	1best Build details register	1best Build details register	8 Build details register-bare-o3	2.5 Build details register-bare-o3at 48 MHz
Control loop, trigger to new output, µslower is better; the same run per board as the Control loops chart: code in flash, from the ADC end-of-conversion interrupt where the board has that run (build named under the number)	88.5 Build details hw	448.1worst Build details sw-eoc-flash	336.6 Build details sw-eoc-flash-os	21.4 Build details sw-eoc-flash	2.58 Build details hw-eoc-flash	1.33best Build details hw-tmu-eoc-flash		10.5 Build details sw-eoc-flash-o3							89.3 Build details hw	180.8 Build details sw-o3	5.03 Build details sw-eoc-flash-o3	298.8 Build details sw-eoc-flashat 48 MHz
One control step, software float, cycleslower is better; the same C on every board	14k Build details sw	13k Build details sw-flash	14kworst Build details sw-flash	671.4 Build details sw-flash	400.2best Build details sw-flash	655.6 Build details sw-flash		593.7 Build details sw-flash				5.4k Build details sw-flash			14k Build details sw	14k Build details sw	531.1 Build details sw-flash	12k Build details sw-flashat 48 MHz
Radios
Bluetooth on the air after flashing, slower is better			1.4worst Build details default-100ms-interval					0.9best Build details default-100ms-interval
Bluetooth echo, 20 bytes, mslower is better			45worst		n/a			29.2best
Zigbee light command from Home Assistant, mslower is better; median of 20			95 Build details zigbee-home-assistant
Zigbee join, slower is better			6.2 Build details zigbee-router-to-cc1354p10-coordinator	6 Build details zigbee-zed-zha
Thread join, slower is better			15.5worst Build details thread-router-to-cc1354p10-leader	8best Build details thread-leader-fresh-image
Thread node on the chipend device or router running on the chip itself			Not for this chip	Works	n/a					Works
Thread radio (RCP) for a host			Works	Works	n/a
Bluetooth and Zigbee or TI 15.4 at once			Not for this chip	Works	n/a			Works
Matter			Not for this chip		n/a				Works
Memory
RAM left beside TI's Bluetooth example, byteshigher is better			5.1 kBworst Build details ble-basic-ble					81 kBbest Build details ble-basic-ble
RAM left beside TI's Zigbee example, byteshigher is better; router on the CC2340R5, end device on the CC1354P10			799worst Build details zigbee-router	210.6 kB Build details zigbee-zed-ti-example				126.5 kB Build details zigbee-router		360.1 kBbest Build details zigbee-router
Blink image RAM, no OS, byteslower is better	512 Build details bare	256best Build details bare	9.9 kB Build details bare	18.6 kB Build details bare TI's default configuration, not ranked	1.5 kB Build details bare	9 kB Build details bare	93.8 kBworst Build details bare	10.1 kB Build details bare								256best Build details bare	256best Build details bareat 32 MHz
Blink image RAM under Zephyr, byteslower is better	3.9 kB Build details zephyr	3.8 kBbest Build details zephyr	4 kB Build details zephyr									3.8 kB Build details zephyr			4 kB Build details zephyr			4.8 kBworst Build details zephyrat 48 MHz
Blink image RAM under FreeRTOS, byteslower is better	1 kBbest Build details freertosat 32 MHz	1 kBbest Build details freertos	13 kB Build details freertos	37.6 kB Build details freertos TI's default configuration, not ranked					36.9 kBworst Build details freertos					1 kBbest Build details freertos
RTOS task switch, µslower is better; FreeRTOS or Zephyr, whichever ran	6.72 Build details zephyr	12 Build details freertos	10.6 Build details freertos	7.38 Build details freertos	Not for this chip	1.49best Build details freertos		3.5 Build details freertos		16.5 Build details freertos		2.76 Build details freertos		19.9worst Build details freertos	6.99 Build details zephyr	7 Build details freertos	11.5 Build details freertosat 32 MHz	9.29 Build details freertosat 48 MHz
Small AI models
Room occupancy, medium, on the CPU, mslower is better; 8.5k weights, fastest plain-C build	35.3	83.7worst	63.2	25.8	17.5	5.05		10.5	6.25	7.14	8.38	12.6		Not for this chip	35.3	35.3	7.38		6.82	0.193best
Room occupancy, medium, on an NPU, mslower is better; 8.5k weights, TI NNC code on the chip's neural processing unit						0.63best										2.12worst
Room occupancy, large, fit33k weights: does it fit and pass	Works	Not for this chip	Works	Works	Works	Works		Works	Works	Works	Works	Works		Not for this chip	Works	Works	Works	Works
Handwritten digit, BitNetMCU 64-wide model, cycleslower is better	2.64M Build details cnn64-ticlang-o3-32mhzat 32 MHz	2.65M Build details cnn64-ticlang-o3-32mhz			2.55M Build details cnn-64-c2000-o4s5-120mhz-ram										2.64M Build details cnn64-ticlang-o3-32mhzat 32 MHz	2.64M Build details cnn64-ticlang-o3-32mhzat 32 MHz
Hardware helpers
sin and cos in hardware against software, times fasterhigher is better	85.6	n/a	n/a	n/a	9.87worst									n/a	93.3best	n/a		n/a
AES in hardware against software, times fasterhigher is better	20.3 Build details 1kb-ticlang-o2		131.5best Build details 1kb-ecb-ticlang-o3-driver-polling	51.1 Build details 1kb-ecb-ticlang-o2-driver-polling		29.9 Build details 1kb-ecb-ticlang-o2-keyout		92.8 Build details ticlang-o2-driver-polling		1.23worst Build details 16b-ecb-mbedtls					45.8 Build details 1kb-ticlang-o2	45.5 Build details 1kb-ticlang-o2	41.9 Build details 1kb-ecb-ticlang-o2-keyoutat 32 MHz	93.9 Build details 1kb-aes256-ecb-gcc12-o2-regat 48 MHz
CRC in hardware against a table, times fasterhigher is better	13.5best	10.9				1.89worst						7.06		3 Build details crc16-256b	8.09	8.12	2.75 Build details at 32 MHz	8.34 Build details gcc12-o3at 48 MHz	5.71 Build details dma-sniffer-riscv
256-point FFT, mslower is better; float where the chip has an FPU		25.7worst Build details f32	18.3 Build details f32	0.473 Build details f32	0.107best Build details f32	0.117 Build details f32		0.224 Build details f32				4.69 Build details f32		Not for this chip	11.1 Build details f32	11 Build details f32	0.204 Build details f32	20 Build details f32at 48 MHz
Fastest ADC reading, million samples per secondhigher is better; each board's best setting. Under each number: the ADC's clock when it differs from the CPU's, DMA or not, and resolution and sample window where recorded	3.93 Build details at 32 MHzCPU at 32 MHz, ADC clock as in the example	1.45 Build details 8-cycle window	1.2 Build details ti-adcbuf-dmaDMA	0.2 Build details ti-adcbuf-dmaDMA	3.98 Build details ti-continuous-9cyc9-cycle window			1.2 Build details ti-adcbuf-dmaDMA	0.17worst Build details 12-bit			2.53 Build details dmaDMA12-bit		1 Build details DMA10-bit, 2-cycle window	3.93 Build details at 32 MHzCPU at 32 MHz, ADC clock as in the example	0.547 Build details at 32 MHzCPU at 32 MHz, ADC clock as in the example12-bit, 62.5 ns window	5best Build details DMA	1.71 Build details internalCPU at 48 MHz, ADC clock as in the exampleDMA12-bit	0.5 Build details pico-sdk-dma-riscvDMA
Power
Deepest sleep with RAM kept, µAlower is better; probe-measured chip figure; modes differ by chip (named under the number)		6.2 Build details STANDBY0, RAM kept			2,590 Build details ti-c2000-standby-borrowed-energytrace STANDBY with the PLL on; HALT is the reference floor; not ranked
Running current per MHz, µAlower is better; slope between two clocks; the workload differs by chip (named under the number)		72.9 Build details energytrace-3.3v-coremark			263 Build details ti-c2000-borrowed-energytrace busy loop at two clocks, not CoreMark; not ranked
Getting started
Plug-in to a changed blink, slower is better; the CC2340R5 figure includes 8 min lost to a half-seated probe	149 Build details ccs-mcp		540worst Build details cli	29 Build details cli		60 Build details ccs-mcp	78 Build details cli	198 Build details ccs-mcp	60 Build details cli	184 Build details cli		1.4best Build details cliat 12 MHz	Not for this chip	77 Build details cli			19 Build details cli	14 Build details cliat 48 MHz	87 Build details cli
Blink build, slower is better	6 Build details ccs-mcp	3.3 Build details cli	22 Build details cli		4.6 Build details ti-led-ex1-blinky-cli	2.7best Build details cli	5.9 Build details cli	14 Build details cli	6 Build details cli	67worst Build details cli	8.6 Build details ti-led-ex1-blinky-cli	30 Build details cliat 12 MHz		3.5 Build details cli	3.2 Build details cliat 32 MHz	4 Build details cliat 32 MHz	4.8 Build details cliat 32 MHz	6 Build details cliat 48 MHz	12 Build details cli
Blink flash, slower is better	6 Build details west	0.6best Build details openocd	7 Build details uniflash		3.9 Build details ccs-scripting-ram-load	3.3 Build details uniflash		27worst Build details ccs-mcp						0.64 Build details openocd	0.6best Build details openocdat 32 MHz	20.7 Build details openocdat 32 MHz	11.3 Build details uniflashat 32 MHz	1.2 Build details openocdat 48 MHz
Examples safe to run that ran with no crash, %higher is better; of the images safe to run, not the whole SDK: 213 of 270 on the G3507, 148 of 170 on the L1306	100 Build details mspm0-sdk-2.11at 32 MHz	100 Build details mspm0-sdk-2.11	100 Build details cc35x1-twins-checked		100 Build details c2000ware-26.02	98.8 Build details am13e230-sdk-26.01	118best Build details mcu-plus-sdk-26.01-lp	98.8 Build details sdk-f3-9.21	93.5 Build details wifi-sdk-10.20.00.39	87.1 Build details esp-idf-v6.1		97.7 Build details mcux-26.09.00		100 Build details mspm0-sdk-2.11	100 Build details mspm0-sdk-2.11	90.8 Build details mspm0-sdk-2.11	99.4 Build details mspm33-sdk-1.04at 32 MHz	100 Build details stm32cubeu0-1.3.0at 48 MHz		60worst Build details edgeai-gst-apps-11.00.00.08
Full redraw of a 3.5 inch SPI screen, mslower is better; fastest setting	67.2 Build details spi40-dma16														67.2 Build details spi40-dma16

Speed

CoreMark per MHz, TI compiler, no OS, lowest clock run

higher is better; the lowest clock run, so flash wait states cost the least (clock under the number); the C2000 build is unofficial, not ranked

LP-MSPM0G35072.46
Build details
ticlang-bare-o3-32mhzat 32 MHz
LP-MSPM0L13062.63
Build details
ticlang-bare-o3-4mhzat 4 MHz
LP-EM-CC2340R52.31worst
Build details
ticlang-bare-o3
LP-CC1354P10-14.17
Build details
ticlang-bare-o3
LAUNCHXL-F28001371.18
Build details
c28patch-ti-c2000-o4s5-flash
unofficial build, not ranked
LP-AM13E2304.42
Build details
ticlang-bare-o3-100mhzat 100 MHz
LP-EM-CC2755P104.44best
Build details
ticlang-bare-o3
LAUNCHXL-F28P65X1.06
Build details
c28patch-ti-c2000-o4s5-flash
unofficial build, not ranked
LP-MSPM0C1104Not for this chip
LP-MSPM0G35192.46
Build details
ticlang-bare-o3-32mhzat 32 MHz
LP-MSPM0G51872.46
Build details
ticlang-bare-o3-32mhzat 32 MHz
LP-MSPM33C321A4.38
Build details
ticlang-bare-o3-32mhzat 32 MHz

CoreMark per MHz, TI compiler, no OS, top clock

higher is better; each board at the clock its run used, usually its top clock, where flash wait states cost cycles

LP-MSPM0G35072.3
Build details
ticlang-bare-o3
LP-MSPM0L13062.45
Build details
ticlang-bare-o3
LP-EM-CC2340R52.31
Build details
ticlang-bare-o3
LP-CC1354P10-14.17
Build details
ticlang-bare-o3
LAUNCHXL-F28001371.18
Build details
c28patch-ti-c2000-o4s5-flash
unofficial build, not ranked
LP-AM13E2304.36
Build details
ticlang-bare-o3
LP-EM-CC2755P104.44best
Build details
ticlang-bare-o3
LAUNCHXL-F28P65X1.06
Build details
c28patch-ti-c2000-o4s5-flash
unofficial build, not ranked
LP-MSPM0C1104Not for this chip
LP-MSPM0G35192.28worst
Build details
ticlang-bare-o3
LP-MSPM0G51872.29
Build details
ticlang-bare-o3
LP-MSPM33C321A3.4
Build details
ticlang-bare-o3

CoreMark per MHz, GCC under Zephyr

higher is better

LP-MSPM0G35071.75
Build details
gcc-zephyr-o2unroll
LP-MSPM0L13061.87best
Build details
gcc-zephyr-o2unroll
LP-EM-CC2340R51.75
Build details
gcc-zephyr-o2unroll
LP-MSPM0G35191.69worst
Build details
gcc-zephyr-o2unroll

Dhrystone, DMIPS per MHz

higher is better

LP-MSPM0G3507
LP-MSPM0L13060.731
LP-EM-CC2340R50.602worst
Build details
ticlang-bare-o3-noinline
LP-CC1354P10-11.2
Build details
ticlang-bare-o3-noinline
LAUNCHXL-F28001370.442
Build details
ti-c2000-o3-noinline-ram
C28x: 16-bit bytes, not ranked
LP-AM13E2301.05
Build details
ticlang-bare-o3-noinline
LP-AM263P1.5
Build details
ticlang-bare-o3-noinline
TMDS62LEVM (AM62L EVM)
LP-EM-CC2755P101.32
Build details
ticlang-bare-o3-noinline
LP-EM-CC35X11.34
Build details
ticlang-freertos-o3-noinline
LAUNCHXL-F28P65X0.444
Build details
ti-c2000-o3-noinline-ram
C28x: 16-bit bytes, not ranked
FRDM-MCXA1530.835
Build details
gcc-bare-o3
LP-MSPM0C1104Not for this chip
LP-MSPM0G35190.652
Build details
ticlang-bare-o3-noinline
LP-MSPM0G51870.651
Build details
ticlang-bare-o3-noinline
LP-MSPM33C321A1.22
Build details
ticlang-bare-o3-noinline-32mhzat 32 MHz
NUCLEO-U083RC0.672
Build details
at 48 MHz
Raspberry Pi Pico 2 H1.32
Build details
gcc-bare-o3-noinline-m33
SK-TDA4VM4.83best
Build details
gcc-linux-o3-noinline

Embench programs that fit and pass, of 19

higher is better

LP-MSPM0G350718
LP-MSPM0L130613
LP-EM-CC2340R518
Build details
ticlang-o2
LP-CC1354P10-119best
Build details
ticlang-o2
LAUNCHXL-F280013715
Build details
ti-c2000-o2-flash
LP-AM13E23019best
Build details
ticlang-o2
LP-EM-CC2755P1019best
Build details
ticlang-o2
LP-EM-CC35X1Not for this chip
FRDM-MCXA15319best
LP-MSPM0C11049worst
LP-MSPM0G351918
LP-MSPM0G518718
LP-MSPM33C321A19best
Build details
ticlang-o2
NUCLEO-U083RC19best
Build details
at 48 MHz
Raspberry Pi Pico 2 H19best
Build details
gcc-o2-m33
SK-TDA4VM

Reaction

Interrupt latency, no OS, cycles

lower is better; signal to the first line of the handler

LP-MSPM0G350729
Build details
bare
LP-MSPM0L130625
Build details
bare
LP-EM-CC2340R528
Build details
bare
LP-CC1354P10-123
Build details
bare
LAUNCHXL-F280013730
Build details
bare
C28x: different core family, not ranked
LP-AM13E230
LP-AM263P195worst
Build details
bare
LP-EM-CC2755P1021
Build details
bare
FRDM-MCXA15317best
Build details
bare
LP-MSPM0C110423
Build details
bare
LP-MSPM0G351931
Build details
bare
LP-MSPM0G518736
Build details
bare
LP-MSPM33C321A32
Build details
bare
NUCLEO-U083RC22
Build details
bareat 48 MHz

Interrupt latency under Zephyr, cycles

lower is better

LP-MSPM0G3507163worst
Build details
zephyr
LP-MSPM0L1306119
Build details
zephyr
LP-EM-CC2340R548best
Build details
zephyr
LP-MSPM0G3519157
Build details
zephyr
NUCLEO-U083RC108
Build details
zephyrat 48 MHz

GPIO toggle by register write, cycles

lower is better

LP-MSPM0G35071best
Build details
register
LP-MSPM0L13061best
Build details
register
LP-EM-CC2340R52.1
Build details
register-bare-o3
LP-CC1354P10-11.51
Build details
register-bare-o3
LAUNCHXL-F28001371.56
Build details
register
LP-AM13E230
LP-AM263P10
Build details
register
LP-EM-CC2755P103.99
Build details
register-bare-o3
ESP32-C6-DevKitC-1-N838worst
Build details
register-bare-o3
LAUNCHXL-F28P65X1.56
Build details
register
FRDM-MCXA1531.12
Build details
register
LP-MSPM0C11042
Build details
register-bare-o3
LP-MSPM0G35191best
Build details
register
LP-MSPM0G51871best
Build details
register
LP-MSPM33C321A8
Build details
register-bare-o3
NUCLEO-U083RC2.5
Build details
register-bare-o3at 48 MHz

Control loop, trigger to new output, µs

lower is better; the same run per board as the Control loops chart: code in flash, from the ADC end-of-conversion interrupt where the board has that run (build named under the number)

LP-MSPM0G350788.5
Build details
hw
LP-MSPM0L1306448.1worst
Build details
sw-eoc-flash
LP-EM-CC2340R5336.6
Build details
sw-eoc-flash-os
LP-CC1354P10-121.4
Build details
sw-eoc-flash
LAUNCHXL-F28001372.58
Build details
hw-eoc-flash
LP-AM13E2301.33best
Build details
hw-tmu-eoc-flash
LP-EM-CC2755P1010.5
Build details
sw-eoc-flash-o3
LP-MSPM0G351989.3
Build details
hw
LP-MSPM0G5187180.8
Build details
sw-o3
LP-MSPM33C321A5.03
Build details
sw-eoc-flash-o3
NUCLEO-U083RC298.8
Build details
sw-eoc-flashat 48 MHz

One control step, software float, cycles

lower is better; the same C on every board

LP-MSPM0G350714k
Build details
sw
LP-MSPM0L130613k
Build details
sw-flash
LP-EM-CC2340R514kworst
Build details
sw-flash
LP-CC1354P10-1671.4
Build details
sw-flash
LAUNCHXL-F2800137400.2best
Build details
sw-flash
LP-AM13E230655.6
Build details
sw-flash
LP-EM-CC2755P10593.7
Build details
sw-flash
FRDM-MCXA1535.4k
Build details
sw-flash
LP-MSPM0G351914k
Build details
sw
LP-MSPM0G518714k
Build details
sw
LP-MSPM33C321A531.1
Build details
sw-flash
NUCLEO-U083RC12k
Build details
sw-flashat 48 MHz

Radios

Bluetooth on the air after flashing, s

lower is better

LP-EM-CC2340R51.4worst
Build details
default-100ms-interval
LP-EM-CC2755P100.9best
Build details
default-100ms-interval

Bluetooth echo, 20 bytes, ms

lower is better

LP-EM-CC2340R545worst
LP-CC1354P10-1
LAUNCHXL-F2800137n/a
LP-EM-CC2755P1029.2best

Zigbee light command from Home Assistant, ms

lower is better; median of 20

LP-EM-CC2340R595
Build details
zigbee-home-assistant

Zigbee join, s

lower is better

LP-EM-CC2340R56.2
Build details
zigbee-router-to-cc1354p10-coordinator
LP-CC1354P10-16
Build details
zigbee-zed-zha

Thread join, s

lower is better

LP-EM-CC2340R515.5worst
Build details
thread-router-to-cc1354p10-leader
LP-CC1354P10-18best
Build details
thread-leader-fresh-image

Thread node on the chip

end device or router running on the chip itself

LP-EM-CC2340R5Not for this chip
LP-CC1354P10-1Works
LAUNCHXL-F2800137n/a
ESP32-C6-DevKitC-1-N8Works

Thread radio (RCP) for a host

LP-EM-CC2340R5Works
LP-CC1354P10-1Works
LAUNCHXL-F2800137n/a

Bluetooth and Zigbee or TI 15.4 at once

LP-EM-CC2340R5Not for this chip
LP-CC1354P10-1Works
LAUNCHXL-F2800137n/a
LP-EM-CC2755P10Works

Matter

LP-EM-CC2340R5Not for this chip
LP-CC1354P10-1
LAUNCHXL-F2800137n/a
LP-EM-CC35X1Works

Memory

RAM left beside TI's Bluetooth example, bytes

higher is better

LP-EM-CC2340R55.1 kBworst
Build details
ble-basic-ble
LP-EM-CC2755P1081 kBbest
Build details
ble-basic-ble

RAM left beside TI's Zigbee example, bytes

higher is better; router on the CC2340R5, end device on the CC1354P10

LP-EM-CC2340R5799worst
Build details
zigbee-router
LP-CC1354P10-1210.6 kB
Build details
zigbee-zed-ti-example
LP-EM-CC2755P10126.5 kB
Build details
zigbee-router
ESP32-C6-DevKitC-1-N8360.1 kBbest
Build details
zigbee-router

Blink image RAM, no OS, bytes

lower is better

LP-MSPM0G3507512
Build details
bare
LP-MSPM0L1306256best
Build details
bare
LP-EM-CC2340R59.9 kB
Build details
bare
LP-CC1354P10-118.6 kB
Build details
bare
TI's default configuration, not ranked
LAUNCHXL-F28001371.5 kB
Build details
bare
LP-AM13E2309 kB
Build details
bare
LP-AM263P93.8 kBworst
Build details
bare
LP-EM-CC2755P1010.1 kB
Build details
bare
LP-MSPM0G5187256best
Build details
bare
LP-MSPM33C321A256best
Build details
bareat 32 MHz

Blink image RAM under Zephyr, bytes

lower is better

LP-MSPM0G35073.9 kB
Build details
zephyr
LP-MSPM0L13063.8 kBbest
Build details
zephyr
LP-EM-CC2340R54 kB
Build details
zephyr
FRDM-MCXA1533.8 kB
Build details
zephyr
LP-MSPM0G35194 kB
Build details
zephyr
NUCLEO-U083RC4.8 kBworst
Build details
zephyrat 48 MHz

Blink image RAM under FreeRTOS, bytes

lower is better

LP-MSPM0G35071 kBbest
Build details
freertosat 32 MHz
LP-MSPM0L13061 kBbest
Build details
freertos
LP-EM-CC2340R513 kB
Build details
freertos
LP-CC1354P10-137.6 kB
Build details
freertos
TI's default configuration, not ranked
LP-EM-CC35X136.9 kBworst
Build details
freertos
LP-MSPM0C11041 kBbest
Build details
freertos

RTOS task switch, µs

lower is better; FreeRTOS or Zephyr, whichever ran

LP-MSPM0G35076.72
Build details
zephyr
LP-MSPM0L130612
Build details
freertos
LP-EM-CC2340R510.6
Build details
freertos
LP-CC1354P10-17.38
Build details
freertos
LAUNCHXL-F2800137Not for this chip
LP-AM13E2301.49best
Build details
freertos
LP-EM-CC2755P103.5
Build details
freertos
LP-EM-CC35X1
ESP32-C6-DevKitC-1-N816.5
Build details
freertos
FRDM-MCXA1532.76
Build details
freertos
LP-MSPM0C110419.9worst
Build details
freertos
LP-MSPM0G35196.99
Build details
zephyr
LP-MSPM0G51877
Build details
freertos
LP-MSPM33C321A11.5
Build details
freertosat 32 MHz
NUCLEO-U083RC9.29
Build details
freertosat 48 MHz

Small AI models

Room occupancy, medium, on the CPU, ms

lower is better; 8.5k weights, fastest plain-C build

LP-MSPM0G350735.3
LP-MSPM0L130683.7worst
LP-EM-CC2340R563.2
LP-CC1354P10-125.8
LAUNCHXL-F280013717.5
LP-AM13E2305.05
LP-EM-CC2755P1010.5
LP-EM-CC35X16.25
ESP32-C6-DevKitC-1-N87.14
LAUNCHXL-F28P65X8.38
FRDM-MCXA15312.6
LP-MSPM0C1104Not for this chip
LP-MSPM0G351935.3
LP-MSPM0G518735.3
LP-MSPM33C321A7.38
Raspberry Pi Pico 2 H6.82
SK-TDA4VM0.193best

Room occupancy, medium, on an NPU, ms

lower is better; 8.5k weights, TI NNC code on the chip's neural processing unit

LP-AM13E2300.63best
LP-MSPM0G51872.12worst

Room occupancy, large, fit

33k weights: does it fit and pass

LP-MSPM0G3507Works
LP-MSPM0L1306Not for this chip
LP-EM-CC2340R5Works
LP-CC1354P10-1Works
LAUNCHXL-F2800137Works
LP-AM13E230Works
LP-EM-CC2755P10Works
LP-EM-CC35X1Works
ESP32-C6-DevKitC-1-N8Works
LAUNCHXL-F28P65XWorks
FRDM-MCXA153Works
LP-MSPM0C1104Not for this chip
LP-MSPM0G3519Works
LP-MSPM0G5187Works
LP-MSPM33C321AWorks
NUCLEO-U083RCWorks

Handwritten digit, BitNetMCU 64-wide model, cycles

lower is better

LP-MSPM0G35072.64M
Build details
cnn64-ticlang-o3-32mhzat 32 MHz
LP-MSPM0L13062.65M
Build details
cnn64-ticlang-o3-32mhz
LAUNCHXL-F28001372.55M
Build details
cnn-64-c2000-o4s5-120mhz-ram
LP-MSPM0G35192.64M
Build details
cnn64-ticlang-o3-32mhzat 32 MHz
LP-MSPM0G51872.64M
Build details
cnn64-ticlang-o3-32mhzat 32 MHz

Hardware helpers

sin and cos in hardware against software, times faster

higher is better

LP-MSPM0G350785.6
LP-MSPM0L1306n/a
LP-EM-CC2340R5n/a
LP-CC1354P10-1n/a
LAUNCHXL-F28001379.87worst
LP-MSPM0C1104n/a
LP-MSPM0G351993.3best
LP-MSPM0G5187n/a
NUCLEO-U083RCn/a

AES in hardware against software, times faster

higher is better

LP-MSPM0G350720.3
Build details
1kb-ticlang-o2
LP-EM-CC2340R5131.5best
Build details
1kb-ecb-ticlang-o3-driver-polling
LP-CC1354P10-151.1
Build details
1kb-ecb-ticlang-o2-driver-polling
LP-AM13E23029.9
Build details
1kb-ecb-ticlang-o2-keyout
LP-EM-CC2755P1092.8
Build details
ticlang-o2-driver-polling
ESP32-C6-DevKitC-1-N81.23worst
Build details
16b-ecb-mbedtls
LP-MSPM0G351945.8
Build details
1kb-ticlang-o2
LP-MSPM0G518745.5
Build details
1kb-ticlang-o2
LP-MSPM33C321A41.9
Build details
1kb-ecb-ticlang-o2-keyoutat 32 MHz
NUCLEO-U083RC93.9
Build details
1kb-aes256-ecb-gcc12-o2-regat 48 MHz

CRC in hardware against a table, times faster

higher is better

LP-MSPM0G350713.5best
LP-MSPM0L130610.9
LP-AM13E2301.89worst
FRDM-MCXA1537.06
LP-MSPM0C11043
Build details
crc16-256b
LP-MSPM0G35198.09
LP-MSPM0G51878.12
LP-MSPM33C321A2.75
Build details
at 32 MHz
NUCLEO-U083RC8.34
Build details
gcc12-o3at 48 MHz
Raspberry Pi Pico 2 H5.71
Build details
dma-sniffer-riscv

256-point FFT, ms

lower is better; float where the chip has an FPU

LP-MSPM0G3507
LP-MSPM0L130625.7worst
Build details
f32
LP-EM-CC2340R518.3
Build details
f32
LP-CC1354P10-10.473
Build details
f32
LAUNCHXL-F28001370.107best
Build details
f32
LP-AM13E2300.117
Build details
f32
LP-EM-CC2755P100.224
Build details
f32
FRDM-MCXA1534.69
Build details
f32
LP-MSPM0C1104Not for this chip
LP-MSPM0G351911.1
Build details
f32
LP-MSPM0G518711
Build details
f32
LP-MSPM33C321A0.204
Build details
f32
NUCLEO-U083RC20
Build details
f32at 48 MHz

Fastest ADC reading, million samples per second

higher is better; each board's best setting. Under each number: the ADC's clock when it differs from the CPU's, DMA or not, and resolution and sample window where recorded

LP-MSPM0G35073.93
Build details
at 32 MHzCPU at 32 MHz, ADC clock as in the example
LP-MSPM0L13061.45
Build details
8-cycle window
LP-EM-CC2340R51.2
Build details
ti-adcbuf-dmaDMA
LP-CC1354P10-10.2
Build details
ti-adcbuf-dmaDMA
LAUNCHXL-F28001373.98
Build details
ti-continuous-9cyc9-cycle window
LP-EM-CC2755P101.2
Build details
ti-adcbuf-dmaDMA
LP-EM-CC35X10.17worst
Build details
12-bit
FRDM-MCXA1532.53
Build details
dmaDMA12-bit
LP-MSPM0C11041
Build details
DMA10-bit, 2-cycle window
LP-MSPM0G35193.93
Build details
at 32 MHzCPU at 32 MHz, ADC clock as in the example
LP-MSPM0G51870.547
Build details
at 32 MHzCPU at 32 MHz, ADC clock as in the example12-bit, 62.5 ns window
LP-MSPM33C321A5best
Build details
DMA
NUCLEO-U083RC1.71
Build details
internalCPU at 48 MHz, ADC clock as in the exampleDMA12-bit
Raspberry Pi Pico 2 H0.5
Build details
pico-sdk-dma-riscvDMA

Power

Deepest sleep with RAM kept, µA

lower is better; probe-measured chip figure; modes differ by chip (named under the number)

LP-MSPM0G3507
LP-MSPM0L13066.2
Build details
STANDBY0, RAM kept
LP-EM-CC2340R5
LP-CC1354P10-1
LAUNCHXL-F28001372,590
Build details
ti-c2000-standby-borrowed-energytrace
STANDBY with the PLL on; HALT is the reference floor; not ranked
LP-AM13E230
LP-AM263P
TMDS62LEVM (AM62L EVM)
LP-EM-CC35X1
FRDM-MCXA153
LP-MSPM0C1104
LP-MSPM0G3519
LP-MSPM0G5187
LP-MSPM33C321A
NUCLEO-U083RC

Running current per MHz, µA

lower is better; slope between two clocks; the workload differs by chip (named under the number)

LP-MSPM0L130672.9
Build details
energytrace-3.3v-coremark
LP-EM-CC2340R5
LP-CC1354P10-1
LAUNCHXL-F2800137263
Build details
ti-c2000-borrowed-energytrace
busy loop at two clocks, not CoreMark; not ranked
LP-AM13E230
LP-AM263P
LP-EM-CC35X1
FRDM-MCXA153
LP-MSPM0G3519
LP-MSPM0G5187
LP-MSPM33C321A
NUCLEO-U083RC

Getting started

Plug-in to a changed blink, s

lower is better; the CC2340R5 figure includes 8 min lost to a half-seated probe

LP-MSPM0G3507149
Build details
ccs-mcp
LP-MSPM0L1306
LP-EM-CC2340R5540worst
Build details
cli
LP-CC1354P10-129
Build details
cli
LAUNCHXL-F2800137
LP-AM13E23060
Build details
ccs-mcp
LP-AM263P78
Build details
cli
LP-EM-CC2755P10198
Build details
ccs-mcp
LP-EM-CC35X160
Build details
cli
ESP32-C6-DevKitC-1-N8184
Build details
cli
FRDM-MCXA1531.4best
Build details
cliat 12 MHz
IWRL6432BOOSTNot for this chip
LP-MSPM0C110477
Build details
cli
LP-MSPM0G3519
LP-MSPM0G5187
LP-MSPM33C321A19
Build details
cli
NUCLEO-U083RC14
Build details
cliat 48 MHz
Raspberry Pi Pico 2 H87
Build details
cli
SK-TDA4VM

Blink build, s

lower is better

LP-MSPM0G35076
Build details
ccs-mcp
LP-MSPM0L13063.3
Build details
cli
LP-EM-CC2340R522
Build details
cli
LAUNCHXL-F28001374.6
Build details
ti-led-ex1-blinky-cli
LP-AM13E2302.7best
Build details
cli
LP-AM263P5.9
Build details
cli
LP-EM-CC2755P1014
Build details
cli
LP-EM-CC35X16
Build details
cli
ESP32-C6-DevKitC-1-N867worst
Build details
cli
LAUNCHXL-F28P65X8.6
Build details
ti-led-ex1-blinky-cli
FRDM-MCXA15330
Build details
cliat 12 MHz
LP-MSPM0C11043.5
Build details
cli
LP-MSPM0G35193.2
Build details
cliat 32 MHz
LP-MSPM0G51874
Build details
cliat 32 MHz
LP-MSPM33C321A4.8
Build details
cliat 32 MHz
NUCLEO-U083RC6
Build details
cliat 48 MHz
Raspberry Pi Pico 2 H12
Build details
cli

Blink flash, s

lower is better

LP-MSPM0G35076
Build details
west
LP-MSPM0L13060.6best
Build details
openocd
LP-EM-CC2340R57
Build details
uniflash
LAUNCHXL-F28001373.9
Build details
ccs-scripting-ram-load
LP-AM13E2303.3
Build details
uniflash
LP-EM-CC2755P1027worst
Build details
ccs-mcp
LP-MSPM0C11040.64
Build details
openocd
LP-MSPM0G35190.6best
Build details
openocdat 32 MHz
LP-MSPM0G518720.7
Build details
openocdat 32 MHz
LP-MSPM33C321A11.3
Build details
uniflashat 32 MHz
NUCLEO-U083RC1.2
Build details
openocdat 48 MHz

Examples safe to run that ran with no crash, %

higher is better; of the images safe to run, not the whole SDK: 213 of 270 on the G3507, 148 of 170 on the L1306

LP-MSPM0G3507100
Build details
mspm0-sdk-2.11at 32 MHz
LP-MSPM0L1306100
Build details
mspm0-sdk-2.11
LP-EM-CC2340R5100
Build details
cc35x1-twins-checked
LP-CC1354P10-1
LAUNCHXL-F2800137100
Build details
c2000ware-26.02
LP-AM13E23098.8
Build details
am13e230-sdk-26.01
LP-AM263P118best
Build details
mcu-plus-sdk-26.01-lp
LP-EM-CC2755P1098.8
Build details
sdk-f3-9.21
LP-EM-CC35X193.5
Build details
wifi-sdk-10.20.00.39
ESP32-C6-DevKitC-1-N887.1
Build details
esp-idf-v6.1
FRDM-MCXA15397.7
Build details
mcux-26.09.00
LP-MSPM0C1104100
Build details
mspm0-sdk-2.11
LP-MSPM0G3519100
Build details
mspm0-sdk-2.11
LP-MSPM0G518790.8
Build details
mspm0-sdk-2.11
LP-MSPM33C321A99.4
Build details
mspm33-sdk-1.04at 32 MHz
NUCLEO-U083RC100
Build details
stm32cubeu0-1.3.0at 48 MHz
Raspberry Pi Pico 2 H
SK-TDA4VM60worst
Build details
edgeai-gst-apps-11.00.00.08

Full redraw of a 3.5 inch SPI screen, ms

lower is better; fastest setting

LP-MSPM0G350767.2
Build details
spi40-dma16
LP-MSPM0G351967.2
Build details
spi40-dma16

best in the rowbetweenworst in the rowNo color: a spread under 5 percent, one number only, or a figure that is not like for like (named under it). Hover Not for this chip for the reason. Small gray lines: the build behind the number. Hover a cell for more.

Linux boards: their own numbers

A Linux board runs its own tests: boot time, memory bandwidth and latency, model inference, video, Ethernet and storage. Most have no MCU to compare with, so the board sits behind the scorecard toggle. These are its headline numbers. The board page has the conditions and limits of each.

SK-TDA4VM 2x Cortex-A72 + C7x/MMA + 2x C66x + 6x R5F, 2000 MHz, 4 GB RAM

Measured	Best measured	Variants
Power in to the Linux login prompt on the serial console	24.2 to 24.5 s warm	best of 2 variants; every row
Vision app launch to the first frame-rate line	3.09 to 3.84 s edgeai-app-yolox-pose-file	1 variant; every row
CoreMark	19,579.9 to 19,583.4 iter/s gcc-linux-o3-2t	best of 7 variants; every row
Dhrystone 2.1	5.26 to 5.68 DMIPS/MHz gcc-linux-o3-inline	best of 5 variants; every row
Memory bandwidth, STREAM-style triad	5,385 MB/s triad-2t	best of 3 variants; every row
Memory load latency by pointer chase	2.5 ns l1-16k	best of 7 variants; every row
DDR bandwidth used during a pipeline (TI perf_stats)	1,957 MB/s yolox-zerocopy-111fps	first of 2 variants; every row
Deep-learning model: one inference, median wall time	2.17 ms mobilenetv1-mlperf-224-tflitert-tidl-c7x	best of 22 variants; every row
Whole vision graph: decode, preprocess, inference, postprocess, sink	216.4 fps three-pipelines-yolox-c7x	best of 8 variants; every row
Hardware video decode throughput	499.6 fps 1280x768-h264-zerocopy	best of 7 variants; every row
Hardware video encode throughput	131 fps 720p-h264	best of 4 variants; every row
Wired Ethernet TCP throughput (iperf3)	939.7 to 939.9 Mbit/s tcp-rx-from-host	best of 3 variants; every row
Wired Ethernet round trip to a host on the same switch	0.313 to 0.37 ms wired-host-same-switch	1 variant; every row
Timer wake-up latency (cyclictest), max over 60 s	41 us idle	best of 2 variants; every row
Boot medium sequential read	44.8 MB/s microsd-fio-1m	best of 2 variants; every row
Boot medium sequential write	30.5 MB/s microsd-dd-direct	1 variant; every row
openssl speed, EVP, 8 KB blocks	1,560 MB/s aes-128-gcm-cpu	best of 10 variants
Room-occupancy model, one inference	0.0174 ms Room occupancy: tiny, GCC -O3, on the A72 cores	best of 4 variants; every row

TMDS62LEVM (AM62L EVM) 2x Cortex-A53, 1250 MHz, 2 GB RAM

Measured	Best measured	Variants	Source
Power in to the Linux login prompt on the serial console	3.9 to 3.91 s systemd-kernel	best of 9 variants; every row	README notes
CoreMark	8,428.2 iter/s gcc15.3-linux-o2-2t	best of 7 variants; every row	README notes
Memory bandwidth, STREAM-style triad	2,418.5 MB/s copy-2t	best of 4 variants; every row	README notes
Memory load latency by pointer chase	2.41 ns l1-16k	best of 3 variants; every row	README notes
Whole vision graph: decode, preprocess, inference, postprocess, sink	30 fps kmscube-1080p60-llvmpipe	best of 4 variants; every row	README notes
Wired Ethernet TCP throughput (iperf3)	927 Mbit/s tx-to-host	best of 2 variants; every row	README notes
Wired Ethernet round trip to a host on the same switch	0.368 ms host-same-switch	1 variant; every row	README notes
Timer wake-up latency (cyclictest), max over 60 s	442 to 645 us idle-60s-x3	best of 4 variants; every row	AM62L realtime jitter notes
Boot medium sequential read	38.1 MB/s usb-stick-dd-direct	1 variant; every row	README notes
Boot medium sequential write	10.4 MB/s usb-stick-dd-direct	best of 3 variants; every row	README notes

What runs on each board

What you want to do

LP-MSPM0G3507Cortex-M0+, 32 kB RAM

LP-MSPM0L1306Cortex-M0+, 4 kB RAM

LP-EM-CC2340R5Cortex-M0+, 36 kB RAM

LP-CC1354P10-1Cortex-M33, 296 kB RAM

LAUNCHXL-F2800137C28x, 36 kB RAM

LP-AM13E230Cortex-M33 with TMU and NPU, 128 kB RAM

LP-AM263P4x Cortex-R5F with TMU, 3 MB RAM

TMDS62LEVM (AM62L EVM)2x Cortex-A53, 2 GB RAM

LP-EM-CC2755P10Cortex-M33, 162 kB RAM

LP-EM-CC35X1Cortex-M33, 1.1 MB RAM

ESP32-C6-DevKitC-1-N8RISC-V, 512 kB RAM

LAUNCHXL-F28P65XC28x, 248 kB RAM

FRDM-MCXA153Cortex-M33, 32 kB RAM

IWRL6432BOOSTCortex-M4F, 1 MB RAM

LP-MSPM0C1104Cortex-M0+, 1 kB RAM

LP-MSPM0G3519Cortex-M0+, 128 kB RAM

LP-MSPM0G5187Cortex-M0+ with NPU, 32 kB RAM

LP-MSPM33C321ACortex-M33, 256 kB RAM

NUCLEO-G071RBCortex-M0+, 36 kB RAM

NUCLEO-U083RCCortex-M0+, 40 kB RAM

Raspberry Pi Pico 2 HCortex-M33, 520 kB RAM

SK-TDA4VM2x Cortex-A72 + C7x/MMA + 2x C66x + 6x R5F, 4 GB RAM

Blink an LED

LP-MSPM0G3507Works

LP-MSPM0L1306Works

LP-EM-CC2340R5Works

LP-CC1354P10-1

LAUNCHXL-F2800137Works

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Build and flash from the terminal in one command

LP-MSPM0G3507Works

LP-MSPM0L1306Works

LP-EM-CC2340R5Works

LP-CC1354P10-1

LAUNCHXL-F2800137Works

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Run the Zephyr real-time operating system

LP-MSPM0G3507Works

LP-MSPM0L1306Could work

LP-EM-CC2340R5Not for this chip

LP-CC1354P10-1

LAUNCHXL-F2800137

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Run the FreeRTOS real-time operating system

LP-MSPM0G3507Should work

LP-MSPM0L1306Works

LP-EM-CC2340R5Works

LP-CC1354P10-1

LAUNCHXL-F2800137Not for this chip

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Run a small AI model on the chip

LP-MSPM0G3507Works

LP-MSPM0L1306Works

LP-EM-CC2340R5Works

LP-CC1354P10-1

LAUNCHXL-F2800137Works

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Send a sensor reading to a phone over Bluetooth

LP-MSPM0G3507Not for this chip

LP-MSPM0L1306Not for this chip

LP-EM-CC2340R5Could work

LP-CC1354P10-1Could work

LAUNCHXL-F2800137Not for this chip

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Switch a board's LED from Home Assistant over Zigbee

LP-MSPM0G3507Not for this chip

LP-MSPM0L1306Not for this chip

LP-EM-CC2340R5Could work

LP-CC1354P10-1Works

LAUNCHXL-F2800137Not for this chip

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Join a Thread network

LP-MSPM0G3507Not for this chip

LP-MSPM0L1306Not for this chip

LP-EM-CC2340R5Not for this chip

LP-CC1354P10-1Works

LAUNCHXL-F2800137Not for this chip

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Run Bluetooth and Zigbee on one chip at once

LP-MSPM0G3507Not for this chip

LP-MSPM0L1306Not for this chip

LP-EM-CC2340R5Not for this chip

LP-CC1354P10-1

LAUNCHXL-F2800137Not for this chip

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Update a board's firmware over the air

LP-MSPM0G3507Not for this chip

LP-MSPM0L1306Not for this chip

LP-EM-CC2340R5Could work

LP-CC1354P10-1

LAUNCHXL-F2800137Not for this chip

LP-AM13E230

LP-AM263P

TMDS62LEVM (AM62L EVM)

LP-EM-CC2755P10

LP-EM-CC35X1

ESP32-C6-DevKitC-1-N8

LAUNCHXL-F28P65X

FRDM-MCXA153

IWRL6432BOOST

LP-MSPM0C1104

LP-MSPM0G3519

LP-MSPM0G5187

LP-MSPM33C321A

NUCLEO-G071RB

NUCLEO-U083RC

Raspberry Pi Pico 2 H

SK-TDA4VM

Works ran here and did what it shouldShould work it builds; only naming, packaging or build glue is in the way; hover for whichCould work some of it ran; real porting effort is left; hover for whatNot for this chip the chip or its vendor software cannot; hover for whyNot for this chip not enough RAM for itNot for this chip the chip has no radio for itNot for this chip the vendor offers no software for it on this chipNot for this chip a tool or build problem stopped it, not the chipNot for this chip it stopped on the air; the radio link in the room is the suspect, not the chip (not proven)

Every test, every board

The whole store, by test class; the tests that explain a chart sit under that chart. An empty cell is a gap, not a zero. Variant codes:

Code	Means
`ticlang`	TI Arm Clang, TI's own compiler
`gcc`	GCC, the open-source compiler Zephyr uses
`armgnu14`	GCC 14 from the Arm GNU Toolchain
`ti-c2000`	TI's C2000 compiler
`bare`	no operating system
`zephyr`	under the Zephyr operating system
`o2, o3, os, oz`	compiler optimization level: o3 is fastest code, os and oz are smallest
`o2unroll`	Zephyr's speed setting: -O2 with loop unrolling. Its menu calls it -O3.
`flash, ram`	where the code ran from
`c28patch`	CoreMark patched for the C28x 16-bit byte: unofficial
`hw, sw`	math unit in hardware, or the same math in software
`nnc`	built by TI's Neural Network Compiler
`c-o3`	the model as plain C at that optimization level
`energytrace`	measured with the board's own EnergyTrace probe
`eoc, early`	control loop timed from the ADC end-of-conversion interrupt, or from the earlier ADC-start interrupt
`ramfunc`	the interrupt handler copied to RAM
`o4, o4s0, o4s5`	C2000 compiler -O4 with size-against-speed settings 0 or 5
`q15, q31, f32`	FFT in 16-bit fixed point, 32-bit fixed point, or 32-bit float
`-32mhz, -4mhz`	the same build run at that clock instead of the board's top clock
`cli, ccs-mcp, west, uniflash, openocd, dslite`	the path or tool used: command line, CCStudio's AI tools, Zephyr's tool, TI's flashers, OpenOCD
`register, driver`	one register write, or the vendor's driver call
`(one way)`	the test has only one way to run it, so the row has no variant code

Setup: the other tests in this class

Box or plug-in to a changed blink s, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	CC3200-LAUNCHXL	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	IWRL6432BOOST	LP-AM263P
arduino									124 (1 run)48 MHz		146 (1 run)150 MHz						Not for this chip
ccs-mcp	149 (1 run)													60 (1 run)	198 (1 run)
cli			77 (1 run)24 MHz			540 (1 run)	29 (1 run)		14 (1 run)48 MHz	1.4 (1 run)12 MHz	87 (1 run)150 MHz	184 (1 run)160 MHz	19 (1 run)	22 (1 run)	41 (1 run)	60 (1 run)	Not for this chip	78 (1 run)
cli-eyes-check
cli-script					13.2 to 14.632 MHz
cli-scripted				15.2 to 15.732 MHz
digital-lab-chrome-webhid		13.3 (1 run)32 MHz			22.9 (1 run)80 MHz
linkserver
micropython											65 (1 run)150 MHz						Not for this chip
oneline-terminal		481 (1 run)
oneline-terminal-start-sh		75 (1 run)
platformio																	Not for this chip
rust
rust-from-nothing		160 (1 run)32 MHz
ti-led-ex1-blinky-cli
vendor-ide
zephyr	570 (1 run)		Not for this chip			82 (1 run)			471 (1 run)48 MHz	128 (1 run)96 MHz
zephyr-script	180 to 186			Not for this chip
(one way)								Not for this chip

Sources

c2000-launchpad/tools/c2kbuild_f28p65x.sh, cc1354p10/blink/README.md, cc2340r5/zephyr/README.md, cc3200-launchxl/README.md, iwrl6432boost/README.md, mspm0g3507/zephyr/README.md, mspm0l1306/README.md, ports/rust-mspm0/README.md

Blink example build time s, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X	LP-AM263P	TMDS62LEVM (AM62L EVM)
arduino								2148 MHz, 1 run
arduino-cli										301 run
ccs-mcp	61 run												91 run	101 run
cli		3.332 MHz, 1 run	3.5 to 4.324 MHz	3.2 to 4.132 MHz	4 to 5.832 MHz	221 run		648 MHz, 1 run	3012 MHz, 1 run	121 run	671 run	4.832 MHz, 1 run	2.71 run	141 run	61 run		5.91 run
cli-wrapper				5.5 to 5.732 MHz
driverlib		18.432 MHz, 1 run	2.424 MHz, 1 run	3.932 MHz, 1 run	4.532 MHz, 1 run
driverlib-am13													14.3200 MHz, 1 run
driverlib-hsadc												9.932 MHz, 1 run
esphome-cc23xx-compile-cold						31.5
esphome-cc23xx-compile-warm						22.4 to 23.3
esphome-cc27xx-compile-cold														35.1 to 40.1
esphome-cc27xx-compile-warm														27 to 31.9
esphome-cc35xx-compile-cold															55.71 run
esphome-cc35xx-compile-warm															23.3 to 25.9
moonlight-embedded-f32e415-cmake-make																		2821250 MHz, 1 run
rust-cold		10732 MHz, 1 run
rust-incremental		3.0532 MHz, 1 run
three-ways-app-c2000ware							2.7 to 4120 MHz, 1 run
three-ways-app-linux-gcc14-o2																		1.23 to 1.271250 MHz
three-ways-app-tidrivers						37.11 run								54.11 run	8.41 run
ti-led-ex1-blinky-cli							4.61 run									8.61 run
zephyr	621 run			51.5 to 52.880 MHz		471 run		42.748 MHz, 1 run	12496 MHz, 1 run
(one way)

Sources

c2000-launchpad/README.md, c2000-launchpad/tools/c2kbuild_f28p65x.sh, cc2340r5/README.md, cc2340r5/zephyr/README.md, cc3200-launchxl/README.md, data/benchmark-suite.json, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, docs/experiments/2026-10-13-esphome-cc35xx-platform.md, mspm0l1306/README.md, portability/three-ways/README.md, portability/three-ways/tidrivers/build_cc2755p10.sh, portability/three-ways/tidrivers/build_cc35x1.sh, ports/rust-mspm0/README.md

Blink flash time s, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	CC3200-LAUNCHXL	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P
ccs-mcp	381 run														271 run
ccs-scripting-ram-load								3.91 run
digital-lab-chrome-webhid		0.732 MHz, 1 run			0.880 MHz, 1 run
digital-lab-node-saci															6.696 MHz, 1 run
dsflash-wrapper				5.2 to 5.332 MHz
dslite			5.4 to 5.624 MHz	4.6 to 4.832 MHz	4.7 to 6.432 MHz										91 run		5
dslite-esphome-upload															15 to 16.1
dslite-esphome-upload-cc23xx						15.1 to 16.7
dslite-flash-cli
esptool												2.4
frl			124 MHz, 1 run	0.932 MHz, 1 run	0.832 MHz, 1 run			9.4 to 9.8					0.532 MHz, 1 run	1.31 run
openocd		0.632 MHz, 1 run	0.64 to 0.6724 MHz	0.6 to 20.632 MHz	20.732 MHz, 1 run		Not for this chip		1.248 MHz, 1 run
openocd-am13e230x
openocd-cc23xx						0.371 run
openocd-mspm0				0.95 to 1.0932 MHz
openocd-ti-fork-cc23xx						6.21 run
openocd-ti-fork-mspm33
picotool											2
pyocd										10.812 MHz, 1 run
pyocd-zephyr										3.696 MHz, 1 run
rust-openocd		2.632 MHz, 1 run
stm32cubeprogrammer									548 MHz, 1 run
stm32duino-swd									348 MHz, 1 run
toolbox																201 run
uf2											11 run
uniflash		1532 MHz, 1 run				71 run	Not for this chip						11.332 MHz, 1 run	3.31 run
uniflash-zephyr						71 run
west	61 run			4.5 to 4.880 MHz
(one way)

Sources

c2000-launchpad/README.md, c2000-launchpad/tools/F28P650DK9_CL650001.ccxml, cc2340r5/README.md, cc2340r5/zephyr/README.md, cc3200-launchxl/README.md, data/benchmark-suite.json, docs/experiments/2026-10-13-esphome-cc27xx-platform.md, docs/experiments/2026-10-18-open-flash-path-survey.md, docs/frl.md, mspm0l1306/README.md, ports/openocd-cc23xx/README.md, ports/rust-mspm0/README.md

One step of frl's connect ladder reaches the chip on its own (frl identify --connect-step) yes/no

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM33C321A	LP-AM13E230
clock-100-khz	132 MHz, 1 run
clock-1000-khz	132 MHz, 1 run
connect-under-reset			11 run	1200 MHz, 1 run
probe-reset-xdsdfu		Not for this chip

Sources

docs/frl.md

Print-a-message write and serial check, three runs back to back s, lower is better

Variant	LP-MSPM0L1306	LP-EM-CC2340R5	LP-EM-CC35X1
frl	132 MHz
frl-dslite		9.2 to 9.848 MHz
frl-toolbox			23.3 to 24.2

Sources

docs/frl.md

frl's own check fails on the board when asked for something the program does not do yes/no

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC35X1
led-log-other-pattern					148 MHz, 1 run
led-other-pattern	11 run	11 run	11 run			11 run	11 run	11 run
uart-other-message				148 MHz, 1 run					11 run

Sources

docs/frl.md

frl refuses an unsafe write or restore before anything is written yes/no

Variant	LP-CC1354P10-1	LP-EM-CC35X1
layout-change		11 run
old-backup		11 run
power-cycle-other-port	11 run
stale-bundle		11 run

Sources

docs/frl.md

Print-a-message write and serial check, one run s, lower is better

Variant	LP-MSPM0L1306	LP-EM-CC35X1
frl	1.5132 MHz, 1 run
frl-toolbox-hashed-copy		20.21 run

Sources

docs/frl.md

Blink write and check through frl's second write path (TI's DSLite), once s, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230
dslite	632 MHz, 1 run	6.124 MHz, 1 run	5.932 MHz, 1 run	61 run	11.51 run	5.71 run	5.2200 MHz, 1 run

Sources

docs/frl.md

Blink write and check, three runs back to back s, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230
frl	0.732 MHz	0.624 MHz	0.7 to 20.632 MHz	0.732 MHz			9.2 to 9.9	0.532 MHz	1.1
frl-ccs13						7.4 to 34.648 MHz
frl-dslite					10.2 to 10.548 MHz
frl-verify-path			0.732 MHz

Sources

docs/frl.md

Debug probe on the board yes/no

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	CC3200-LAUNCHXL	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X	LP-AM263P
(one way)	yes1 run	yes1 run	yes1 run	yes1 run	yes1 run	no	yes	yes	yes1 run	yes1 run	yes1 run	no1 run	yes	yes1 run	yes1 run	no	no	yes1 run	yes1 run

Sources

c2000-launchpad/README.md, c2000-launchpad/tools/F28P650DK9_CL650001.ccxml, cc1354p10/README.md, cc2340r5/README.md, cc2755p10/README.md, cc3200-launchxl/README.md, esp32c6/README.md, mspm0l1306/README.md, pico2/README.md

Power in to the Linux login prompt on the serial console s, lower is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
booti-to-login-median3		21.4 to 21.71250 MHz
first-boot	27.71 run
kernel-6.18.38-ti-kvm-reboot-to-login		32.81250 MHz, 1 run
kvm-guest-alpine-qemu-start-to-poweroff		19.9 to 20.21250 MHz
rom-phase1-zephyr-sram
sd-boot-to-login
systemd-kernel		3.9 to 3.911250 MHz
systemd-userspace-total		135.3 to 135.51250 MHz
uart-rom-ping-to-uboot-prompt		420 to 422.5
uart-tiboot3-xmodem		28 to 28.2
uart-tispl-xmodem		190 to 191.9
uart-uboot-img-ymodem		192.2 to 192.4
warm	24.2 to 24.5

Sources

am62l-evm/README.md, am62l-evm/zephyr/mkphase1.sh, docs/experiments/2026-10-24-am62l-sd-boot-stall.md

LED pin toggles under the Rust build (probe GPIO read, changes counted) count, higher is better

Variant	LP-MSPM0L1306	LP-AM13E230
msp430-fram-persist	132 MHz, 1 run
rust-embassy	532 MHz, 1 run

Sources

ports/msp430-to-mspm0/fram-persist/PORT.md, ports/rust-mspm0/README.md

Serial console shows the program banner and a running count yes/no

Variant	LP-MSPM0L1306	LP-AM13E230
cli-wrapper
msp430-lpm3-wake	yes32 MHz, 1 run	yes200 MHz, 1 run
msp430-uart-echo	yes32 MHz, 1 run
rust-embassy	Works32 MHz, 1 run

Sources

ports/msp430-to-mspm0/uart-echo-lpm3/PORT.md, ports/rust-mspm0/README.md

Core: the other tests in this class

GPIO toggle cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-EM-CC2755P10	LAUNCHXL-F28P65X	LP-AM263P
driver	1.11 to 880 MHz, 1 run	1.05 to 1332 MHz, 1 run	2.04 to 524 MHz	1.11 to 880 MHz	1.08 to 880 MHz			5 to 9120 MHz, 1 run	2148 MHz	3.0196 MHz, 1 run					5 to 9.99200 MHz, 1 run	10400 MHz
driver-bare-o3						19.248 MHz, 1 run	15.148 MHz					68160 MHz		8.0696 MHz
driver-bare-oz						3148 MHz, 1 run	1948 MHz							1196 MHz
driver-m33											7150 MHz
driver-m33-gcc12											7150 MHz
driver-o3
driver-riscv											6150 MHz
register	1 to 1.1280 MHz, 1 run	1 to 1.1232 MHz, 1 run		180 MHz	1 to 1.1280 MHz			1.56120 MHz, 1 run		1.1296 MHz, 1 run					1.56200 MHz, 1 run	10400 MHz
register-bare-o3			224 MHz, 1 run			2.148 MHz, 1 run	1.5148 MHz		2.5 to 448 MHz			38160 MHz	8160 MHz	3.9996 MHz
register-bare-o3-32mhz													3.0232 MHz
register-bare-oz						948 MHz, 1 run	8.0148 MHz							7.0196 MHz
register-m33											4150 MHz
register-m33-gcc12											4150 MHz
register-riscv											3150 MHz
register-zephyr-o2unroll						2.848 MHz, 1 run
zephyr	36.2 to 4980 MHz, 1 run	26 to 4232 MHz, 1 run							21.5 to 3048 MHz, 1 run
zephyr-o2unroll						26.848 MHz, 1 run

Sources

cc1354p10/baremetal/README.md, cc2340r5/baremetal/README.md, mspm0l1306/zephyr/README.md

256-point complex FFT ms, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10
armgnu14-f32									4.6696 MHz, 1 run
armgnu14-q15									0.33296 MHz, 1 run
armgnu14-q31									0.25796 MHz, 1 run
cmsis-dsp
f32	25.732 MHz		11.180 MHz	1180 MHz	18.348 MHz	0.47348 MHz	0.107120 MHz	2048 MHz	4.6996 MHz, 1 run		0.204160 MHz	0.117200 MHz	0.22496 MHz
f32-32mhz											0.71632 MHz
f32-flash							0.106120 MHz
f32-m33										0.17150 MHz
f32-riscv										2.46150 MHz
q15	2.2532 MHz, 1 run	Not for this chip	0.91880 MHz	0.91880 MHz	1.6948 MHz	0.3448 MHz		1.4448 MHz	0.32996 MHz, 1 run		0.135160 MHz	0.0755200 MHz	0.1696 MHz
q15-32mhz											0.49832 MHz
q15-m33										0.0969150 MHz
q15-riscv										0.179150 MHz
q31	7.6832 MHz, 1 run		3.2180 MHz	3.2780 MHz	5.3848 MHz	0.55848 MHz		3.7448 MHz	0.25496 MHz, 1 run		0.24160 MHz	0.136200 MHz	0.26896 MHz
q31-32mhz											0.82332 MHz
q31-m33										0.161150 MHz
q31-riscv										0.135150 MHz

Sources

data/boards.csv

sin plus cos cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
fpu-flash											154.9150 MHz
fpu-flash-gcc12											156.9150 MHz
fpu-flash-o3											154.9150 MHz
fpu-flash-o3-gcc12											156.9150 MHz
fpu-flash-os											157.1150 MHz
fpu-flash-os-gcc12											157.1150 MHz
fpu-gcc-bare-o3-m33											152.7150 MHz
fpu-gcc-bare-o3-m33-gcc12											152.7150 MHz
hw	115.8 to 132.180 MHz, 1 run	n/a	n/a	109.8 to 143.180 MHz, 1 run	n/a	n/a	n/a	16.5120 MHz	n/a
hw-flash								24120 MHz
hw-tmu														11.6200 MHz		26.9400 MHz
riscv-nofpu-gcc-bare-o3-riscv											2.2k150 MHz
sdk																77.8400 MHz
sw	9.9k to 10k80 MHz, 1 run	8.7k to 9.1k32 MHz, 1 run		10k to 11k80 MHz, 1 run	8k to 8k80 MHz	9.7k48 MHz, 1 run		162.8120 MHz		2.5k96 MHz, 1 run		3.4k160 MHz	495.5160 MHz, 1 run	512.7200 MHz		300.2400 MHz
sw-32mhz													454.432 MHz, 1 run
sw-flash		7.1k32 MHz					522.548 MHz	190.7120 MHz	6.3k48 MHz						451.996 MHz
sw-flash-o3															451.996 MHz
sw-flash-o3-riscv											2.3k150 MHz
sw-flash-os															461.696 MHz
sw-flash-os-riscv											2.3k150 MHz
sw-flash-riscv											2.3k150 MHz
sw-ticlang-bare-o3						8.2k48 MHz	527.948 MHz								45796 MHz
sw-ticlang-bare-oz						8.2k48 MHz	53948 MHz								465.296 MHz

Sources

mspm0l1306/zephyr/README.md

Divide cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
fpu-flash							18.848 MHz				21.1150 MHz				18.896 MHz
fpu-flash-gcc12											21.1150 MHz
fpu-flash-o3											21.1150 MHz				18.896 MHz
fpu-flash-o3-gcc12											21.1150 MHz
fpu-flash-os											20.1150 MHz				2396 MHz
fpu-flash-os-gcc12											21.1150 MHz
fpu-gcc-bare-o3-m33											17.6150 MHz
fpu-gcc-bare-o3-m33-gcc12											17.6150 MHz
hw	100 to 11280 MHz, 1 run	n/a	n/a	94 to 119.180 MHz, 1 run	n/a	n/a	n/a	8120 MHz	n/a	13.696 MHz, 1 run		9160 MHz
hw-flash								12120 MHz
hw-tmu														7.86200 MHz		Not for this chip
riscv-nofpu-gcc-bare-o3-riscv											26150 MHz
sw	262.4 to 267.580 MHz, 1 run	242.2 to 253.232 MHz, 1 run		263.3 to 271.580 MHz, 1 run	368 to 373.180 MHz	283.548 MHz, 1 run		241.2120 MHz					15.1160 MHz, 1 run	15.1200 MHz		15.9400 MHz
sw-32mhz													15.132 MHz, 1 run
sw-flash		509.232 MHz						250.2120 MHz	442.948 MHz
sw-flash-o3-riscv											173.2150 MHz
sw-flash-os-riscv											172.3150 MHz
sw-flash-riscv											173.2150 MHz
sw-ticlang-bare-o3						457.448 MHz	13.848 MHz								13.796 MHz
sw-ticlang-bare-oz						461.648 MHz	21.648 MHz								20.696 MHz

Sources

mspm0l1306/zephyr/README.md

Hardware AES against software (ratio at one message size) x, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10
16b-ecb-mbedtls												1.23160 MHz
16b-ecb-ticlang-o2-keyin													25.632 MHz, 1 run	20.1200 MHz, 1 run
16b-ecb-ticlang-o2-keyout													29.932 MHz, 1 run	28.6200 MHz, 1 run
1kb-aes256-ecb-gcc12-o2-hal									33.648 MHz
1kb-aes256-ecb-gcc12-o2-reg									93.948 MHz
1kb-aes256-ecb-gcc12-o3-hal									25.448 MHz
1kb-aes256-ecb-gcc12-o3-reg									69.748 MHz
1kb-cbc-ticlang-o2-driver-polling							45.9 to 93.748 MHz
1kb-cbc-ticlang-o3-driver-polling							45.8 to 93.948 MHz
1kb-ecb-mbedtls												0.83160 MHz
1kb-ecb-ticlang-o2-driver-polling						128.248 MHz	51.148 MHz
1kb-ecb-ticlang-o2-keyin													41.732 MHz, 1 run	29.8200 MHz, 1 run
1kb-ecb-ticlang-o2-keyout													41.932 MHz, 1 run	29.9200 MHz, 1 run
1kb-ecb-ticlang-o3-driver-polling						131.548 MHz	51.148 MHz
1kb-gcc12-o2-hal									26 to 44.148 MHz
1kb-gcc12-o2-reg									83.9 to 90.848 MHz
1kb-gcc12-o3-hal									19.6 to 37.148 MHz
1kb-gcc12-o3-reg									65.1 to 74.448 MHz
1kb-ticlang-o2	20.3 to 36.580 MHz, 1 run			45.8 to 79.180 MHz	45.5 to 78.580 MHz
hsm-psa
none								n/a
ticlang-o2-driver-polling															92.896 MHz
ticlang-o3-driver-polling															92.896 MHz
(one way)		n/a	n/a							n/a	Not for this chip

Sources

mspm0l1306/third-party/README.md

Hardware CRC against a table x, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1
crc16-256b			324 MHz, 1 run
dma-sniffer-m33											5.63150 MHz
dma-sniffer-riscv											5.71150 MHz
gcc12-o3									8.3448 MHz
none								n/a
(one way)	13.580 MHz, 1 run	10.932 MHz, 1 run		8.0980 MHz	8.1280 MHz	n/a	n/a			7.0696 MHz, 1 run		n/a	2.7532 MHz, 1 run	1.89200 MHz, 1 run	n/a	Not for this chip

Sources

cc2755p10/README.md, mspm0l1306/third-party/README.md

Blink image flash B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P
arduino									12.9 kB48 MHz, 1 run		50.9 kB1 run
bare	4721 run	4401 run			5281 run	5.6 kB	11.5 kB	8.4 kB1 run					80032 MHz, 1 run	7 kB1 run	7 kB1 run		n/a
bare-m33											33 kB1 run
bare-riscv											39.1 kB1 run
driverlib-bare			42424 MHz, 1 run
esp-idf-v6.1-og												147.6 kB160 MHz, 1 run
freertos	3.7 kB32 MHz, 1 run	3.7 kB1 run	4.1 kB24 MHz, 1 run			11.7 kB	17.5 kB									17.6 kB
mcux-bare										3.7 kB12 MHz, 1 run
rust-embassy		2.6 kB32 MHz, 1 run
st-hal-bare									3.8 kB48 MHz, 1 run
stm32g4-hal														8.4 kB1 run
zephyr	11.7 kB1 run	11.1 kB1 run	Not for this chip	11.8 kB1 run	Not for this chip	9.9 kB			16.5 kB48 MHz, 1 run	17.1 kB96 MHz, 1 run

Sources

am13e230/stm32g4-hal/run_all.sh, cc2755p10/README.md, cc35x1/README.md, mspm0c1104/freertos/README.md, mspm0g3519/FAQ.md, mspm0l1306/freertos/README.md, mspm0l1306/zephyr/README.md, ports/rust-mspm0/README.md

Blink image RAM B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P
arduino									91648 MHz, 1 run		9.5 kB1 run
bare	5121 run	2561 run			2561 run	9.9 kB	18.6 kB	1.5 kB1 run					25632 MHz, 1 run	9 kB1 run	10.1 kB1 run		93.8 kB1 run
bare-m33											2.6 kB1 run
bare-riscv											2.3 kB1 run
driverlib-bare			6424 MHz, 1 run
esp-idf-v6.1-og												48.5 kB160 MHz, 1 run
freertos	1 kB32 MHz, 1 run	1 kB1 run	1 kB24 MHz, 1 run			13 kB	37.6 kB									36.9 kB
mcux-bare										3.2 kB12 MHz, 1 run
rust-embassy		11232 MHz, 1 run
st-hal-bare									1.6 kB48 MHz, 1 run
stm32g4-hal														9 kB1 run
zephyr	3.9 kB1 run	3.8 kB1 run		4 kB1 run		4 kB			4.8 kB48 MHz, 1 run	3.8 kB96 MHz, 1 run

Sources

Zephyr upstream support status

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	IWRL6432BOOST	LP-AM263P
out-of-tree port		Works1 run	Not for this chip														Works400 MHz, 1 run
upstream	Works1 run	Not for this chip	Not for this chip	Works	Not for this chip	Works	Not for this chip	Not for this chip	Works48 MHz, 1 run	Works1 run	partly1 run		Works	Not for this chip	Not for this chip	Not for this chip	Not for this chip
upstream-4.4.2-spi	Not for this chip			Not for this chip
upstream-main-boosterpack												Not for this chip
upstream-main-micropython												Works160 MHz, 1 run
upstream-main-spi	Works1 run			Works1 run
vendor-fork						Not for this chip								Works96 MHz, 1 run

Sources

am263p/README.md, cc1354p10/README.md, cc2340r5/zephyr-ti/README.md, cc2340r5/zephyr/README.md, cc2755p10/README.md, cc35x1/README.md, docs/experiments/2026-10-08-zephyr-lp-mspm0g3519.md, docs/experiments/2026-10-09-am13e230-zephyr-upstream.md, iwrl6432boost/README.md, mspm0g3507/zephyr/README.md, mspm0l1306/zephyr/README.md, ports/micropython/zephyr-mspm33c321a/README.md

RTOS task switch round trip us, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-AM263P
freertos		12 to 1632 MHz, 1 run	19.9 to 25.324 MHz		7 to 8.9880 MHz	10.6 to 1448 MHz	7.38 to 9.4248 MHz	Not for this chip	9.29 to 13.248 MHz	2.7696 MHz, 1 run		16.5160 MHz	11.5 to 14.532 MHz	1.49 to 2.1200 MHz, 1 run	3.5 to 4.4596 MHz
freertos-lean			16.3 to 21.724 MHz
freertos-m33											1.91150 MHz
freertos-o3						9.98 to 1348 MHz	7.25 to 9.1948 MHz								3.44 to 4.5296 MHz
freertos-riscv											2.58150 MHz
freertos-ti-example-notify																5.13400 MHz, 1 run
freertos-ti-example-semaphore																6.06400 MHz, 1 run
zephyr	6.7280 MHz, 1 run			6.9980 MHz, 1 run		10.5 to 12.648 MHz, 1 run			18.8 to 21.848 MHz, 1 run

Sources

mspm0c1104/freertos/README.md, mspm0l1306/freertos/README.md

Minimal tasks that fit tasks, higher is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-EM-CC2755P10
freertos	1132 MHz, 1 run	124 MHz, 1 run					Not for this chip	14948 MHz, 1 run			513 to 514160 MHz
freertos-heap			20580 MHz	5180 MHz	4148 MHz	38848 MHz			996 MHz, 1 run			39732 MHz, 1 run	23896 MHz
freertos-heap-m33										775150 MHz
freertos-heap-riscv										731150 MHz
freertos-static					4148 MHz	40848 MHz							25296 MHz
freertos-static48				12280 MHz
freertos-unchecked		224 MHz, 1 run

Sources

mspm0c1104/freertos/README.md, mspm0g3519/RESULTS-mspm0g3519.md, mspm0l1306/freertos/README.md

AES-128 CBC encrypt of 1 kB in software (tiny-AES-c) cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10
gcc12-o2-flash								538k48 MHz
gcc12-o3-flash								405k48 MHz
ticlang-o2-flash	435k80 MHz, 1 run	416k32 MHz, 1 run	439k80 MHz	439k80 MHz	424k48 MHz	249k48 MHz	772k120 MHz, 1 run		316k32 MHz, 1 run	359k200 MHz, 1 run	233k96 MHz
ticlang-o2-ram							628k120 MHz, 1 run
ticlang-o3-flash						249k48 MHz					234k96 MHz
ticlang-o4s5-ram							405k120 MHz, 1 run
tinyaes-o2

Sources

Run records are in the notebook, not this copy.

AES-128 in hardware, cycles per byte cycles/B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	TMDS62LEVM (AM62L EVM)
16b-cbc-ticlang-o2-keyin				125.548 MHz	129.1 to 129.248 MHz				72.296 MHz
16b-cbc-ticlang-o3-keyin				126.148 MHz	128.848 MHz				71.296 MHz
16b-ecb-gcc12-o2-keyin-hal						43.848 MHz
16b-ecb-gcc12-o2-keyin-reg						14.548 MHz
16b-ecb-gcc12-o3-keyin-hal						39.848 MHz
16b-ecb-gcc12-o3-keyin-reg						8.8848 MHz
16b-ecb-ticlang-o2-keyin				73.648 MHz	91.648 MHz		15.132 MHz, 1 run	21.5200 MHz, 1 run	40.296 MHz
16b-ecb-ticlang-o2-keyout							10.332 MHz, 1 run	12.2200 MHz, 1 run
16b-ecb-ticlang-o3-keyin				72.948 MHz	91.848 MHz				39.496 MHz
1kb-cbc-gcc12-o2-keyin-hal						19.848 MHz
1kb-cbc-gcc12-o2-keyin-reg						6.1648 MHz
1kb-cbc-gcc12-o3-keyin-hal						19.848 MHz
1kb-cbc-gcc12-o3-keyin-reg						5.9748 MHz
1kb-cbc-ticlang-o2-keyin	20.980 MHz, 1 run	9.3580 MHz	9.4280 MHz	3.9948 MHz	5.2948 MHz				2.8496 MHz
1kb-cbc-ticlang-o3-keyin				3.9848 MHz	5.348 MHz				2.8496 MHz
1kb-ecb-gcc12-o2-keyin-hal						19.848 MHz
1kb-ecb-gcc12-o2-keyin-reg						6.1348 MHz
1kb-ecb-gcc12-o3-keyin-hal						19.748 MHz
1kb-ecb-gcc12-o3-keyin-reg						5.9548 MHz
1kb-ecb-ticlang-o2-keyin	1780 MHz, 1 run	9.3380 MHz	9.3980 MHz	3.1748 MHz	4.6548 MHz		7.4332 MHz, 1 run	11.8200 MHz, 1 run	2.3996 MHz
1kb-ecb-ticlang-o2-keyout							7.3532 MHz, 1 run	11.7200 MHz, 1 run
1kb-ecb-ticlang-o3-keyin				3.1748 MHz	4.6448 MHz				2.3896 MHz
accelerator										Not for this chip
hsm-psa

Sources

am62l-evm/README.md

AES-128 in software (tiny-AES-c), cycles per byte cycles/B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	TMDS62LEVM (AM62L EVM)
16b-ecb-ticlang-o2-keyin					501.948 MHz	311.6 to 311.848 MHz			385.132 MHz, 1 run	432.9200 MHz, 1 run	288.896 MHz
16b-ecb-ticlang-o2-keyout					400.448 MHz	231.448 MHz			308.332 MHz, 1 run	348.1200 MHz, 1 run	216.496 MHz
16b-ecb-ticlang-o3-keyin					512.648 MHz	310.848 MHz					288.496 MHz
16b-ecb-ticlang-o3-keyout					410.848 MHz	231.148 MHz					216.496 MHz
16kb-cbc-openssl-armv8ce												1.331250 MHz, 1 run
1kb-cbc-gcc12-o2-flash-keyout								525.448 MHz
1kb-cbc-gcc12-o3-flash-keyout								395.548 MHz
1kb-cbc-ticlang-o2-flash-keyout	424.780 MHz, 1 run	405.932 MHz, 1 run	428.680 MHz	428.680 MHz			754.3120 MHz, 1 run
1kb-cbc-ticlang-o2-keyout					404.748 MHz	235.448 MHz					220.396 MHz
1kb-cbc-ticlang-o2-ram-keyout							613.7120 MHz, 1 run
1kb-cbc-ticlang-o3-keyout					417.448 MHz	235.148 MHz					221.596 MHz
1kb-cbc-ticlang-o4s5-ram-keyout							395.4120 MHz, 1 run
1kb-ecb-ticlang-o2-keyin									309.332 MHz, 1 run	350.5200 MHz, 1 run
1kb-ecb-ticlang-o2-keyout					397.948 MHz	229.748 MHz			308.232 MHz, 1 run	349.4200 MHz, 1 run	214.596 MHz
1kb-ecb-ticlang-o3-keyout					408.248 MHz	229.348 MHz					214.296 MHz
tinyaes-o2

Sources

am62l-evm/README.md

Memory bandwidth, STREAM-style triad MB/s, higher is better

Variant	LP-AM263P	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
bytecopy-64k-ocram-beside-loop-core1	59.2400 MHz
bytecopy-64k-ocram-beside-loop-core2	58.6400 MHz
bytecopy-64k-ocram-beside-loop-core3	61.4400 MHz
bytecopy-64k-ocram-cores1to3-core1	59.2400 MHz
bytecopy-64k-ocram-cores1to3-core2	58.6400 MHz
bytecopy-64k-ocram-cores1to3-core3	61.4400 MHz
copy-1t			1,6351250 MHz, 1 run
copy-2t			2,418.51250 MHz, 1 run
triad-1t		4,789	1,521.51250 MHz, 1 run
triad-1t-3runs		4,793 to 4,795
triad-2t		5,3852000 MHz	2,392.11250 MHz, 1 run
wordcopy-64k-ocram-beside-loop-core1	247.4 to 247.5400 MHz
wordcopy-64k-ocram-beside-loop-core2	235.4400 MHz
wordcopy-64k-ocram-beside-loop-core3	296.5400 MHz
wordcopy-64k-ocram-cores1to3-core1	247.5 to 248400 MHz
wordcopy-64k-ocram-cores1to3-core2	235.3 to 235.4400 MHz
wordcopy-64k-ocram-cores1to3-core3	296.5400 MHz
wordcopy-nearcore0-64k-ocram-beside-loop-core1	326.7400 MHz
wordcopy-nearcore0-64k-ocram-beside-loop-core2	261.8400 MHz
wordcopy-nearcore0-64k-ocram-beside-loop-core3	308.5 to 308.6400 MHz
wordcopy-nearcore0-64k-ocram-cores1to3-core1	327.2 to 327.4400 MHz
wordcopy-nearcore0-64k-ocram-cores1to3-core2	262.2 to 262.3400 MHz
wordcopy-nearcore0-64k-ocram-cores1to3-core3	309.1 to 309.3400 MHz

Sources

am62l-evm/README.md

TI OptiFlash benchmark: 16 tasks, no copy (the 0 B row), code where the variant says ms, lower is better

Variant	LP-AM263P
ocram	14.6400 MHz
xip	126.6 to 129400 MHz
xip-l2-16k	91.8400 MHz
xip-l2-32k	65.5400 MHz
xip-l2-64k	34.6400 MHz
xip-l2-8k	110.4 to 111.2400 MHz
xip-sp	13.3400 MHz

Sources

Run records are in the notebook, not this copy.

Hardware video decode throughput fps, higher is better

Variant	SK-TDA4VM
1080p-h264-1	66.7 to 67.5
1080p-h264-2	91.11 run
1080p-h264-4	89.41 run
1080p-h264-zerocopy	Not for this chip
1280x768-h264-1	4521 run
1280x768-h264-2	4611 run
1280x768-h264-4	4491 run
1280x768-h264-zerocopy	499.61 run

Sources

Run records are in the notebook, not this copy.

Hardware video encode throughput fps, higher is better

Variant	SK-TDA4VM
1080p-h264	42
1080p-h264-hw-transcode	53.7
1080p-jpeg-software	411 run
720p-h264	1311 run

Sources

Run records are in the notebook, not this copy.

Holoscan Sensor Bridge device emulation: CPU cost of the packetizer process %, lower is better

Variant	SK-TDA4VM
1080p-raw8-30fps-ticks	37.21 run
1080p-raw8-30fps-wire-perf	37.41998 MHz, 1 run
1080p-raw8-60fps-mtu1500-perf	98.11998 MHz, 1 run
1080p-raw8-60fps-mtu9000-perf	371998 MHz, 1 run
1080p-raw8-60fps-ticks	52.21 run
720p-raw8-30fps-ticks	16.51 run
720p-raw8-30fps-wire-perf	21.91998 MHz, 1 run
720p-raw8-60fps-ticks	331 run

Sources

Run records are in the notebook, not this copy.

Holoscan Sensor Bridge stream: frames received intact by the host-side check %, higher is better

Variant	SK-TDA4VM
1080p-raw8-30fps-wire-macos	Not for this chip
720p-raw8-10fps-python	1001 run
720p-raw8-10fps-wire-macos	1001 run
720p-raw8-30fps-python	Not for this chip
720p-raw8-30fps-wire-macos	1001 run

Sources

Run records are in the notebook, not this copy.

Software H.264 decode (baseline profile, no video hardware): frames per second fps, higher is better

Variant	LP-MSPM33C321A	LP-AM13E230	TMDS62LEVM (AM62L EVM)
h264-1280x720-main-600k30-ffmpeg8-t1			52.61250 MHz
h264-1280x720-main-600k30-ffmpeg8-t2			771250 MHz
h264-1280x720-main-600k30-h264bsd-o3			Not for this chip
h264-1280x720-ref1-600k30-ffmpeg8-t1			75.71250 MHz
h264-1280x720-ref1-600k30-ffmpeg8-t2			110.31250 MHz
h264-1280x720-ref1-600k30-h264bsd-o3			40.81250 MHz
h264-128x96-ref1-os		149.8200 MHz
h264-160x112-ref1-os		109.9200 MHz
h264-176x144-ref1-o3	59.8160 MHz	84.7200 MHz
h264-176x144-ref1-os	57.4160 MHz	86.1200 MHz
h264-1920x1080-main-600k30-ffmpeg8-t1			23.51250 MHz
h264-1920x1080-main-600k30-ffmpeg8-t2			35.51250 MHz
h264-1920x1080-ref1-600k30-ffmpeg8-t1			341250 MHz
h264-1920x1080-ref1-600k30-ffmpeg8-t2			50.21250 MHz
h264-1920x1080-ref1-600k30-h264bsd-o3			171250 MHz
h264-240x240-ref1-o3	34.8160 MHz
h264-240x240-ref1-os	33.3160 MHz
h264-320x240-ref1-600k30-ffmpeg8-t1			451.11250 MHz
h264-320x240-ref1-600k30-ffmpeg8-t2			538.61250 MHz
h264-320x240-ref1-os
h264-540x540-ref1-ffmpeg8-t1			172.11250 MHz
h264-540x540-ref1-ffmpeg8-t2			246.51250 MHz
h264-540x540-ref1-h264bsd-o3			97.51250 MHz
h264-640x360-bbb-ffmpeg8-t1			174.71250 MHz
h264-640x360-bbb-ffmpeg8-t2			245.11250 MHz
h264-640x360-bbb-h264bsd-o3			102.81250 MHz
h264-640x360-main-600k30-ffmpeg8-t1			146.51250 MHz
h264-640x360-main-600k30-ffmpeg8-t2			210.41250 MHz
h264-640x360-ref1-600k30-ffmpeg8-t1			231.71250 MHz
h264-640x360-ref1-600k30-ffmpeg8-t2			310.21250 MHz
moonlight-embedded-local-test-stream

Sources

docs/experiments/2026-10-23-am62l-h264-decode.md, portability/h264/SOURCE.md

RAM the software H.264 decode needs: decoder heap peak plus input buffer plus RGB565 strip B, lower is better

Variant	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC35X1	TMDS62LEVM (AM62L EVM)
h264-1280x720-ref1-600k30-h264bsd-o3				3.5 MB1250 MHz
h264-1280x720-ref1-600k30-os	Not for this chip 3.2 MB short	Not for this chip 3.3 MB short	Not for this chip 2.98 MB short
h264-128x96-ref1-os		55.3 kB200 MHz
h264-160x112-ref1-os		76.8 kB200 MHz
h264-176x144-ref1-o3	108.3 kB160 MHz	105.5 kB200 MHz
h264-176x144-ref1-os	108.3 kB160 MHz	105.5 kB200 MHz
h264-1920x1080-ref1-600k30-h264bsd-o3				8 MB1250 MHz
h264-1920x1080-ref1-600k30-os	Not for this chip 7.5 MB short	Not for this chip 7.6 MB short	Not for this chip 7.3 MB short
h264-240x240-ref1-o3	233.5 kB160 MHz
h264-240x240-ref1-os	233.5 kB160 MHz	Not for this chip 99.8 kB short
h264-320x240-ref1-os	Not for this chip 52.6 kB short	Not for this chip 173.1 kB short
h264-480x270-ref1-600k30-os	Not for this chip 260.9 kB short	Not for this chip 381.4 kB short	Not for this chip 69.4 kB short
h264-540x540-ref1-h264bsd-o3				1.2 MB1250 MHz
h264-540x540-ref1-os	Not for this chip 885.8 kB short	Not for this chip 1,007.4 kB short	Not for this chip 694.2 kB short
h264-640x360-bbb-h264bsd-o3				1.6 MB1250 MHz
h264-640x360-ref1-600k30-os	Not for this chip 657.5 kB short	Not for this chip 778 kB short	Not for this chip 466 kB short

Sources

docs/experiments/2026-10-23-am62l-h264-decode.md, portability/h264/SOURCE.md

OpenGL ES on the GPU: kmscube or glmark2 frames per second fps, higher is better

Variant	SK-TDA4VM
kmscube

Sources

Run records are in the notebook, not this copy.

Parallel RGB scan-out from a frame buffer: frames per second with every frame's pixel sum checked against the buffer fps, higher is better

Variant	LP-AM263P
pru-rgb565-480x480-22mhz	76.3400 MHz
pru-rgb565-480x480-22mhz-r5f-copy-same-bank	76.3400 MHz
pru-rgb565-800x480-22mhz	45.6400 MHz
pru-rgb565-800x480-22mhz-r5f-copy-same-bank	Not for this chip
pru-rgb565-800x480-28mhz	Not for this chip

Sources

docs/experiments/2026-10-11-am263p-pru-rgb-scanout.md

PRU read bandwidth into its registers, 64-byte blocks, data checked MB/s, higher is better

Variant	LP-AM263P
lbbo32x2-ocsram	54400 MHz
lbbo64-ocsram	56.9400 MHz
lbbo64-pru-dram	640400 MHz
lbbo64-pru-shared	640400 MHz
xfr2vbus-ocsram	48.9400 MHz

Sources

docs/experiments/2026-10-11-am263p-pru-rgb-scanout.md

What a display scan-out costs the CPU: change in a measured CPU workload's cycles with the scan-out running %, lower is better

Variant	LP-AM263P
pru-h264-320x240	0.01400 MHz

Sources

docs/experiments/2026-10-11-am263p-pru-rgb-scanout.md

CPU cycles per single word load from each memory (unrolled independent loads) cycles, lower is better

Variant	LP-EM-CC35X1
code-ram	1.6160 MHz, 1 run
dram	6.2160 MHz, 1 run
flash	44160 MHz, 1 run
tcm	1.06160 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Data-cache hit and miss counters readable from the application yes/no

Variant	LP-EM-CC35X1
data-cache-hit-and-miss

Sources

docs/experiments/2026-10-20-doom-smallest-ti-chip.md

External flash read mode or clock changeable from the application yes/no

Variant	LP-EM-CC35X1
external-flash-read-mode

Sources

docs/experiments/2026-10-20-doom-smallest-ti-chip.md

Size of the on-board QSPI NOR flash B

Variant	LP-MSPM33C321A
onboard-nor-jedec-c22018	16 MB160 MHz, 1 run

Sources

docs/experiments/2026-10-21-mspm33c321a-qspi-dump-doom-data.md

Time to read 256 bytes from the on-board QSPI flash us, lower is better

Variant	LP-MSPM33C321A
quad-io-0xeb-40mhz-sdk-polled	23.8160 MHz, 1 run

Sources

docs/experiments/2026-10-21-mspm33c321a-qspi-dump-doom-data.md

CPU cycles per instruction of a whole program, from its time and an emulator instruction count cycles/instr, lower is better

Variant	LP-MSPM33C321A
doom-m33whx-board	1.65 calc.

Sources

docs/experiments/2026-10-21-mspm33c321a-qspi-dump-doom-data.md

Time for the Lisp Tak benchmark (tak 18 12 6) s, lower is better

Variant	LP-MSPM0G5187
ulisp-4.9a	7.580 MHz

Sources

Run records are in the notebook, not this copy.

Time for the Lisp (fib 23) benchmark s, lower is better

Variant	LP-MSPM0G5187
ulisp-4.9a	5.280 MHz

Sources

Run records are in the notebook, not this copy.

Time for the Lisp (fib 20) benchmark s, lower is better

Variant	LP-MSPM0G5187
ulisp-4.9a	1.280 MHz

Sources

Run records are in the notebook, not this copy.

Time for the Lisp queens benchmark (q2 7 8) s, lower is better

Variant	LP-MSPM0G5187
ulisp-4.9a	19.280 MHz

Sources

Run records are in the notebook, not this copy.

Time of one garbage collection in the Lisp interpreter us, lower is better

Variant	LP-MSPM0G5187
ulisp-4.9a	1,07480 MHz

Sources

Run records are in the notebook, not this copy.

Free objects in the Lisp workspace after boot and a collection count, higher is better

Variant	LP-MSPM0G5187
ulisp-4.9a	2,80080 MHz

Sources

Run records are in the notebook, not this copy.

The Lisp interpreter reads an ADC channel yes/no

Variant	LP-MSPM0G5187
ulisp-4.9a

Sources

ports/ulisp/README.md

The Lisp interpreter saves its workspace to flash and restores it yes/no

Variant	LP-MSPM0G5187
ulisp-4.9a

Sources

ports/ulisp/README.md

A long program pasted in one burst while another runs loses no characters yes/no

Variant	LP-MSPM0G5187
ulisp-4.9a	no80 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Analog: the other tests in this class

Fastest ADC with DMA Msps, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	LP-MSPM33C321A	LP-EM-CC2755P10	LP-EM-CC35X1
12bit-s2			0.84124 MHz, 1 run
12bit-s4			0.73924 MHz, 1 run
12bit-s8			0.59524 MHz, 1 run
dma										2.5396 MHz, 1 run
internal									1.7148 MHz
pico-sdk-dma-m33											0.499150 MHz
pico-sdk-dma-riscv											0.5150 MHz
ti-adcbuf-dma						1.248 MHz	0.248 MHz						1.296 MHz
ti-continuous-36cyc								2.1120 MHz, 1 run
ti-continuous-9cyc								3.98120 MHz, 1 run
ti-example-single-block												Not for this chip
(one way)	3.9332 MHz, 1 run	1.4532 MHz, 1 run	124 MHz	3.9332 MHz, 1 run	0.54732 MHz							5160 MHz, 1 run		0.17160 MHz

Sources

cc35x1/adc/build.sh, mspm33c321a/tools/build_adc.sh

DAC to ADC loop error %, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	NUCLEO-U083RC	LP-AM13E230	LP-EM-CC2755P10
dac10-ain16								9.4200 MHz
internal							Not for this chip
pwm-rc
(one way)	0.3432 MHz	n/a	n/a	0.05532 MHz, 1 run	n/a	n/a			n/a

Sources

cc2755p10/README.md

Fastest clean UART loopback Mbaud, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LAUNCHXL-F28P65X
internal							Not for this chip
internal-loopback											12.5200 MHz, 1 run
uart0
uart0-to-uart0										51 run
uart1									396 MHz
uc0-to-uc1								6.25200 MHz, 1 run
wire						7.5120 MHz, 1 run
(one way)	10.732 MHz	10.732 MHz	824 MHz	10.732 MHz	10.732 MHz

Sources

am13e230/tools/build_example.sh, cc35x1/uartloop/uart_loop.c

Power: the other tests in this class

Deepest sleep with RAM kept (STANDBY-class) uA, lower is better

Variant	LP-MSPM0L1306	LAUNCHXL-F2800137
ddr-self-refresh-rtcwake-mem
energytrace-3.3v
energytrace-3.3v-standby0	6.21 run
meter-3.3v
msp430-lpm3-wake
ti-c2000-standby-borrowed-energytrace		2,590120 MHz, 1 run
(one way)

Sources

am62l-evm/README.md, cc1354p10/README.md, ports/msp430-to-mspm0/lpm3-wake/PORT.md

Energy per BLE connection (BlueJoule job) uJ, lower is better

Variant	LP-MSPM0L1306	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC
none				n/a
(one way)	n/a				n/a

Sources

cc1354p10/README.md, mspm0l1306/README.md

Energy per CoreMark iteration uJ, lower is better

Variant	LP-MSPM0L1306	LAUNCHXL-F2800137
meter
ti-c2000-o4s5-flash-borrowed-energytrace		828.1120 MHz
ticlang-bare-o3-4mhz-energytrace	175.54 MHz, 1 run
ticlang-bare-o3-energytrace	109.632 MHz, 1 run
(one way)

Sources

cc1354p10/README.md

Energy per occupancy inference uJ, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230
Room occupancy: small, TI neural-network compiler, on the CPU, EnergyTrace			673.480 MHz
Room occupancy: small, TI neural-network compiler, on the CPU, EnergyTrace, floor subtracted			21.3 to 22.680 MHz
Room occupancy: small, TI neural-network compiler, on the NPU, EnergyTrace			81.280 MHz
Room occupancy: small, TI neural-network compiler, on the NPU, EnergyTrace, floor subtracted			2.36 to 2.680 MHz
Room occupancy: large, TI neural-network compiler, on the CPU, EnergyTrace			8,927.580 MHz
Room occupancy: large, TI neural-network compiler, on the CPU, EnergyTrace, floor subtracted			303.2 to 353.680 MHz
Room occupancy: large, TI neural-network compiler, on the NPU, EnergyTrace			3,856.580 MHz
Room occupancy: large, TI neural-network compiler, on the NPU, EnergyTrace, floor subtracted			92.3 to 120.980 MHz
Room occupancy: tiny, TI neural-network compiler, on the CPU, EnergyTrace			276.380 MHz
Room occupancy: tiny, TI neural-network compiler, on the CPU, EnergyTrace, floor subtracted			8.73 to 8.9680 MHz
Room occupancy: tiny, TI neural-network compiler, on the NPU, EnergyTrace			72.180 MHz
Room occupancy: tiny, TI neural-network compiler, on the NPU, EnergyTrace, floor subtracted			2.31 to 2.3380 MHz
Room occupancy: medium, plain C -O3, metered run
Room occupancy: medium, TI neural-network compiler, on the CPU, EnergyTrace			2,552.980 MHz
Room occupancy: medium, TI neural-network compiler, on the CPU, EnergyTrace, floor subtracted			85.7 to 94.380 MHz
Room occupancy: medium, TI neural-network compiler, on the NPU, EnergyTrace			118.980 MHz
Room occupancy: medium, TI neural-network compiler, on the NPU, EnergyTrace, floor subtracted			4.23 to 4.9280 MHz
ti-c2000-borrowed-energytrace
(one way)

Sources

cc1354p10/README.md

Active current per MHz running CoreMark (slope between two clocks) uA/MHz, lower is better

Variant	LP-MSPM0L1306	LAUNCHXL-F2800137
energytrace-3.3v
energytrace-3.3v-coremark	72.932 MHz, 1 run
ti-c2000-borrowed-energytrace		263120 MHz, 1 run
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

Lightest sleep, CPU off, clocks on (SLEEP) uA, lower is better

Variant	LP-MSPM0L1306	LAUNCHXL-F2800137
energytrace-3.3v-sleep0	94032 MHz, 1 run
ti-c2000-idle-borrowed-energytrace		23,240120 MHz
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

STOP-class mode, low-speed clock kept uA, lower is better

Variant	LP-MSPM0L1306	LAUNCHXL-F2800137
energytrace-3.3v-stop0	32932 MHz, 1 run
none		Not for this chip
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

Deepest mode, RAM lost, pin wake uA, lower is better

Variant	LP-MSPM0L1306	LAUNCHXL-F2800137
energytrace-3.3v
energytrace-3.3v-shutdown	Reference: 1.500 mA raw
meter-3.3v
msp430-fram-persist
ti-c2000-halt-borrowed-energytrace		Reference: 1.500 mA raw
(one way)

For the shutdown-current reference rows, the raw 1.500 mA reading is subtracted from every other row.

Sources

cc1354p10/README.md, cc2340r5/FAQ.md, ports/msp430-to-mspm0/fram-persist/PORT.md

Deepest sleep with a 1 s timer wake and nothing to do uA, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-EM-CC35X1
energytrace-3.3v
energytrace-3.3v-standby0-1s
ti-c2000-standby-wd-borrowed-energytrace						3,540
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

Deepest sleep with RAM kept to first instruction us, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	LP-MSPM33C321A	LP-EM-CC35X1
energytrace-3.3v
msp430-lpm3-wake
pin-toggle
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md, ports/msp430-to-mspm0/lpm3-wake/PORT.md

Energy per ADC sample at 1 ksps from sleep uJ, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC
energytrace-3.3v
ti-c2000-borrowed-energytrace
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

Energy per 1 kB flash write uJ, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC
energytrace-3.3v
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

Energy per radio transmit event (one advertisement or one packet) uJ, lower is better

Variant	LP-MSPM0L1306	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC
none	n/a
(one way)				n/a	n/a

Sources

cc1354p10/README.md, cc2340r5/FAQ.md

Days on a CR2032 (225 mAh) at the reference duty cycle: wake once a minute, 50 ms active at top clock, deepest sleep with RAM kept between days, higher is better

Variant	LP-MSPM0L1306	LP-MSPM0G5187	LAUNCHXL-F2800137
cc1407p-shed-sensor-every-1-04s-2xaa		304 calc.
cc1407p-shed-sensor-every-10s-2xaa		2,729 calc.
cc1407p-shed-sensor-every-300s-2xaa		26,178 calc.
cc1407p-shed-sensor-every-60s-2xaa		11,979 calc.
energytrace-3.3v	1,12032 MHz, 1 run
ti-c2000-borrowed-energytrace			3.6120 MHz, 1 run
(one way)

Sources

cc1354p10/README.md, cc2340r5/FAQ.md, site/data/radio-demos.json

Hottest thermal zone after 10 minutes of sustained load C, lower is better

Variant	SK-TDA4VM
coremark2t-plus-tidl-reset-at-60s	92.61 run

Sources

Run records are in the notebook, not this copy.

CPU clock ticks counted across one sleep (a stopped clock counts only the wake) ticks

Variant	LP-MSPM0L1306
msp430-lpm3-wake-sleep	15,171,11332 MHz
msp430-lpm3-wake-standby0	33,80332 MHz

Sources

Run records are in the notebook, not this copy.

Period between wakes from deep sleep, read by the host clock ms

Variant	LP-MSPM0L1306
msp430-lpm3-wake	997.432 MHz

Sources

Run records are in the notebook, not this copy.

Display: the other tests in this class

Full redraw of a 320x480 RGB565 SPI screen (ST7796S) ms, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LP-MSPM33C321A
doom-view-80-rows-dma-16mhz							14.6 to 17.6160 MHz, 1 run
edumkii-core-dma-16mhz							19.532 MHz
edumkii-core-dma-8mhz							3932 MHz
edumkii-core-fill-dma-16mhz							2032 MHz
edumkii-core-polled-16mhz							55.332 MHz
edumkii-core-polled-8mhz							67.632 MHz
edumkii-st7735s-128x128-16mhz							35.332 MHz, 1 run
edumkii-st7735s-128x128-8mhz							56.932 MHz, 1 run
edumkii-st7735s-core		Not for this chip
none
sharp128-ls013b7dh03-1mhz-bands-16row			41.124 MHz, 1 run
sharp128-ls013b7dh03-1mhz-bands-1row			153.724 MHz, 1 run
sharp128-ls013b7dh03-1mhz-framebuffer
spi16-dma16	168.132 MHz, 1 run			168.132 MHz, 1 run
spi20-cpu80-polled-fill				134.480 MHz
spi20-dma16	134.480 MHz, 1 run			134.480 MHz, 1 run
spi40-cpu80-polled-fill				67.280 MHz
spi40-dma16	67.280 MHz, 1 run			67.280 MHz, 1 run
spi8-cpu8	364.932 MHz, 1 run			364.832 MHz, 1 run
touch-games-2048-spi20				324.780 MHz
touch-games-2048-spi40				165.680 MHz
touch-games-menu-spi20				201.780 MHz
touch-games-menu-spi40				103.580 MHz
touch-games-mines-spi20				331.980 MHz
touch-games-mines-spi40				206.980 MHz
(one way)					Not for this chip	Not for this chip

Sources

data/benchmark-suite.json, docs/experiments/2026-10-20-doom-smallest-ti-chip.md, mspm0g3507/display-test/README.md, portability/boosterpacks/edumkii/boards/mspm0l1306/PINS.md, portability/boosterpacks/edumkii/build.sh, portability/boosterpacks/edumkii/examples/framebench/framebench.c, portability/boosterpacks/sharp128/README.md

Power-up to SPI screen initialized and on ms, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1
edumkii-screen-pattern
edumkii-st7735s-core		Not for this chip
fast	12.232 MHz, 1 run		12.232 MHz, 1 run
none
sharp128-power-to-picture
usual	370.232 MHz, 1 run		370.232 MHz, 1 run
(one way)				Not for this chip	Not for this chip

Sources

data/benchmark-suite.json, mspm0g3507/display-test/README.md, portability/boosterpacks/edumkii/boards/mspm0l1306/PINS.md, portability/boosterpacks/edumkii/build.sh, portability/boosterpacks/sharp128/README.md

LVGL demo screen refreshes per second fps, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LP-MSPM33C321A	TMDS62LEVM (AM62L EVM)
arduboy-crate-confusion-edumkii					30.832 MHz, 1 run
arduboy-snake-edumkii					32.932 MHz, 1 run
arduboy-tileworld-edumkii					30.632 MHz, 1 run
chocolate-doom-3.1.0-timedemo-demo1-kmsdrm-1080p						61250 MHz, 1 run
chocolate-doom-3.1.0-timedemo-demo1-weston-pixman-1080p						6.541250 MHz, 1 run
doom-cpu-only					94 calc.
doom-m33whx-board-cached-16kb					21.9160 MHz, 1 run
doom-m33whx-board-cached-8kb					19.9160 MHz, 1 run
doom-m33whx-cached-16kb-emulated					19.7 calc.
doom-m33whx-cpu-only					42 calc.
doom-m33whx-edumkii-c16-realtime					24.5160 MHz, 1 run
doom-m33whx-edumkii-c8-realtime					22.7160 MHz, 1 run
doom-m33whx-edumkii-c8-stepped					19.3160 MHz, 1 run
doom-m33whx-edumkii-play-by-hand
doom-m33whx-edumkii-play-e1m1					31160 MHz, 1 run
doom-on-board
doomgeneric-fbdev-timedemo-demo1-1280x1000-scale4x5						185.8 to 192.71250 MHz
doomgeneric-fbdev-timedemo-demo1-320x200-scale1x1						412.91250 MHz, 1 run
doomgeneric-fbdev-timedemo-demo1-960x800-scale3x4						240.61250 MHz, 1 run
dsda-doom-0.29-timedemo-demo1-1280x720						10.51250 MHz, 1 run
dsda-doom-0.29-timedemo-demo1-640x480						27.21250 MHz, 1 run
lvgl-9.3-drm-1080p						181250 MHz
lvgl-9.3-fbdev-1080p						54.11250 MHz
lvgl9.6-32mhz-spi16	732 MHz, 1 run	732 MHz
lvgl9.6-32mhz-spi16-buf40		10 to 1332 MHz, 1 run
lvgl9.6-80mhz-spi20		13 to 1880 MHz, 1 run
lvgl9.6-80mhz-spi40	15 to 1980 MHz, 1 run	15 to 1980 MHz
lvgl9.6-80mhz-spi40-buf40		15 to 1980 MHz, 1 run
lvgl9.6-full-demo-32mhz-spi16		732 MHz, 1 run
none
peanut-gb-tobu-edumkii					6.732 MHz, 1 run
peanut-gb-tobu-edumkii-160mhz-full
peanut-gb-tobu-edumkii-160mhz-full-model					42 calc.
peanut-gb-tobu-edumkii-160mhz-skip
peanut-gb-tobu-edumkii-160mhz-skip-model					84 calc.
peanut-gb-tobu-edumkii-160mhz-skip-paced					59.7160 MHz, 1 run
peanut-gb-tobu-edumkii-32mhz-full
peanut-gb-tobu-edumkii-32mhz-full-model					17 calc.
peanut-gb-tobu-edumkii-32mhz-skip
peanut-gb-tobu-edumkii-32mhz-skip-model					27 calc.
qt-quick-software-1080p
redroid-12-scrcpy-5.0.1-hdmi-1080p-settings-scroll						1.1 to 21250 MHz, 1 run
retroarch-1.20-gambatte-tobu-tobu-girl-hdmi-1920x1080						10.41250 MHz, 1 run
retroarch-1.20-gambatte-tobu-tobu-girl-hdmi-640x480						481250 MHz, 1 run
retroarch-core-controller
sharp128-player-fatfs
sharp128-player-pff
slint-1.17.1-linuxkms-software-1080p						41.9 to 42.11250 MHz
steam-light-2d-game-box64-llvmpipe
supertuxkart-1.5-arm64-hdmi-640x480-fullscreen						21250 MHz, 1 run
supertuxkart-1.5-arm64-profile-1280x720						2.421250 MHz, 1 run
supertuxkart-1.5-arm64-profile-640x480						4.481250 MHz, 1 run
supertuxkart-1.5-x86_64-box64-0.3.4-profile-640x480						Not for this chip
thermostat-80mhz-spi40		8 to 1080 MHz, 1 run
zephyr-lvgl9.5		Not for this chip
(one way)			Not for this chip	Not for this chip

Sources

am62l-evm/bench/logs/showcase/stk-640x480-race.jpg, data/benchmark-suite.json, docs/experiments/2026-10-20-doom-smallest-ti-chip.md, docs/experiments/2026-10-21-mspm33c321a-qspi-dump-doom-data.md, docs/experiments/2026-10-24-am62l-doom-controller.md, docs/experiments/2026-10-24-am62l-hmi-shootout.md, docs/experiments/2026-10-25-am62l-steam-box64.md, portability/boosterpacks/edumkii/examples/gameboy/gb_app.cpp, portability/boosterpacks/edumkii/examples/snake/frl_edumkii.cpp, portability/boosterpacks/sharp128/README.md, ports/doom/README.md, ports/peanut-gb/README.md, thermostat-system/g3519-zephyr/README.md

Baseline JPEG decode of one 160 x 120 frame (picojpeg), one block row at a time cycles, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G3519	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137
ticlang-o2-flash	6.37M80 MHz, 1 run	5.88M32 MHz, 1 run	6.52M80 MHz	6.61M48 MHz	3.42M48 MHz	9.43M120 MHz, 1 run
ticlang-o2-ram						7.46M120 MHz, 1 run
ticlang-o3-flash				5.62M48 MHz	2.99M48 MHz
ticlang-o4s5-ram						7.09M120 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

SPI controller transmit at a stated clock, nothing attached kB/s, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137
16bit-12mhz			1,33948 MHz	1,33948 MHz
16bit-13mhz	1,48880 MHz, 1 run
16bit-16mhz		1,78532 MHz, 1 run
16bit-24mhz-over-spec			2,67848 MHz	2,67848 MHz
16bit-30mhz					3,662120 MHz, 1 run
16bit-40mhz-over-spec	4,46380 MHz, 1 run
8bit-12mhz				1,23448 MHz
8bit-24mhz-over-spec				1,95348 MHz
8bit-30mhz					1,957120 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

SPI transmit as a percent of the wire rate %, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137
16bit-12mhz				91.448 MHz	91.448 MHz
16bit-13mhz	91.480 MHz, 1 run
16bit-16mhz		91.432 MHz, 1 run
16bit-24mhz-over-spec				91.448 MHz	91.448 MHz
16bit-30mhz						100120 MHz, 1 run
16bit-40mhz-over-spec	91.480 MHz, 1 run
8bit-12mhz				84.248 MHz	84.248 MHz
8bit-13mhz	84.280 MHz, 1 run
8bit-16mhz		84.232 MHz, 1 run
8bit-24mhz-over-spec					66.748 MHz
8bit-30mhz						53.5120 MHz, 1 run
8bit-40mhz-over-spec	84.280 MHz, 1 run
cc1407p-spi-1mhz-ofdm-opt1-mcs6-200b			7880 MHz
cc1407p-spi-2mhz-ofdm-opt1-mcs6-200b			6580 MHz
cc1407p-spi-4mhz-ofdm-opt1-mcs6-200b			4880 MHz
cc1407p-spi-8mhz-ofdm-opt1-mcs6-200b			3180 MHz

Sources

Run records are in the notebook, not this copy.

2048: one swipe (slide, spawn, redraw changed tiles) on a touch panel ms, lower is better

Variant	LP-MSPM0G3519
spi20-cpu80-first10	47.380 MHz
spi20-cpu80-mean60	61.6 to 89.680 MHz
spi40-cpu80-first10	24.980 MHz
spi40-cpu80-mean60	32.6 to 47.580 MHz

Sources

Run records are in the notebook, not this copy.

Minesweeper: one tap that redraws one cell ms, lower is better

Variant	LP-MSPM0G3519
spi20-cpu80	1.680 MHz
spi40-cpu80	1.0980 MHz

Sources

Run records are in the notebook, not this copy.

Minesweeper: first tap of a board (places mines, may open a region) ms, lower is better

Variant	LP-MSPM0G3519
spi20-cpu80	32.5 to 102.280 MHz
spi40-cpu80	21.7 to 67.980 MHz

Sources

Run records are in the notebook, not this copy.

Minesweeper: long press that places a flag ms, lower is better

Variant	LP-MSPM0G3519
spi20-cpu80	8.3880 MHz
spi40-cpu80	4.9480 MHz

Sources

Run records are in the notebook, not this copy.

One read of the touch chip registers us, lower is better

Variant	LP-MSPM0G3519
ft6336u-softi2c-5byte	88680 MHz

Sources

Run records are in the notebook, not this copy.

Deepest stack use of the touch games after the scripted run B, lower is better

Variant	LP-MSPM0G3519
touch-games	59280 MHz

Sources

Run records are in the notebook, not this copy.

Touch to changed pixel (computed from measured parts, or measured with a real finger) ms, lower is better

Variant	LP-MSPM0G3519
real-finger
swipe-2048-spi40-estimate	31 calc.
tap-one-cell-spi40-estimate	7 calc.

Sources

ports/touch-games/README.md

A person looked at the games on the panel and touch worked yes/no

Variant	LP-MSPM0G3519
2048-and-mines	yes1 run

Sources

ports/touch-games/README.md

Instructions per drawn frame of an emulator port, counted in an Arm emulator instructions, lower is better

Variant	LP-MSPM33C321A
peanut-gb-acc0	2.85M to 2.73M1 run
peanut-gb-base	2.85M to 2.73M1 run
peanut-gb-lut	1.19M to 1.09M1 run
peanut-gb-o2	1.04M to 921k1 run
peanut-gb-o3	1.06M to 935k1 run
peanut-gb-skip	593k to 488k1 run
peanut-gb-skip-ram	594k to 488k1 run
peanut-gb-zero	934k to 816k1 run

Sources

ports/peanut-gb/speed/emu_pgb.py

Emulator frames that finished after their due time, paced to the console clock count, lower is better

Variant	LP-MSPM33C321A
peanut-gb-tobu-edumkii-160mhz-skip-paced	0160 MHz, 1 run

Sources

ports/peanut-gb/README.md

Instructions per second of an interpreter port with no screen and the display wait off instructions/s, higher is better

Variant	LP-MSPM0G3519
chip8-core-flat-out	526,153 to 472,66980 MHz, 1 run

Sources

ports/chip8/README.md

Interpreter speed divided by the pace a typical program needs x, higher is better

Variant	LP-MSPM0G3519
chip8-vs-700ips	751 calc.

Sources

ports/chip8/README.md

Time to redraw the whole emulated screen to the panel us, lower is better

Variant	LP-MSPM0G3519
chip8-64x32-5x	14,21280 MHz, 1 run

Sources

ports/chip8/README.md

A finger on the panel pad reaches the program as a key press yes/no

Variant	LP-MSPM0G3519
chip8-pad

Sources

ports/chip8/README.md

A finger on the touch keyboard reaches the program as a typed character yes/no

Variant	LP-MSPM0G3519
zx48-keyboard

Sources

ports/zx-spectrum/README.md

Emulated machine speed as per mille of the real machine, 159 meaning 15.9 percent per mille, higher is better

Variant	LP-MSPM0G3519
zx48-estimate-from-coremark	200 calc.
zx48-o2-flash-core	15980 MHz, 1 run
zx48-o2-flash-total	12380 MHz, 1 run
zx48-o2-sram-core	18980 MHz, 1 run
zx48-o2-sram-total	14080 MHz, 1 run
zx48-o3-flash-core	14080 MHz, 1 run
zx48-opense-idle-sram-core	19480 MHz
zx48-opense-idle-sram-total	19080 MHz

Sources

ports/zx-spectrum/README.md

Linux: the other tests in this class

Memory load latency by pointer chase ns, lower is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
ddr-16m	169.22000 MHz, 1 run
ddr-2m	91.52000 MHz, 1 run
ddr-64m	191.62000 MHz, 1 run	271.11250 MHz, 1 run
l1-16k	2.52000 MHz, 1 run	2.411250 MHz, 1 run
l1-32k	5.092000 MHz, 1 run
l2-128k		121250 MHz, 1 run
l2-1m	24.52000 MHz, 1 run
l2-512k	13.82000 MHz, 1 run

Sources

am62l-evm/README.md

Timer wake-up latency (cyclictest), max over 60 s us, lower is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
idle	412000 MHz, 1 run	4741 run
idle-60s-x3		442 to 6451250 MHz
load-2cpu	592000 MHz, 1 run	5461250 MHz, 1 run
load-2cpu-60s-x5		1,074 to 93,6991250 MHz
openplc-v4-cycle
preempt-rt-idle

Sources

am62l-evm/README.md, docs/experiments/2026-10-24-am62l-realtime-jitter.md

Boot medium sequential read MB/s, higher is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
microsd-fio-1m	44.81 run
microsd-hdparm	43.8 to 43.9
usb-stick-dd-direct		38.11 run

Sources

am62l-evm/README.md

Boot medium sequential write MB/s, higher is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
microsd-dd-direct	30.51 run
sd-image-write-over-ssh		81 run
usb-stick-dd-direct		10.41 run
windows11-setup-apply-qcow2-on-sd-avg		2.51250 MHz, 1 run

Sources

am62l-evm/README.md, docs/experiments/2026-10-24-am62l-sd-card-and-ps5-driver.md, docs/experiments/2026-10-25-am62l-windows-arm-vm.md

Boot medium 4 kB random read IOPS, higher is better

Variant	SK-TDA4VM
microsd-fio-4k-qd8	3,8231 run

Sources

Run records are in the notebook, not this copy.

Vision app launch to the first frame-rate line s, lower is better

Variant	SK-TDA4VM
edgeai-app-yolox-pose-file	3.09 to 3.84

Sources

Run records are in the notebook, not this copy.

Linux to remote core message round trip (rpmsg echo) us, lower is better

Variant	LP-AM263P	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
freertos-to-nortos-1-remote	3.49 to 3.5400 MHz
freertos-to-nortos-3-remotes	3.81400 MHz
main-r5f-c66x-c7x
mcu-r5f0-0-rpmsg14		74 to 75
threadx-to-nortos-1-remote	3.36400 MHz
threadx-to-nortos-3-remotes	3.87400 MHz
zrpmsg-char-userspace-496b			35.1 to 35.31250 MHz
zrpmsg-char-userspace-56b			18.11250 MHz
zrpmsg-irq-496b			18.41250 MHz
zrpmsg-irq-496b-thread-wake			25.2 to 25.31250 MHz
zrpmsg-irq-56b			5.661250 MHz
zrpmsg-irq-56b-thread-wake			12.71250 MHz
zrpmsg-poll-496b			17.21250 MHz
zrpmsg-poll-56b			4.03 to 4.041250 MHz
zrpmsg-zirq-lpoll-496b			17.61250 MHz
zrpmsg-zirq-lpoll-56b			4.39 to 4.411250 MHz
zrpmsg-zpoll-lirq-496b			17.9 to 181250 MHz
zrpmsg-zpoll-lirq-56b			5.08 to 5.091250 MHz

Sources

docs/experiments/2026-10-24-am62l-linux-zephyr-link.md

Linux to remote core message stream, payload rate (1 kB = 1,000 bytes) kB/s, higher is better

Variant	TMDS62LEVM (AM62L EVM)
linux-to-zephyr-496b-irq	331,2121250 MHz
linux-to-zephyr-496b-poll	330,7681250 MHz
linux-to-zephyr-64b-irq	43,077.51250 MHz
linux-to-zephyr-64b-poll	43,1541250 MHz
zephyr-to-linux-496b-irq	146,5161250 MHz
zephyr-to-linux-496b-poll	257,7481250 MHz
zephyr-to-linux-64b-irq	19,146.51250 MHz
zephyr-to-linux-64b-poll	34,121.51250 MHz

Sources

docs/experiments/2026-10-24-am62l-linux-zephyr-link.md

Linux to remote core message stream, messages per second msgs/s, higher is better

Variant	TMDS62LEVM (AM62L EVM)
linux-to-zephyr-496b-irq	667,7671250 MHz
linux-to-zephyr-496b-poll	666,8731250 MHz
linux-to-zephyr-64b-irq	673,0971250 MHz
linux-to-zephyr-64b-poll	674,2901250 MHz
zephyr-to-linux-496b-irq	295,3961250 MHz
zephyr-to-linux-496b-poll	519,6551250 MHz
zephyr-to-linux-64b-irq	299,1701250 MHz
zephyr-to-linux-64b-poll	533,1561250 MHz

Sources

docs/experiments/2026-10-24-am62l-linux-zephyr-link.md

Remote core load during a pipeline (TI perf_stats) %

Variant	SK-TDA4VM
c7x-pose-app-30fps	351 run
c7x-yolox-zerocopy-111fps	45

Sources

Run records are in the notebook, not this copy.

DDR bandwidth used during a pipeline (TI perf_stats) MB/s

Variant	SK-TDA4VM
pose-app-30fps	4,228 to 4,322
yolox-zerocopy-111fps	1,9571 run

Sources

Run records are in the notebook, not this copy.

Wired Ethernet TCP throughput (iperf3) Mbit/s, higher is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
rx-from-host		9001 run
tcp-bidir	891.5 to 934.1
tcp-rx-from-host	939.7 to 939.9
tcp-tx-to-host	939.5 to 939.6
tx-to-host		9271 run

Sources

am62l-evm/README.md

Wired Ethernet round trip to a host on the same switch ms, lower is better

Variant	SK-TDA4VM	TMDS62LEVM (AM62L EVM)
host-same-switch		0.3681 run
wired-host-same-switch	0.313 to 0.371 run

Sources

am62l-evm/README.md

Board rail power from the on-board current monitors (sum of the monitored rails, 30 s mean) mW, lower is better

Variant	TMDS62LEVM (AM62L EVM)
app-baseline-docker-idle	567.61250 MHz, 1 run
app-ha-alone	571.51250 MHz, 1 run
app-ha-plus-mosquitto	569.51250 MHz, 1 run
app-mosquitto	569.11250 MHz, 1 run
app-nodered	571.81250 MHz, 1 run
app-pihole	572.41250 MHz, 1 run
dvdd1v8-linux-idle	34.31250 MHz, 1 run
dvdd3v3-linux-idle	78.91250 MHz, 1 run
lpddr4-linux-idle	90.61250 MHz, 1 run
total-linux-1core-busy	651.91250 MHz, 1 run
total-linux-coremark-zephyr-coremark	659.21250 MHz, 1 run
total-linux-idle	560.91250 MHz, 1 run
total-linux-idle-zephyr-coremark	633.51250 MHz, 1 run
total-linux-maxcpus1-busy-zephyr-hello	6701250 MHz, 1 run
total-linux-maxcpus1-zephyr-hello-idle	552.51250 MHz, 1 run
vdd-core-linux-idle	312.51250 MHz, 1 run
vdda1v8-linux-idle	44.61250 MHz, 1 run
zephyr-sram-ddr-off

Sources

am62l-evm/README.md, am62l-evm/linux-apps/README.md, am62l-evm/zephyr/mkphase1.sh

Memory a container app uses on a Linux board (docker stats, and whole-system RAM in use from /proc/meminfo) MB, lower is better

Variant	TMDS62LEVM (AM62L EVM)
baseline-system-used	5211250 MHz, 1 run
ha-container	5571250 MHz, 1 run
ha-plus-mosquitto-system-used	1,1311250 MHz, 1 run
ha-system-used	1,0791250 MHz, 1 run
mosquitto-container	2.61250 MHz, 1 run
mosquitto-native-rss	7.71250 MHz, 1 run
mosquitto-system-used	5981250 MHz, 1 run
nodered-container	1101250 MHz, 1 run
nodered-system-used	6391250 MHz, 1 run
pihole-container	24.81250 MHz, 1 run
pihole-system-used	5761250 MHz, 1 run
redroid-12-arm64-boot-system-used-peak	9891250 MHz, 1 run
steam-arm64-hotpatch-bigpicture-board-used-peak	1,4211250 MHz, 1 run
steam-arm64-hotpatch-board-used-peak	1,302 to 1,4091250 MHz
steam-arm64-hotpatch-lean-board-used-peak	1,224 to 1,3411250 MHz
steam-arm64-native-lsetrap-system-used	1,3291250 MHz, 1 run
steam-client-update-box64-system-used	1,0901250 MHz, 1 run
ti-lvgl-demo-lvglsim-rss	1181250 MHz, 1 run
windows11-22621-arm64-kvm-qemu-rss-peak	1,5871250 MHz, 1 run

Sources

am62l-evm/linux-apps/README.md, am62l-evm/showcase/run.sh, docs/experiments/2026-10-25-am62l-windows-arm-vm.md

CPU idle share with a container app running, both cores, 30 s %, higher is better

Variant	TMDS62LEVM (AM62L EVM)
baseline	931250 MHz, 1 run
ha-alone	91.51250 MHz, 1 run
ha-plus-mosquitto	82.51250 MHz, 1 run
mosquitto	92.61250 MHz, 1 run
nodered	88.91250 MHz, 1 run
pihole	911250 MHz, 1 run

Sources

am62l-evm/linux-apps/README.md

Seconds from docker run to the app's first useful answer s, lower is better

Variant	TMDS62LEVM (AM62L EVM)
fex-armv8a-x86-uname	11250 MHz, 1 run
ha-mqtt-sensor-after-publish	29.21250 MHz, 1 run
ha-onboarding-api	67.71250 MHz
mosquitto-port-open	2.41250 MHz, 1 run
nodered-flow-deploy-to-reply	2.51250 MHz, 1 run
nodered-ui	17.71250 MHz, 1 run
pihole-dns	4.61250 MHz, 1 run
redroid-12-arm64-boot-completed	861250 MHz, 1 run
redroid-12-arm64-boot-completed-showcase	701250 MHz, 1 run
steam-arm64-hotpatch-bigpicture-signin-painted	1371250 MHz, 1 run
steam-arm64-hotpatch-lean-signin-painted	471250 MHz
steam-arm64-hotpatch-lean-swap1g-signin-painted	461250 MHz, 1 run
steam-arm64-hotpatch-nogpu-signin-painted	441250 MHz, 1 run
steam-arm64-hotpatch-nosandbox-signin-painted	471250 MHz, 1 run
steam-arm64-hotpatch-signin-painted	46 to 491250 MHz
steam-arm64-native-lsetrap-login-window	20 to 431250 MHz, 1 run
steam-arm64-native-plain	Not for this chip
steam-arm64-trap-signin-painted	56 to 571250 MHz
steam-proton-backyard-baseball-97-window	Not for this chip
steam-x86-box64-git-box32-login-window
steam-x86-box64-login-window	Not for this chip
steam-x86-box86-armhf-login-window
windows11-22621-arm64-kvm-desktop	Not for this chip
windows11-22621-arm64-kvm-oobe-start	6,3021250 MHz, 1 run
windows11-22621-arm64-kvm-setup-copying	2701250 MHz, 1 run
windows11-22621-arm64-kvm-setup-first-reboot	5,0731250 MHz, 1 run
windows11-22621-arm64-kvm-winpe-up	801250 MHz, 1 run

Sources

am62l-evm/linux-apps/README.md, am62l-evm/showcase/run.sh, docs/experiments/2026-10-25-am62l-steam-box64.md, docs/experiments/2026-10-25-am62l-windows-arm-vm.md

Seconds to pull the container image over Ethernet to the board's SD card s, lower is better

Variant	TMDS62LEVM (AM62L EVM)
fex-rootfs-arch-download	1201250 MHz, 1 run
ha	654.91250 MHz, 1 run
mosquitto	9.11250 MHz, 1 run
nodered	197.11250 MHz, 1 run
pihole	33.41250 MHz, 1 run
steam-anonymous-depotdownloader-arm64-app1007	28.91250 MHz, 1 run
steam-anonymous-depotdownloader-arm64-app896660	3711250 MHz, 1 run
steam-client-32bit-update-box32	4361250 MHz, 1 run
steam-client-steamrt64-update-box64	3571250 MHz, 1 run
steamcmd-box64-0.3.4-box32-anonymous-app1007	Not for this chip

Sources

am62l-evm/linux-apps/README.md, docs/experiments/2026-10-25-am62l-steam-box64.md

Container image size on the board's disk (docker images) MB, lower is better

Variant	TMDS62LEVM (AM62L EVM)
ha	3,4201250 MHz, 1 run
mosquitto	43.21250 MHz, 1 run
nodered	9921250 MHz, 1 run
pihole	1661250 MHz, 1 run
steam-arm64-native-client-unpacked	3241250 MHz, 1 run
steam-client-installed-dir	2,4671250 MHz, 1 run

Sources

am62l-evm/linux-apps/README.md

Speech to text on the CPU, processing time divided by audio length (whisper.cpp, 11.0 s clip) x real time, lower is better

Variant	TMDS62LEVM (AM62L EVM)
base.en-t1	8.761250 MHz
base.en-t2	4.751250 MHz
tiny.en-t1	41250 MHz
tiny.en-t2	2.231250 MHz

Sources

docs/experiments/2026-10-24-am62l-edge-ai-cpu.md

Small language model on the CPU (llama.cpp llama-bench) tokens/s, higher is better

Variant	TMDS62LEVM (AM62L EVM)
qwen2.5-0.5b-q4_0-gen-t1	1.681250 MHz
qwen2.5-0.5b-q4_0-gen-t2	2.891250 MHz
qwen2.5-0.5b-q4_0-prompt-t1	2.841250 MHz
qwen2.5-0.5b-q4_0-prompt-t2	5.231250 MHz

Sources

docs/experiments/2026-10-24-am62l-edge-ai-cpu.md

One model inference on the CPU, median latency ms, lower is better

Variant	TMDS62LEVM (AM62L EVM)
silero-en_v5-ort-t1	1,4401250 MHz
silero-en_v5-ort-t2	8681250 MHz
yamnet-litert-t1	30.41250 MHz
yamnet-litert-t2	20.21250 MHz

Sources

docs/experiments/2026-10-24-am62l-edge-ai-cpu.md

Frames per second an emulator core delivers to the HDMI connector, locked to vblank (free-run rows: no vblank wait) fps, higher is better

Variant	TMDS62LEVM (AM62L EVM)
gambatte-tobu-tobu-girl-1920x1080	601250 MHz
gambatte-tobu-tobu-girl-640x480	601250 MHz
pcsx-rearmed-gpu-benchmark-1280x720	601250 MHz
pcsx-rearmed-gpu-benchmark-1920x1080	601250 MHz
pcsx-rearmed-gpu-benchmark-1920x1080-thread	601250 MHz
pcsx-rearmed-gpu-benchmark-640x480	601250 MHz
pcsx-rearmed-gpu-benchmark-640x480-audio	601250 MHz
pcsx-rearmed-gpu-benchmark-640x480-freerun	250.3 to 254.81250 MHz
snes9x2005-effects-music-1920x1080-audio-thread	601250 MHz
snes9x2005-effects-music-640x480-audio	601250 MHz
snes9x2005-mode7-1280x720	601250 MHz
snes9x2005-mode7-1920x1080	601250 MHz
snes9x2005-mode7-1920x1080-thread	601250 MHz
snes9x2005-mode7-640x480	601250 MHz
snes9x2005-mode7-800x600	60.31250 MHz
snes9x2005-mode7-plane-scaling-2x-4x	Not for this chip
snes9x2005-transparency-1920x1080	601250 MHz
snes9x2005-transparency-1920x1080-freerun	193.9 to 194.31250 MHz
snes9x2005-transparency-640x480	601250 MHz
snes9x2005-transparency-640x480-freerun	272.3 to 273.11250 MHz

Sources

docs/experiments/2026-11-01-am62l-lean-emulation.md

Time the emulator core spends in one retro_run frame, video callback excluded ms, lower is better

Variant	TMDS62LEVM (AM62L EVM)
gambatte-tobu-tobu-girl-1920x1080	0.87 to 0.881250 MHz
gambatte-tobu-tobu-girl-640x480	0.871250 MHz
pcsx-rearmed-gpu-benchmark-1280x720	3.7 to 3.741250 MHz
pcsx-rearmed-gpu-benchmark-1920x1080	3.86 to 3.891250 MHz
pcsx-rearmed-gpu-benchmark-1920x1080-thread	5.15 to 5.161250 MHz
pcsx-rearmed-gpu-benchmark-640x480	3.61 to 3.621250 MHz
pcsx-rearmed-gpu-benchmark-640x480-audio	3.651250 MHz
pcsx-rearmed-gpu-benchmark-640x480-freerun	3.61 to 3.721250 MHz
snes9x2005-effects-music-1920x1080-audio-thread	1.11 to 1.121250 MHz
snes9x2005-effects-music-640x480-audio	0.73 to 0.741250 MHz
snes9x2005-mode7-1280x720	2.56 to 2.571250 MHz
snes9x2005-mode7-1920x1080	2.58 to 2.631250 MHz
snes9x2005-mode7-1920x1080-thread	3.11 to 3.171250 MHz
snes9x2005-mode7-640x480	2.48 to 2.551250 MHz
snes9x2005-mode7-800x600	2.56 to 2.581250 MHz
snes9x2005-transparency-1920x1080	3.55 to 3.591250 MHz
snes9x2005-transparency-1920x1080-freerun	3.53 to 3.631250 MHz
snes9x2005-transparency-640x480	3.39 to 3.41250 MHz
snes9x2005-transparency-640x480-freerun	3.29 to 3.471250 MHz

Sources

docs/experiments/2026-11-01-am62l-lean-emulation.md

Time from a gamepad button event (kernel timestamp) to the vblank at which the frame drawn after it is flipped onto the connector ms, lower is better

Variant	TMDS62LEVM (AM62L EVM)
snes9x2005-breakout-1920x1080	31.9 to 155.91250 MHz, 1 run

Sources

docs/experiments/2026-11-01-am62l-lean-emulation.md

Sweep: the other tests in this class

Vendor examples that build from a terminal %, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-CC1354P10-1	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	Raspberry Pi Pico 2 H	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P	SK-TDA4VM
all-sdks						98.11 run
am13e230-sdk-26.01													100200 MHz, 1 run
c2000ware-26.02							1001 run
edgeai-gst-apps-11.00.00.08																	n/a
esp-idf-v6.1											93.5160 MHz, 1 run
mcu-plus-sdk-26.01-lp																249 to 284400 MHz, 1 run
mcux-26.09.00									981 run
mspm0-sdk-2.11	971 run	98.21 run	1001 run	971 run	85.71 run
mspm33-sdk-1.04												10032 MHz, 1 run
pico-sdk-2.3.1-arm										1001 run
pico-sdk-2.3.1-riscv										1001 run
sdk-f3-9.21														1001 run
stm32cubeu0-1.3.0								1001 run
wifi-sdk-10.20.00.39															96.91 run
(one way)

Sources

cc35x1/prebuild.sh, data/benchmark-suite.json

Examples run with no crash %, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LAUNCHXL-F2800137	NUCLEO-U083RC	FRDM-MCXA153	ESP32-C6-DevKitC-1-N8	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	LP-AM263P	SK-TDA4VM
all-sdks
am13e230-sdk-26.01												98.8200 MHz, 1 run
c2000ware-26.02							100120 MHz, 1 run
cc35x1-twins-checked						10048 MHz, 1 run
edgeai-gst-apps-11.00.00.08																601 run
esp-idf-v6.1										87.1160 MHz, 1 run
esp-idf-v6.1-sleep
f3-app-twins-programmed														100160 MHz, 1 run
f3-app-twins-serial-checked														54.5160 MHz, 1 run
f3-app-twins-silent
g3519-port-adc12_max_freq_dma					10080 MHz, 1 run
g3519-port-adc12_max_freq_dma_8bit
g3519-port-adc12_monitor_supply
g3519-port-adc12_sequence_conversion					10080 MHz, 1 run
g3519-port-adc12_single_conversion
g3519-port-adc12_single_conversion_vref_external
g3519-port-adc12_single_conversion_vref_internal
g3519-port-adc12_triggered_by_timer_event
g3519-port-adc12_triggered_by_timer_event_stop
g3519-port-adc12_window_comparator
g3519-port-aesadv_cbc_128_enc_dec
g3519-port-aesadv_ecb_128_enc_dec
g3519-port-aesadv_ecb_256_encrypt
g3519-port-blinky_asm					Not for this chip
g3519-port-cinit_bypass
g3519-port-comm_modules-i2c_controller
g3519-port-comm_modules-i2c_target
g3519-port-comm_modules-uart_controller
g3519-port-comm_modules-uart_target
g3519-port-comp_analog_filter
g3519-port-comp_dac_to_timer_event
g3519-port-comp_hs_tima_pwm_fault
g3519-port-comp_lp_dac_vref_internal
g3519-port-crcp_calculate_checksum					10080 MHz, 1 run
g3519-port-crcp_calculate_checksum_dma					10080 MHz, 1 run
g3519-port-dma_block_transfer					Not for this chip
g3519-port-dma_fill_data					Not for this chip
g3519-port-dma_table_transfer					Not for this chip
g3519-port-drivers-uart_echo					Not for this chip
g3519-port-empty
g3519-port-empty_asm
g3519-port-empty_cpp
g3519-port-empty_driverlib_src
g3519-port-event_input_triggers_output
g3519-port-flashctl_blank_verify					10080 MHz, 1 run
g3519-port-flashctl_dynamic_memory_protection
g3519-port-flashctl_ecc_error_injection
g3519-port-flashctl_multiple_size_read_verify					10080 MHz, 1 run
g3519-port-flashctl_multiple_size_write					10080 MHz, 1 run
g3519-port-flashctl_nonmain_memory_write
g3519-port-flashctl_program_with_ecc					10080 MHz, 1 run
g3519-port-gpio_input_capture
g3519-port-gpio_software_poll
g3519-port-gpio_toggle_output_hiz					10080 MHz, 1 run
g3519-port-iwdt_periodic_reset					10080 MHz, 1 run
g3519-port-nvic_interrupt_disable					10080 MHz, 1 run
g3519-port-rtc_b_offset_calibration_lfxt
g3519-port-spi_controller_internal_loopback_poll					10080 MHz, 1 run
g3519-port-spi_controller_multibyte_fifo_poll
g3519-port-spi_controller_register_format
g3519-port-spi_peripheral_echo_interrupts
g3519-port-spi_peripheral_multibyte_fifo_poll
g3519-port-spi_peripheral_register_format
g3519-port-sram_parity
g3519-port-sysctl_frequency_clock_counter
g3519-port-sysctl_hfxt_run					Not for this chip
g3519-port-sysctl_lfxt_standby
g3519-port-sysctl_power_policy_sleep_to_standby
g3519-port-sysctl_power_policy_sleep_to_stop
g3519-port-sysctl_shutdown
g3519-port-systick_periodic_timer					10080 MHz, 1 run
g3519-port-tima_timer_mode_periodic_repeat_count					10080 MHz, 1 run
g3519-port-tima_timer_mode_pwm_dead_band
g3519-port-tima_trigger_fail_mechanism
g3519-port-timx_timer_mode_capture_duty_and_period
g3519-port-timx_timer_mode_capture_edge_capture
g3519-port-timx_timer_mode_compare_edge_count
g3519-port-timx_timer_mode_one_shot_standby
g3519-port-timx_timer_mode_periodic_sleep
g3519-port-timx_timer_mode_periodic_standby
g3519-port-timx_timer_mode_periodic_stop
g3519-port-timx_timer_mode_pwm_center_stop
g3519-port-timx_timer_mode_pwm_edge_sleep
g3519-port-timx_timer_mode_pwm_edge_sleep_shadow
g3519-port-timx_timer_mode_pwm_x_trig_stop_restore
g3519-port-uart_echo_interrupts_standby
g3519-port-uart_extend_irda_send_packet
g3519-port-uart_extend_manchester_echo
g3519-port-uart_extend_manchester_send_packet
g3519-port-uart_external_loopback_interrupt
g3519-port-uart_rw_multibyte_fifo_poll-rx
g3519-port-uart_rw_multibyte_fifo_poll-tx					10080 MHz, 1 run
g3519-port-wwdt_interval_timer_lfosc_standby
g3519-port-wwdt_interval_timer_lfxt_standby
g3519-port-wwdt_window_mode_periodic_reset
mcu-plus-sdk-26.01-lp															118 to 208400 MHz, 1 run
mcux-26.09.00									97.796 MHz, 1 run
mspm0-sdk-2.11	10032 MHz, 1 run	10032 MHz, 1 run	10024 MHz, 1 run	10080 MHz, 1 run	90.880 MHz, 1 run
mspm33-sdk-1.04											99.432 MHz, 1 run
mspm33-sdk-1.04-second-pass											10032 MHz, 1 run
pico-sdk-2.3.1
ported-g3507-adc-to-uart
ported-g3507-adc12-monitor-supply			10024 MHz, 1 run
ported-g3507-adc12-sequence-conversion			10024 MHz, 1 run
ported-g3507-adc12-single-conversion			10024 MHz, 1 run
ported-g3507-adc12-single-conversion-vref-external			10024 MHz, 1 run
ported-g3507-adc12-single-conversion-vref-internal			10024 MHz, 1 run
ported-g3507-adc12-window-comparator			10024 MHz, 1 run
ported-g3507-blinky-asm			024 MHz, 1 run
ported-g3507-cinit-bypass			10024 MHz, 1 run
ported-g3507-cmsis-dsp-empty			10024 MHz, 1 run
ported-g3507-comp-analog-filter			10024 MHz, 1 run
ported-g3507-dma-block-transfer			10024 MHz, 1 run
ported-g3507-driverlib-timg-qei-mode			10024 MHz, 1 run
ported-g3507-driverlib-uart-extend-irda-receive-packet
ported-g3507-driverlib-uart-extend-irda-send-packet
ported-g3507-driverlib-uart-extend-manchester-echo			10024 MHz, 1 run
ported-g3507-driverlib-uart-extend-manchester-send-packet			10024 MHz, 1 run
ported-g3507-driverlib-uart-external-loopback-interrupt			10024 MHz, 1 run
ported-g3507-driverlib-uart-rs485-receive-packet			10024 MHz, 1 run
ported-g3507-drivers-uart-echo
ported-g3507-eeprom-eeprom-emulation-type-a
ported-g3507-eeprom-eeprom-emulation-type-a-test
ported-g3507-eeprom-eeprom-emulation-type-b
ported-g3507-eeprom-eeprom-emulation-type-b-test
ported-g3507-eeprom-emulation-type-a
ported-g3507-eeprom-emulation-type-a-test
ported-g3507-eeprom-emulation-type-b
ported-g3507-eeprom-emulation-type-b-test
ported-g3507-empty			10024 MHz, 1 run
ported-g3507-empty-asm			10024 MHz, 1 run
ported-g3507-empty-cpp			10024 MHz, 1 run
ported-g3507-empty-driverlib-src			10024 MHz, 1 run
ported-g3507-empty-non-sysconfig			024 MHz, 1 run
ported-g3507-i2c-controller			10024 MHz, 1 run
ported-g3507-iqmath-iqmath-rts-ops-test			10024 MHz, 1 run
ported-g3507-iqmath-rts-ops-test
ported-g3507-mpesti-mpesti-target			10024 MHz, 1 run
ported-g3507-mpesti-target
ported-g3507-msp-subsystems-adc-to-uart			10024 MHz, 1 run
ported-g3507-msp-subsystems-pwm-led-driver			10024 MHz, 1 run
ported-g3507-msp-subsystems-uart-to-i2c-bridge			10024 MHz, 1 run
ported-g3507-opa-burnout-current-source-to-adc			10024 MHz, 1 run
ported-g3507-opa-general-purpose-rri			10024 MHz, 1 run
ported-g3507-opa-inverting-pga-with-dac			10024 MHz, 1 run
ported-g3507-opa-non-inverting-pga			10024 MHz, 1 run
ported-g3507-out-of-box			024 MHz, 1 run
ported-g3507-pmbus-pmbus-target			10024 MHz, 1 run
ported-g3507-pmbus-target
ported-g3507-pwm-led-driver
ported-g3507-smbus-controller
ported-g3507-smbus-controller-arp
ported-g3507-smbus-smbus-controller
ported-g3507-smbus-smbus-controller-arp
ported-g3507-smbus-smbus-target
ported-g3507-smbus-smbus-target-arp
ported-g3507-smbus-target
ported-g3507-smbus-target-arp
ported-g3507-sram-parity			024 MHz, 1 run
ported-g3507-sysctl-mclk-syspll			10024 MHz, 1 run
ported-g3507-tima-timer-mode-pwm-dead-band			024 MHz, 1 run
ported-g3507-timg-qei-mode
ported-g3507-uart-controller			10024 MHz, 1 run
ported-g3507-uart-echo
ported-g3507-uart-extend-irda-receive-packet
ported-g3507-uart-extend-irda-send-packet
ported-g3507-uart-extend-manchester-echo
ported-g3507-uart-extend-manchester-send-packet
ported-g3507-uart-external-loopback-interrupt
ported-g3507-uart-rs485-receive-packet
ported-g3507-uart-to-i2c-bridge
ported-l1306-battery-gauge-gauge-level1
ported-l1306-battery-gauge-self-calibrate-current-detection
ported-l1306-cmsis-dsp-cmsis-dsp-empty			10024 MHz, 1 run
ported-l1306-demos-blood-pressure-monitor
ported-l1306-demos-out-of-box
ported-l1306-driverlib-adc12-monitor-supply			10024 MHz, 1 run
ported-l1306-driverlib-adc12-sequence-conversion			10024 MHz, 1 run
ported-l1306-driverlib-adc12-single-conversion			10024 MHz, 1 run
ported-l1306-driverlib-adc12-single-conversion-vref-external			10024 MHz, 1 run
ported-l1306-driverlib-adc12-single-conversion-vref-internal			10024 MHz, 1 run
ported-l1306-driverlib-adc12-window-comparator			10024 MHz, 1 run
ported-l1306-driverlib-blinky-asm
ported-l1306-driverlib-cinit-bypass			10024 MHz, 1 run
ported-l1306-driverlib-comp-analog-filter			10024 MHz, 1 run
ported-l1306-driverlib-dma-block-transfer
ported-l1306-driverlib-empty			10024 MHz, 1 run
ported-l1306-driverlib-empty-asm
ported-l1306-driverlib-empty-cpp			10024 MHz, 1 run
ported-l1306-driverlib-empty-driverlib-src			10024 MHz, 1 run
ported-l1306-driverlib-empty-non-sysconfig
ported-l1306-driverlib-flashctl-blank-verify
ported-l1306-driverlib-flashctl-dynamic-memory-protection
ported-l1306-driverlib-flashctl-multiple-size-read-verify
ported-l1306-driverlib-flashctl-multiple-size-write
ported-l1306-driverlib-gpio-toggle-output-hiz			10024 MHz, 1 run
ported-l1306-driverlib-i2c-controller-rw-multibyte-fifo-interrupts
ported-l1306-driverlib-i2c-controller-rw-multibyte-fifo-poll
ported-l1306-driverlib-i2c-multicontroller-arbitration
ported-l1306-driverlib-i2c-target-rw-multibyte-fifo-interrupts			10024 MHz, 1 run
ported-l1306-driverlib-i2c-target-rw-multibyte-fifo-interrupts-stop
ported-l1306-driverlib-i2c-target-rw-multibyte-fifo-poll
ported-l1306-driverlib-opa-burnout-current-source-to-adc			10024 MHz, 1 run
ported-l1306-driverlib-opa-general-purpose-rri			10024 MHz, 1 run
ported-l1306-driverlib-opa-non-inverting-pga			10024 MHz, 1 run
ported-l1306-driverlib-spi-controller-command-data-control			10024 MHz, 1 run
ported-l1306-driverlib-spi-controller-echo-interrupts			10024 MHz, 1 run
ported-l1306-driverlib-spi-controller-fifo-dma-interrupts
ported-l1306-driverlib-spi-controller-internal-loopback-poll			Not for this chip
ported-l1306-driverlib-spi-controller-multibyte-fifo-poll
ported-l1306-driverlib-spi-controller-register-format
ported-l1306-driverlib-spi-peripheral-echo-interrupts
ported-l1306-driverlib-spi-peripheral-fifo-dma-interrupts
ported-l1306-driverlib-spi-peripheral-multibyte-fifo-poll
ported-l1306-driverlib-spi-peripheral-register-format
ported-l1306-driverlib-spi-peripheral-repeated-fifo-dma-interrupts
ported-l1306-driverlib-sysctl-frequency-clock-counter
ported-l1306-driverlib-systick-periodic-timer
ported-l1306-driverlib-uart-echo-interrupts-standby
ported-l1306-driverlib-uart-extend-manchester-echo
ported-l1306-driverlib-uart-extend-manchester-send-packet
ported-l1306-driverlib-uart-external-loopback-interrupt
ported-l1306-driverlib-uart-rw-multibyte-fifo-poll
ported-l1306-driverlib-uart-rx-hw-flow-control
ported-l1306-driverlib-uart-rx-multibyte-fifo-dma-interrupts
ported-l1306-driverlib-uart-tx-hw-flow-control
ported-l1306-driverlib-uart-tx-multibyte-fifo-dma-interrupts
ported-l1306-driverlib-wwdt-interval-timer-lfosc-standby
ported-l1306-eeprom-eeprom-emulation-type-a
ported-l1306-eeprom-eeprom-emulation-type-a-test
ported-l1306-eeprom-eeprom-emulation-type-b
ported-l1306-eeprom-eeprom-emulation-type-b-test
ported-l1306-iqmath-iqmath-rts-ops-test
ported-l1306-msp-subsystems-adc-to-i2c-target
ported-l1306-msp-subsystems-adc-to-spi-peripheral
ported-l1306-msp-subsystems-adc-to-uart
ported-l1306-msp-subsystems-digital-mux
ported-l1306-msp-subsystems-io-expander-i2c-host
ported-l1306-msp-subsystems-io-expander-uart-i2c-spi
ported-l1306-msp-subsystems-transimpedance-amplifier
ported-l1306-msp-subsystems-uart-to-i2c-bridge
ported-l1306-sasi-smoke-detector-air-quality-sensing			Not for this chip
ported-l1306-sasi-smoke-detector-sasi-smoke-detector			Not for this chip
ported-l1306-smbus-smbus-controller
ported-l1306-smbus-smbus-controller-arp
ported-l1306-smbus-smbus-target
ported-l1306-smbus-smbus-target-arp			Not for this chip
r5-twins-behavior-checked													66.796 MHz, 1 run
r5-twins-f3-9.21													10096 MHz, 1 run
sdk-f3-9.21													98.81 run
stm32cubeu0-1.3.0								10048 MHz, 1 run
wifi-sdk-10.20.00.39														93.5160 MHz, 1 run
(one way)

Sources

data/benchmark-suite.json, docs/experiments/2026-10-24-cc2340r5-cc35x1-driver-twins.md

Examples that build after moving to another chip %, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A
cc1407p-host-c28x-no-shim								01 run
cc1407p-host-c28x-shim								1001 run
from-cc1354p10-to-cc2340r5-unchanged
from-f280015x-device-only								76.71 run
from-f280015x-pin-unlock								79.11 run
from-mspm0c1104-way3				81.31 run
from-mspm0g3507-agent
from-mspm0g3507-shim							501 run
from-mspm0g3507-way3				68.51 run	31.11 run
from-mspm0g3519-driverlib-fresh-config					64.81 run
from-mspm0g3519-way3					54.71 run
mspm0g3519-to-mspm33-config-port-12									66.71 run
mspm33-to-mspm0g3519-config-port-12				66.71 run
to-cc1354p10-unchanged						1001 run
to-cc2340r5-shim	1001 run
to-cc2340r5-unchanged	01 run
to-cc2755p10-unchanged						1001 run
to-mspm0c1104-device-only	4.81 run	17.61 run
to-mspm0c1104-lines-dropped	17.81 run	44.71 run
to-mspm0c1104-pin-unlock	5.91 run	36.51 run
to-mspm0c1104-way3				34.11 run
to-mspm0g3507-device-only		59.41 run	48.41 run
to-mspm0g3507-lines-dropped		79.41 run	71.41 run
to-mspm0g3507-pin-unlock		77.61 run	71.41 run
to-mspm0g3507-shim						1001 run
to-mspm0g3507-unchanged						501 run
to-mspm0g3507-way3				77.51 run	431 run
to-mspm0g3519-shim				10080 MHz, 1 run
to-mspm0g3519-way3					48.81 run
to-mspm0l1306-device-only	8.11 run		441 run
to-mspm0l1306-lines-dropped	35.61 run		62.61 run
to-mspm0l1306-pin-unlock	9.61 run		62.61 run
(one way)

Sources

portability/README.md, portability/matrix.csv

A TI driver example from another chip's SDK shows the behavior its source defines yes/no

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LP-CC1354P10-1	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	TMDS62LEVM (AM62L EVM)
arduino-zephyr-analogwrite-duty			Not for this chip
arduino-zephyr-analogwrite-leds					148 MHz, 1 run
arduino-zephyr-analogwrite-pwm-output			180 MHz
arduino-zephyr-attachinterrupt-first-mode			180 MHz
arduino-zephyr-attachinterrupt-mode-change			Not for this chip
arduino-zephyr-attachinterrupt-pinmode-flip					Not for this chip
arduino-zephyr-attachinterrupt-pull-flip					148 MHz, 1 run
arduino-zephyr-spi-loopback
arduino-zephyr-spi-transfer					148 MHz, 1 run
arduino-zephyr-spi-transfer-buffer			Not for this chip
arduino-zephyr-spi-transfer-loopback			180 MHz
arduino-zephyr-spi-wire-loopback
arduino-zephyr-wire-ft6336u			180 MHz, 1 run
arduino-zephyr-wire-hdc2080					Not for this chip
arduino-zephyr-wire-opt3001					148 MHz, 1 run
arduino-zephyr-wire-scan-empty					148 MHz, 1 run
arduino-zephyr-wire-scan-pack					Not for this chip
cc2340r5-to-cc1354p10-adcsinglechannel						148 MHz, 1 run
cc2340r5-to-cc1354p10-empty
cc2340r5-to-cc1354p10-gpiointerrupt
cc2340r5-to-cc1354p10-i2ctmp						148 MHz, 1 run
cc2340r5-to-cc1354p10-spicontroller						148 MHz, 1 run
cc2340r5-to-cc1354p10-spiperipheral						148 MHz, 1 run
cc2340r5-to-cc1354p10-uart2callback						148 MHz, 1 run
cc2340r5-to-cc1354p10-uart2echo						148 MHz, 1 run
doomgeneric-fbdev-4x3-aspect-by-eye												11250 MHz, 1 run
doomgeneric-fbdev-pad-buttons-console-layout												11250 MHz, 1 run
doomgeneric-fbdev-pad-stick-up-walks												11250 MHz, 1 run
driverlib	132 MHz, 1 run	124 MHz, 1 run	132 MHz, 1 run	132 MHz, 1 run
driverlib-am13									1200 MHz, 1 run
driverlib-am13-adc-pa16									Not for this chip
driverlib-hsadc								132 MHz, 1 run
driverlib-pa25-floating			132 MHz, 1 run
from-cc35x1-adcsinglechannel					yes1 run
from-cc35x1-empty					yes1 run
from-cc35x1-exception					yes1 run
from-cc35x1-gpiointerrupt					yes1 run
from-cc35x1-i2ccontroller					yes1 run
from-cc35x1-pwmled1
from-cc35x1-sha2hash					yes1 run
from-cc35x1-spicontroller					yes1 run
from-cc35x1-spiperipheral					yes1 run
from-cc35x1-uart2echo					yes1 run
from-cc35x1-watchdog					yes1 run
from-mspm0g3519-adc12_single_conversion								Not for this chip
from-mspm0g3519-crcp_calculate_checksum								yes1 run
from-mspm0g3519-gpio_toggle_output								yes1 run
from-mspm0g3519-i2c_controller_rw_multibyte_fifo_poll								Not for this chip
from-mspm0g3519-spi_controller_internal_loopback_poll								yes1 run
from-mspm0g3519-systick_periodic_timer								yes1 run
from-mspm0g3519-timx_timer_mode_periodic_sleep								yes1 run
from-mspm0g3519-timx_timer_mode_pwm_edge_sleep								yes1 run
from-mspm0g3519-uart_external_loopback_interrupt
from-mspm0g3519-uart_internal_loopback_standby_restore								Not for this chip
from-mspm0g3519-uart_rw_multibyte_fifo_poll								yes1 run
from-mspm0g3519-uart_tx_console_multibyte_repeated_fifo_dma								Not for this chip
from-mspm33-crcp_calculate_checksum			yes1 run
from-mspm33-gpio_toggle_output			yes1 run
from-mspm33-hsadc_single_conversion			Not for this chip
from-mspm33-i2c_controller_rw_byte_poll			Not for this chip
from-mspm33-spi_controller_internal_loopback_poll			yes1 run
from-mspm33-systick_periodic_timer			yes1 run
from-mspm33-timx_timer_mode_periodic_sleep			yes1 run
from-mspm33-timx_timer_mode_pwm_edge_sleep			yes1 run
from-mspm33-uart_internal_loopback_spgss0_uc1			yes1 run
from-mspm33-uart_internal_loopback_with_interrupt_spgss0_uc1			yes1 run
from-mspm33-uart_rx_rw_multibyte_fifo_poll			Not for this chip
from-mspm33-uart_tx_rw_multibyte_fifo_poll			yes1 run
micropython-zephyr-adc								Not for this chip
micropython-zephyr-i2c-hdc2080					Not for this chip
micropython-zephyr-i2c-opt3001					148 MHz, 1 run
micropython-zephyr-i2c-read								Not for this chip
micropython-zephyr-i2c-scan-empty					148 MHz, 1 run
micropython-zephyr-i2c-scan-pack					Not for this chip
micropython-zephyr-i2c-write								1160 MHz, 1 run
micropython-zephyr-loop-10000								1160 MHz, 1 run
micropython-zephyr-pin-irq-init-flip					Not for this chip
micropython-zephyr-pin-irq-pull-flip					148 MHz, 1 run
micropython-zephyr-pin-pull-up-down								1160 MHz, 1 run
micropython-zephyr-pin-value-readback								Not for this chip
micropython-zephyr-pwm								Not for this chip
micropython-zephyr-pwm-duty-ns					148 MHz, 1 run
micropython-zephyr-pwm-duty-u16					Not for this chip
micropython-zephyr-pwm-duty-u16-patched					148 MHz, 1 run
micropython-zephyr-spi-i2c-together
micropython-zephyr-spi-loopback
micropython-zephyr-spi-write-readinto					148 MHz, 1 run			1160 MHz, 1 run
pico-blink_simple			yes32 MHz
pico-bus_scan			yes32 MHz
pico-hello_adc			yes32 MHz, 1 run
pico-hello_serial			yes32 MHz, 1 run
pico-led_fade			yes32 MHz, 1 run
renesas-adc			yes32 MHz, 1 run
renesas-adc_gpt_periodic_sampling			no32 MHz, 1 run
renesas-adc_gpt_periodic_sampling-dtc-chain			Not for this chip
renesas-adc_gpt_periodic_sampling-event-link			yes32 MHz, 1 run
renesas-adc_gpt_periodic_sampling-group-b-trigger			Not for this chip
renesas-adc_gpt_periodic_sampling-one-channel			yes32 MHz, 1 run
renesas-gpt			yes32 MHz, 1 run
renesas-gpt-one-shot			yes32 MHz
renesas-gpt-periodic-period			yes32 MHz
renesas-gpt-pwm-duty			yes32 MHz, 1 run
renesas-iic_master-ft6336u			yes32 MHz
renesas-iic_master-unchanged			no32 MHz, 1 run
renesas-sci_uart			yes32 MHz, 1 run
renesas-sci_uart-pwm-duty			yes32 MHz, 1 run
silent-twins-uart
three-ways-app-c2000ware							yes120 MHz, 1 run
three-ways-app-linux-sysfs												11250 MHz, 1 run
three-ways-app-tidrivers					148 MHz, 1 run					196 MHz, 1 run	1160 MHz, 1 run
twin-f3-adcsinglechannel			Not for this chip
twin-f3-empty			yes32 MHz, 1 run
twin-f3-gpiointerrupt
twin-f3-i2ccontroller
twin-f3-i2ctarget
twin-f3-i2ctmp
twin-f3-spicontroller
twin-f3-spiperipheral
twin-f3-uart2callback			yes32 MHz, 1 run
twin-f3-uart2echo			yes32 MHz, 1 run
x47-wire-packer							yes120 MHz, 1 run
zephyr-board-pinctrl-tima1-ccp1-pb27			Not for this chip
zephyr-mspm0-i2c-hdc2080

Sources

docs/experiments/2026-10-24-cc2340r5-cc35x1-driver-twins.md, portability/arduino/zephyr/README.md, portability/cc1407p-host/wire/README.md, portability/three-ways/README.md, portability/zephyr-mspm0/README.md, ports/doom/linux-fb/README.md, ports/micropython/zephyr-cc2340r5/README.md

Glue lines the portable three-ways app needs on one board and software layer lines, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	TMDS62LEVM (AM62L EVM)
c2000ware-driverlib						35120 MHz, 1 run
driverlib	3932 MHz, 1 run	3924 MHz, 1 run	3932 MHz, 1 run	3932 MHz, 1 run
driverlib-am13								38200 MHz, 1 run
driverlib-hsadc							3732 MHz, 1 run
linux-sysfs											661250 MHz, 1 run
three-ways-app-tidrivers					481 run
tidrivers-f3-sdk									481 run
tidrivers-wifi-sdk										481 run

Sources

portability/three-ways/README.md, portability/three-ways/c2000/glue_c2000.c, portability/three-ways/linux/glue_linux.c, portability/three-ways/tidrivers/cc2755p10/glue_tidrivers_cc2755p10.c, portability/three-ways/tidrivers/cc35x1/glue_tidrivers_cc35x1.c

Configuration lines the portable three-ways app needs on one board and software layer lines, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	TMDS62LEVM (AM62L EVM)
c2000ware-driverlib						29120 MHz, 1 run
driverlib	3432 MHz, 1 run	3024 MHz, 1 run	3432 MHz, 1 run	3332 MHz, 1 run
driverlib-am13								35200 MHz, 1 run
driverlib-hsadc							4132 MHz, 1 run
linux-sysfs											51250 MHz, 1 run
three-ways-app-tidrivers					141 run
tidrivers-f3-sdk									141 run
tidrivers-wifi-sdk										141 run

Sources

portability/three-ways/README.md, portability/three-ways/c2000/glue_c2000.c, portability/three-ways/driverlib/g3519.syscfg, portability/three-ways/linux/config_am62l.h, portability/three-ways/tidrivers/cc2340r5/app.syscfg, portability/three-ways/tidrivers/cc2755p10/app.syscfg, portability/three-ways/tidrivers/cc35x1/app.syscfg

Flash used by an outside project ported to the board, with its shim B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1	TMDS62LEVM (AM62L EVM)
arduboy-crate-confusion-edumkii								20 kB32 MHz, 1 run
arduboy-snake-edumkii								20.2 kB32 MHz, 1 run
arduboy-tileworld-edumkii								16.4 kB32 MHz, 1 run
arduboy2-crate-confusion	9.4 kB1 run			9.4 kB1 run	9.4 kB1 run			9.6 kB1 run
arduboy2-snake	7.7 kB1 run			7.7 kB1 run	7.7 kB1 run			8.2 kB1 run
arduboy2-tileworld	5.5 kB1 run			5.5 kB1 run	5.5 kB1 run			5.8 kB1 run
arduino-layer-analogreadserial	Not for this chip
arduino-layer-blink	2.6 kB32 MHz, 1 run
arduino-zephyr-analogreadserial				22.8 kB80 MHz, 1 run		27.8 kB48 MHz, 1 run
arduino-zephyr-blink				22.8 kB80 MHz, 1 run		27.7 kB48 MHz, 1 run
arduino-zephyr-buttoninterrupt				31.7 kB80 MHz, 1 run
arduino-zephyr-fadecheck						32.9 kB48 MHz, 1 run
arduino-zephyr-fadepwm				32.1 kB80 MHz, 1 run
arduino-zephyr-interruptselftest						32.9 kB48 MHz, 1 run
arduino-zephyr-spiloop						32 kB48 MHz, 1 run
arduino-zephyr-spiloopback				31.5 kB80 MHz, 1 run
arduino-zephyr-wire-ft6336u				31.1 kB80 MHz, 1 run
arduino-zephyr-wirescan						37.4 kB48 MHz, 1 run
cc1407p-host-fake-radio-bench				10.6 kB1 run				10.8 kB1 run
chip8-bench				14.1 kB80 MHz, 1 run
chip8-play				13.5 kB80 MHz, 1 run
doom-m33whx								230.8 kB160 MHz, 1 run
doom-m33whx-edumkii-c8								298.6 kB160 MHz, 1 run
doom-m33whx-page-cache								304.1 kB160 MHz, 1 run
doom-s3	Not for this chip 134 kB short			262 kB calc.				281.6 kB160 MHz, 1 run
doom-s3-wad-xip											4.3 MB160 MHz, 1 run
doom-whx-panel-xip											1.99 MB160 MHz, 1 run
doom-whx-xip											1.97 MB160 MHz, 1 run
driverlib		2.3 kB32 MHz, 1 run	2.3 kB24 MHz, 1 run	2.3 kB32 MHz, 1 run	2.4 kB32 MHz, 1 run
driverlib-am13									9.2 kB200 MHz, 1 run
driverlib-hsadc								2.8 kB32 MHz, 1 run
llama2c-on-edumkii-screen								554 kB160 MHz, 1 run
llama2c-tinystories-260k-float32								Not for this chip 7.8 kB short
llama2c-tinystories-260k-int8								543.6 kB1 run
micropython-full	Not for this chip 45.9 kB short			174.1 kB1 run
micropython-full-spi				175.6 kB1 run
micropython-min	72.9 kB1 run			73.1 kB1 run
micropython-min-heap16k				73.1 kB1 run
micropython-zephyr-basic-heap20k						174.8 kB48 MHz, 1 run
micropython-zephyr-basic-heap20k-pwmfix						174.9 kB48 MHz, 1 run
micropython-zephyr-extra-heap16k						213.6 kB48 MHz, 1 run
micropython-zephyr-stock-prj-gpio								292.3 kB160 MHz, 1 run
micropython-zephyr-stock-prj-i2c								294.5 kB160 MHz, 1 run
micropython-zephyr-stock-prj-spi								294.4 kB160 MHz, 1 run
minimp3-bench				146.8 kB1 run				153.4 kB160 MHz, 1 run
minmea	5.7 kB1 run			5.6 kB1 run
moonlight-embedded-f32e415												157.8 kB1250 MHz, 1 run
msp430-fram-persist		4.5 kB32 MHz, 1 run							12.6 kB200 MHz, 1 run
msp430-lpm3-wake		2 kB32 MHz, 1 run							8.5 kB200 MHz, 1 run
msp430-uart-echo		1.2 kB32 MHz, 1 run							7.6 kB200 MHz, 1 run
peanut-gb-code	12.5 kB1 run			12.5 kB1 run	12.5 kB1 run			13 kB1 run
peanut-gb-dmg-acid2	44.5 kB1 run			44.5 kB1 run	44.5 kB1 run			45 kB1 run
peanut-gb-tobu	Not for this chip 140.4 kB short			268.5 kB1 run	Not for this chip 140.4 kB short			269 kB1 run
peanut-gb-tobu-edumkii								281.1 kB32 MHz, 1 run
peanut-gb-tobu-edumkii-160mhz-skip								284.6 kB160 MHz, 1 run
pico-blink_simple				2.4 kB32 MHz, 1 run
pico-bus_scan				10.6 kB32 MHz, 1 run
pico-hello_adc				10.4 kB32 MHz, 1 run
pico-hello_serial				1.8 kB32 MHz, 1 run
pico-led_fade				3.3 kB32 MHz, 1 run
renesas-adc				13.3 kB32 MHz, 1 run
renesas-adc_gpt_periodic_sampling				15.5 kB32 MHz, 1 run
renesas-gpt				14.9 kB32 MHz, 1 run
renesas-iic_master-ft6336u				12.6 kB32 MHz, 1 run
renesas-sci_uart				13.1 kB32 MHz, 1 run
rust-uart-hello		3.7 kB32 MHz, 1 run
sharp128-hello	12.2 kB32 MHz, 1 run	4.3 kB32 MHz, 1 run		12.2 kB32 MHz, 1 run				29.4 kB160 MHz, 1 run
sharp128-hello-bands			4.8 kB24 MHz, 1 run
sharp128-player-pff			7.2 kB24 MHz, 1 run
three-ways-app-c2000ware							4.9 kB120 MHz, 1 run
three-ways-app-linux-dynamic-stripped												66.1 kB1250 MHz, 1 run
three-ways-app-linux-static-stripped												583.9 kB1250 MHz, 1 run
three-ways-app-tidrivers						18.4 kB48 MHz, 1 run				19.7 kB96 MHz, 1 run	20.7 kB160 MHz, 1 run
touch-games-bench	10.8 kB1 run			11 kB80 MHz				8.1 kB1 run
touch-games-play				9.7 kB80 MHz, 1 run
ulisp-4.9a	98.8 kB80 MHz, 1 run			99 kB80 MHz, 1 run	98.8 kB80 MHz, 1 run
wasm3-v0.9.0				55.6 kB1 run				76.4 kB1 run
wasm3-v0.9.0-tiny	55.4 kB1 run
x47-wire-packer							1.3 kB120 MHz, 1 run
xiaozhi											1.5 MB160 MHz, 1 run
zx48-sram-testimage				62.5 kB80 MHz, 1 run

Sources

docs/experiments/2026-10-20-doom-smallest-ti-chip.md, portability/arduino/zephyr/README.md, portability/boosterpacks/edumkii/build.sh, portability/boosterpacks/sharp128/README.md, portability/cc1407p-host/wire/README.md, portability/three-ways/README.md, ports/chip8/README.md, ports/doom/target-cc35x1-whx/PANEL-WIRING.md, ports/llama2c/README.md, ports/micropython/zephyr-cc2340r5/README.md, ports/minimp3/README.md, ports/msp430-to-mspm0/fram-persist/PORT.md, ports/msp430-to-mspm0/lpm3-wake/PORT.md, ports/msp430-to-mspm0/uart-echo-lpm3/PORT.md, ports/rust-mspm0/README.md, ports/touch-games/README.md, ports/zx-spectrum/README.md

RAM used by an outside project ported to the board, with its shim B, lower is better

Variant	LP-MSPM0G3507	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	LP-EM-CC2340R5	LAUNCHXL-F2800137	LP-MSPM33C321A	LP-AM13E230	LP-EM-CC2755P10	LP-EM-CC35X1
arduboy-crate-confusion-edumkii								6.3 kB32 MHz, 1 run
arduboy-snake-edumkii								6.6 kB32 MHz, 1 run
arduboy-tileworld-edumkii								6.1 kB32 MHz, 1 run
arduboy2-crate-confusion	3.4 kB1 run			3.4 kB1 run	3.4 kB1 run			3.4 kB1 run
arduboy2-snake	3.8 kB1 run			3.8 kB1 run	3.8 kB1 run			3.8 kB1 run
arduboy2-tileworld	3.3 kB1 run			3.3 kB1 run	3.3 kB1 run			3.3 kB1 run
arduino-layer-blink	6.2 kB32 MHz, 1 run
arduino-zephyr-analogreadserial				4.8 kB80 MHz, 1 run		5.2 kB48 MHz, 1 run
arduino-zephyr-blink				4.8 kB80 MHz, 1 run		5.2 kB48 MHz, 1 run
arduino-zephyr-buttoninterrupt				6.3 kB80 MHz, 1 run
arduino-zephyr-fadecheck						6.1 kB48 MHz, 1 run
arduino-zephyr-fadepwm				6.4 kB80 MHz, 1 run
arduino-zephyr-interruptselftest						6.1 kB48 MHz, 1 run
arduino-zephyr-spiloop						6.1 kB48 MHz, 1 run
arduino-zephyr-spiloopback				6.3 kB80 MHz, 1 run
arduino-zephyr-wire-ft6336u				6.3 kB80 MHz, 1 run
arduino-zephyr-wirescan						6.1 kB48 MHz, 1 run
cc1407p-host-fake-radio-bench				5.8 kB1 run				5.8 kB1 run
chip8-play				10.4 kB80 MHz, 1 run
doom-m33whx	Not for this chip 166 kB short			Not for this chip 70 kB short
doom-m33whx-board-page-cache-16kb								253.3 kB160 MHz, 1 run
doom-m33whx-board-page-cache-8kb								245.3 kB160 MHz, 1 run
doom-m33whx-edumkii-c8								247.7 kB160 MHz, 1 run
doom-m33whx-linked-44kb-zone								242 kB160 MHz, 1 run
doom-m33whx-page-cache-12kb								248.6 kB160 MHz, 1 run
doom-m33whx-page-cache-16kb								252.7 kB160 MHz, 1 run
doom-m33whx-page-cache-19kb								255.7 kB160 MHz, 1 run
doom-m33whx-page-cache-8kb								244.6 kB160 MHz, 1 run
doom-m33whx-static								198 kB160 MHz, 1 run
doom-m33whx-total								228 kB calc.
doom-m33whx-zone-finishes								30 kB160 MHz, 1 run
doom-m33whx-zone-hw								28.9 kB160 MHz, 1 run
doom-s3	Not for this chip 132.1 kB short			Not for this chip 36.1 kB short				Not for this chip 707.8 kB short
doom-s3-linked-80kb-zone								243.8 kB160 MHz, 1 run
doom-s3-static								163.8 kB160 MHz, 1 run
doom-s3-static-and-rtos											187.1 kB160 MHz, 1 run
doom-s3-zone											480 kB160 MHz, 1 run
doom-s3-zone-finishes								800 kB160 MHz, 1 run
doom-s3-zone-floor								593 kB160 MHz, 1 run
doom-whx-panel-static-and-zone											494.2 kB160 MHz, 1 run
doom-whx-panel-zone											31 kB160 MHz, 1 run
doom-whx-static-and-rtos											220.2 kB160 MHz, 1 run
doom-whx-zone											40 kB160 MHz, 1 run
doom-whx-zone-hw											28.9 kB160 MHz, 1 run
driverlib		27232 MHz, 1 run	8024 MHz, 1 run	27232 MHz, 1 run	27232 MHz, 1 run
driverlib-am13									9 kB200 MHz, 1 run
driverlib-hsadc								27232 MHz, 1 run
driverlib-stack-highwater			18424 MHz, 1 run
llama2c-on-edumkii-screen								192.8 kB160 MHz, 1 run
llama2c-tinystories-260k-int8								192.6 kB1 run
micropython-full				5.7 kB1 run
micropython-min	2.9 kB1 run			3 kB1 run
micropython-min-heap16k				19 kB1 run
micropython-zephyr-basic-heap20k						29.9 kB48 MHz, 1 run
micropython-zephyr-basic-heap20k-pwmfix						29.9 kB48 MHz, 1 run
micropython-zephyr-extra-heap16k						26 kB48 MHz, 1 run
micropython-zephyr-stock-prj						Not for this chip 46.8 kB short
micropython-zephyr-stock-prj-gpio								83.6 kB160 MHz, 1 run
micropython-zephyr-stock-prj-i2c								83.6 kB160 MHz, 1 run
micropython-zephyr-stock-prj-spi								83.7 kB160 MHz, 1 run
minimp3-bench				33.3 kB1 run				33.2 kB160 MHz, 1 run
minmea	6281 run			3721 run
msp430-fram-persist		36532 MHz, 1 run
msp430-lpm3-wake		26032 MHz, 1 run
msp430-uart-echo		26832 MHz, 1 run
peanut-gb-code	30.8 kB1 run			30.8 kB1 run	30.8 kB1 run			30.8 kB1 run
peanut-gb-dmg-acid2	30.8 kB1 run			30.8 kB1 run	30.8 kB1 run			30.8 kB1 run
peanut-gb-tobu				30.8 kB1 run				30.8 kB1 run
peanut-gb-tobu-edumkii								31.6 kB32 MHz, 1 run
peanut-gb-tobu-edumkii-160mhz-skip								32.8 kB160 MHz, 1 run
pico-blink_simple				2.1 kB32 MHz, 1 run
pico-bus_scan				2.4 kB32 MHz, 1 run
pico-hello_adc				2.4 kB32 MHz, 1 run
pico-hello_serial				2.1 kB32 MHz, 1 run
pico-led_fade				2.1 kB32 MHz, 1 run
renesas-adc				53532 MHz, 1 run
renesas-adc_gpt_periodic_sampling				4.7 kB32 MHz, 1 run
renesas-gpt				58132 MHz, 1 run
renesas-iic_master-ft6336u				51632 MHz, 1 run
renesas-sci_uart				56732 MHz, 1 run
rust-uart-hello		14832 MHz, 1 run
sharp128-hello	5.5 kB32 MHz, 1 run	3.1 kB32 MHz, 1 run		5.5 kB32 MHz, 1 run				3.7 kB160 MHz, 1 run
sharp128-hello-bands			60924 MHz, 1 run
sharp128-player-pff			77924 MHz, 1 run
three-ways-app-c2000ware							1 kB120 MHz, 1 run
three-ways-app-tidrivers						13.6 kB48 MHz, 1 run				37.8 kB96 MHz, 1 run	43.4 kB160 MHz, 1 run
touch-games-bench	3.3 kB1 run			3.4 kB80 MHz				3.3 kB1 run
touch-games-play				3.4 kB80 MHz, 1 run
ulisp-4.9a	23.1 kB80 MHz, 1 run			71.6 kB80 MHz, 1 run	23.1 kB80 MHz, 1 run
wasm3-v0.9.0				24.2 kB1 run				24.5 kB1 run
wasm3-v0.9.0-standard	24.1 kB1 run
wasm3-v0.9.0-tiny	11.1 kB1 run
xiaozhi											189.6 kB160 MHz, 1 run
zx48-sram-testimage				61 kB80 MHz, 1 run

Sources

docs/experiments/2026-10-20-doom-smallest-ti-chip.md, docs/experiments/2026-10-21-mspm33c321a-qspi-dump-doom-data.md, portability/arduino/zephyr/README.md, portability/boosterpacks/edumkii/build.sh, portability/boosterpacks/sharp128/README.md, portability/three-ways/README.md, ports/chip8/README.md, ports/doom/target-cc35x1-whx/PANEL-WIRING.md, ports/llama2c/README.md, ports/micropython/zephyr-cc2340r5/README.md, ports/minimp3/README.md, ports/msp430-to-mspm0/fram-persist/PORT.md, ports/msp430-to-mspm0/lpm3-wake/PORT.md, ports/msp430-to-mspm0/uart-echo-lpm3/PORT.md, ports/rust-mspm0/README.md, ports/touch-games/README.md, ports/zx-spectrum/README.md

SDK examples built with GCC that ran on the board with no fault examples, higher is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187
gcc-12.2-zephyr-sdk	8732 MHz, 1 run	3180 MHz, 1 run	6580 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

GCC-built SDK examples that faulted on the board examples, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187
gcc-12.2-zephyr-sdk	032 MHz, 1 run	080 MHz, 1 run	080 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

GCC-built SDK examples whose pin or serial behavior matched the ticlang build examples, higher is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519	LP-MSPM0G5187
gcc-12.2-zephyr-sdk	2432 MHz, 1 run	580 MHz, 1 run	1580 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Flash used by the ported UART DMA ring-buffer echo (MaJerle stm32-usart-uart-dma-rx-tx) B, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230
gcc-14.3-os					4.7 kB48 MHz, 1 run
ticlang-5.1.1-o2	2.9 kB32 MHz, 1 run	2.2 kB32 MHz, 1 run	3 kB32 MHz, 1 run	2.5 kB32 MHz, 1 run		2.9 kB32 MHz, 1 run	9.6 kB1 run
ticlang-5.1.1-o2-run3			2.6 kB32 MHz, 1 run
ticlang-5.1.1-o2-run4			2.5 kB32 MHz, 1 run
ticlang-5.1.1-o2-skill			2.6 kB32 MHz, 1 run
(one way)

Sources

Run records are in the notebook, not this copy.

Static RAM (data + bss) of the ported UART DMA ring-buffer echo B, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230
gcc-14.3-os					2.3 kB48 MHz, 1 run
ticlang-5.1.1-o2	63632 MHz, 1 run	63232 MHz, 1 run	63632 MHz, 1 run	63232 MHz, 1 run		63232 MHz, 1 run	9.3 kB1 run
ticlang-5.1.1-o2-run3			63232 MHz, 1 run
ticlang-5.1.1-o2-run4			63232 MHz, 1 run
ticlang-5.1.1-o2-skill			63232 MHz, 1 run
(one way)

Sources

Run records are in the notebook, not this copy.

Packets per second answered for 500 back-to-back 5-byte packets, console UART pps, higher is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230
baud-115200	2,248.532 MHz	2,293.832 MHz	2,245.532 MHz	2,284.432 MHz	2,259.448 MHz	2,28432 MHz	2,273.7
baud-115200-run3			2,279.332 MHz
baud-115200-run4			2,291.332 MHz
baud-115200-skill			2,281.132 MHz
baud-921600	9,131.132 MHz	979.332 MHz	9,142.832 MHz	9,012.832 MHz	15,252.848 MHz	6,285.232 MHz	9,194.7
baud-921600-skill			5,938.332 MHz
(one way)

Sources

Run records are in the notebook, not this copy.

Counting lines read on the host after the banner from TI's UART console example count, higher is better

Variant	LP-MSPM0G3519
uart_tx_console_multibyte_repeated_fifo_dma	632 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Replies lost out of 500 back-to-back packets count, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230
baud-115200	032 MHz	032 MHz	032 MHz	032 MHz	048 MHz	032 MHz	0
baud-115200-run3			032 MHz
baud-115200-run4			032 MHz
baud-115200-skill			032 MHz
baud-921600	032 MHz	447 to 45232 MHz	032 MHz	032 MHz	048 MHz	152 to 18132 MHz	0
baud-921600-skill			164 to 23832 MHz
(one way)

Sources

Run records are in the notebook, not this copy.

Median host-measured round trip of one packet to its 4-byte reply ms, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0C1104	LP-MSPM0G3519	LP-MSPM0G5187	NUCLEO-U083RC	LP-MSPM33C321A	LP-AM13E230
baud-115200	2.8532 MHz	1.5332 MHz	2.8632 MHz	1.8532 MHz	1.7248 MHz	1.832 MHz	1.41
baud-921600	28.532 MHz	28.932 MHz	28.732 MHz	28.532 MHz	0.9248 MHz	28.632 MHz	28.4
(one way)

Sources

Run records are in the notebook, not this copy.

Flash used by the ported BLE gamepad report core (lemmingDev ESP32-BLE-Gamepad, generic mode) and its console glue B, lower is better

Variant	LP-MSPM0G3519	LP-EM-CC2340R5	LAUNCHXL-F2800137	NUCLEO-U083RC	ESP32-C6-DevKitC-1-N8
arduino-nimble-library					702.1 kB160 MHz, 1 run
core-arduino					270.9 kB160 MHz, 1 run
core-cl2000-o2-ram			8.1 kB120 MHz, 1 run
core-gcc-os				3.6 kB16 MHz, 1 run
core-ticlang	9.9 kB32 MHz, 1 run
core-ticlang-ble-hogp-build-only		189.4 kB to 189.4 kB48 MHz

Sources

Run records are in the notebook, not this copy.

Static RAM used by the same image (bss plus data, stack not counted) B, lower is better

Variant	LP-MSPM0G3519	LP-EM-CC2340R5	LAUNCHXL-F2800137	NUCLEO-U083RC	ESP32-C6-DevKitC-1-N8
arduino-nimble-library					24 kB160 MHz, 1 run
core-arduino					15.5 kB160 MHz, 1 run
core-cl2000-o2-ram			1.5 kB120 MHz, 1 run
core-gcc-os				88016 MHz, 1 run
core-ticlang	2 kB32 MHz, 1 run
core-ticlang-ble-hogp-build-only		33.8 kB to 33.9 kB48 MHz

Sources

Run records are in the notebook, not this copy.

Average time to build the HID report descriptor, 200 builds, on the board's own timer us, lower is better

Variant	LP-MSPM0G3519	LP-EM-CC2340R5	LAUNCHXL-F2800137	NUCLEO-U083RC	ESP32-C6-DevKitC-1-N8
arduino-nimble-library					5 to 9160 MHz
core-arduino					14160 MHz
core-cl2000-o2-ram			8120 MHz, 1 run
core-gcc-os				38916 MHz, 1 run
core-ticlang	3532 MHz, 1 run
core-ticlang-ble-hogp		20748 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Average time to pack one 19-byte input report, 1000 packs, on the board's own timer us, lower is better

Variant	LP-MSPM0G3519	LP-EM-CC2340R5	LAUNCHXL-F2800137	NUCLEO-U083RC	ESP32-C6-DevKitC-1-N8
arduino-nimble-library					112 to 131160 MHz
core-arduino					1160 MHz
core-cl2000-o2-ram			5.64120 MHz, 1 run
core-gcc-os				4116 MHz, 1 run
core-ticlang	1732 MHz, 1 run
core-ticlang-ble-hogp		1748 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Flash used by the ported UART bootloader (ferenc-nemeth/stm32-bootloader), text plus data B, lower is better

Variant	LP-MSPM0G3519
ticlang-5.1.1-os	3.2 kB32 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Static RAM (data + bss, including the stack the linker reserves) of the ported UART bootloader B, lower is better

Variant	LP-MSPM0G3519
ticlang-5.1.1-os	4.1 kB32 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Host checks passed out of 6 for the ported UART bootloader (prompt, update, flash content, reset jump, bad CRC retry, abort and recovery) count, higher is better

Variant	LP-MSPM0G3519
fresh-agent-port	632 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Seconds from the first Xmodem packet to the ACK of EOT, including the erase of the application area s, lower is better

Variant	LP-MSPM0G3519
1024-byte-packets-4480-b	1.5432 MHz
128-byte-packets-1416-b	1.1932 MHz

Sources

Run records are in the notebook, not this copy.

DNS sinkhole answers a blocked name with the fixed address yes/no

Variant	LP-EM-CC35X1
fixed-a-record	yes1 run

Sources

ports/dns-sinkhole/measure.py

Median DNS reply time for a blocked name, from a Mac on the same network ms, lower is better

Variant	LP-EM-CC35X1
71492-domains	6

Sources

ports/dns-sinkhole/measure.py

95th percentile DNS reply time for a blocked name ms, lower is better

Variant	LP-EM-CC35X1
71492-domains	20

Sources

ports/dns-sinkhole/measure.py

Median DNS reply time for an allowed name (forwarded to the upstream resolver) ms, lower is better

Variant	LP-EM-CC35X1
71492-domains	70

Sources

ports/dns-sinkhole/measure.py

95th percentile DNS reply time for an allowed name ms, lower is better

Variant	LP-EM-CC35X1
71492-domains	172

Sources

ports/dns-sinkhole/measure.py

DNS transaction IDs come from the chip hardware random generator yes/no

Variant	LP-EM-CC35X1
psa_generate_random	yes1 run

Sources

ports/dns-sinkhole/firmware/dns_sinkhole.c

Flash taken by the blocklist table per 10,000 domains B, lower is better

Variant	LP-EM-CC35X1
40-bit-hash	48.8 kB1 run

Sources

ports/dns-sinkhole/gen_blocklist.py

Static RAM of the DNS blocker code B, lower is better

Variant	LP-EM-CC35X1
full	7301 run

Sources

ports/dns-sinkhole/build.sh

Flash of the whole image with the blocklist table B, lower is better

Variant	LP-EM-CC35X1
71492-domains	1.5 MB1 run

Sources

ports/dns-sinkhole/build.sh

Largest blocklist the application slot holds (link arithmetic) count, higher is better

Variant	LP-EM-CC35X1
flash-slot	304,2981 run

Sources

ports/dns-sinkhole/README.md

Expected console lines the MicroPython image printed in the Cortex-M0 emulator (host only) count, higher is better

Variant	LP-MSPM0G3507	LP-MSPM0G3519
full-unicorn-cortex-m0		111 run
min-unicorn-cortex-m0	71 run	71 run

Sources

Run records are in the notebook, not this copy.

MicroPython REPL banner on the real UART yes/no

Variant	LP-MSPM0G3507	LP-MSPM0G3519
full		yes80 MHz, 1 run
min

Sources

ports/micropython/README.md

machine.Pin toggle read back on the pin yes/no

Variant	LP-MSPM0G3519
full

Sources

ports/micropython/README.md

Pystone iterations per second at 80 MHz iter/s, higher is better

Variant	LP-MSPM0G3519
full

Sources

ports/micropython/README.md

The data file the port reads in place hashes to the expected SHA-1 at boot yes/no

Variant	LP-EM-CC35X1
doom-s3-wad-xip	yes160 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Checksum of the first 400 rendered frames equals the reference build yes/no

Variant	LP-EM-CC35X1
doom-demo1	yes160 MHz, 1 run
doom-demo2
doom-demo3	no160 MHz, 1 run
doom-whx-attract	yes160 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Frames per second of a ported engine running a timedemo, no screen output fps, higher is better

Variant	LP-EM-CC35X1
doom-demo1	61.5160 MHz
doom-demo3	42.6160 MHz, 1 run
doom-whx-all-after400	22.5160 MHz, 1 run
doom-whx-crc-every-frame	20.3160 MHz, 1 run
doom-whx-panel
doom-whx-xip-fast-ram	36.7160 MHz, 1 run
doom-whx-xip-fast-ram-o2	39.2160 MHz, 1 run
doom-whx-xip-tcm-statics-only	34.5160 MHz, 1 run

Sources

docs/experiments/2026-10-20-doom-smallest-ti-chip.md, ports/doom/target-cc35x1-whx/PANEL-WIRING.md

A byte sent over the serial port reaches the firmware (demo select) yes/no

Variant	LP-EM-CC35X1
doom	no160 MHz

Sources

Run records are in the notebook, not this copy.

Disk used by downloaded crates and one release build MB, lower is better

Variant	LP-MSPM0L1306
embassy-cache	23132 MHz, 1 run

Sources

ports/rust-mspm0/README.md

A recorded demo plays to its end on the board yes/no

Variant	LP-EM-CC35X1
doom-whx-demo1	yes160 MHz, 1 run
doom-whx-demo2	yes160 MHz, 1 run
doom-whx-demo3	yes160 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

CRC chain over all frames of a run equals the emulator run yes/no

Variant	LP-EM-CC35X1
doom-whx-attract	yes160 MHz, 1 run
doom-whx-o2-all-frames	yes160 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Disk used by the Rust toolchain for the target MB, lower is better

Variant	LP-MSPM0L1306
rustup-minimal	60232 MHz, 1 run

Sources

ports/rust-mspm0/README.md

WebAssembly interpreter slowdown against the same C (fib 20, instruction counts) x, lower is better

Variant	LP-MSPM0G3519	LP-MSPM33C321A
wasm3-v0.9.0	37.1 calc.	27.6 calc.
wasm3-v0.9.0-board	43.580 MHz, 1 run

Sources

ports/wasm3/README.md

WebAssembly interpreter slowdown against the same C (20k-iteration loop, instruction counts) x, lower is better

Variant	LP-MSPM0G3519	LP-MSPM33C321A
wasm3-v0.9.0	18.6 calc.	21 calc.
wasm3-v0.9.0-board	28.980 MHz, 1 run

Sources

ports/wasm3/README.md

Stored count survives a software reset storm yes/no

Variant	LP-MSPM0L1306	LP-EM-CC35X1
msp430-fram-persist	yes32 MHz, 1 run
wifi-credentials-nvocmp

Sources

ports/wifi-provisioning-ble/README.md

Stored count survives a program load that leaves the data sector alone yes/no

Variant	LP-MSPM0L1306
msp430-fram-persist	yes32 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Stored count survives power removed yes/no

Variant	LP-MSPM0L1306
msp430-fram-persist	yes32 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Time to store one word in non-volatile memory us, lower is better

Variant	LP-MSPM0L1306
msp430-fram-persist	5632 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Time to erase the flash sector that holds the stored data us, lower is better

Variant	LP-MSPM0L1306	LP-MSPM0G3519
msp430-fram-persist	1,98632 MHz, 1 run
zephyr-flash-driver		2,625 to 2,62680 MHz

Sources

Run records are in the notebook, not this copy.

Bytes of a burst echoed correctly while the chip sleeps between bytes bytes, higher is better

Variant	LP-MSPM0L1306
msp430-uart-echo-bus-standby0	19032 MHz, 1 run
msp430-uart-echo-lfclk-standby0	532 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Test programs or checks whose output on the board equals the host reference count, higher is better

Variant	LP-MSPM0G3519
chip8-test-suite-frame-buffer	580 MHz, 1 run
zx48-screen-bytes-vs-host	280 MHz, 1 run

Sources

ports/chip8/tools/verify_board.py, ports/zx-spectrum/tools/verify_board.py

Interactive interpreter checks that pass on the board count, higher is better

Variant	LP-MSPM0G5187	LP-EM-CC2340R5	LP-MSPM33C321A
micropython-zephyr-basic-heap20k		648 MHz, 1 run
micropython-zephyr-stock-prj-gpio			7160 MHz, 1 run
ulisp-4.9a	1980 MHz

Sources

ports/ulisp/tests/repl_check.py

Endless recursion ends in an error message and the next line still runs yes/no

Variant	LP-MSPM0G5187
ulisp-4.9a	180 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

LED pin changes seen by a pin read while a program blinks it count, higher is better

Variant	LP-MSPM0G5187
ulisp-4.9a-autoexec	680 MHz, 1 run

Sources

Run records are in the notebook, not this copy.

Restoring a board's saved flash image after a test (done, or waiting with the reason) yes/no

Variant	LP-MSPM0C1104	LP-EM-CC2340R5	LP-CC1354P10-1	LP-EM-CC2755P10	LP-EM-CC35X1
digital-lab-node-saci				196 MHz, 1 run
flash-1mb-ccfg-2kb			11 run
frl-ccs13			113.548 MHz, 1 run
frl-main-and-ccfg		148 MHz
frl-toolbox-hashed-copy					11 run
frl-toolbox-reprogram					56.9
main-restore-hex
oob-after-factory-reset	124 MHz, 1 run
oob-after-factory-reset-3	124 MHz, 1 run
oob-after-factory-reset-4	124 MHz, 1 run

Sources

docs/frl.md, portability/boosterpacks/audio/tools/am13_algos_quiet.py

What is left to run, per board

LP-MSPM0G3507 56 of 301 tests measured

Could work Loop rate with 20 percent of the period free (computed, superseded)

Dhrystone 2.1; 256-point complex FFT; Minimal tasks that fit

Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Energy per occupancy inference; The part works on this board from this repo's driver; MicroPython REPL banner on the real UART

LP-MSPM0L1306 88 of 265 tests measured

Could work Loop rate with 20 percent of the period free (computed, superseded)

n/a Hardware AES against software (ratio at one message size); DAC to ADC loop error; Energy per BLE connection (BlueJoule job); Energy per radio transmit event (one advertisement or one packet)

Not for this chip Vendor AI example, one inference

Energy per occupancy inference; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; Energy per ADC sample at 1 ksps from sleep; Energy per 1 kB flash write

LP-MSPM0C1104 34 of 213 tests measured

Not for this chip CoreMark; CoreMark per MHz; Dhrystone 2.1; 256-point complex FFT

Not for this chip One step of frl's connect ladder reaches the chip on its own (frl identify --connect-step)

n/a sin plus cos; Divide; Hardware AES against software (ratio at one message size); DAC to ADC loop error

Not for this chip Zephyr upstream support

Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration

LP-MSPM0G3519 121 of 301 tests measured

Could work Touch to changed pixel (computed from measured parts, or measured with a real finger); Interpreter speed divided by the pace a typical program needs; Mean CPU cycles to encode one 10 ms LC3 frame (value is the mean; the note holds the worst frame); Mean CPU cycles to decode one 10 ms LC3 frame

Not for this chip openssl speed, EVP, 8 KB blocks

Dominant frequency the pack microphone hears from a DAC8311 test tone (1000 Hz expected; about 60 Hz is the silent mains floor); Pystone iterations per second at 80 MHz

Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Active current per MHz running CoreMark (slope between two clocks); Lightest sleep, CPU off, clocks on (SLEEP); STOP-class mode, low-speed clock kept; Deepest mode, RAM lost, pin wake; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; Energy per ADC sample at 1 ksps from sleep; Energy per 1 kB flash write; machine.Pin toggle read back on the pin; MicroPython: a finger drag is tracked on screen (a person looked); Serial console shows the program banner and a running count; A finger on the panel pad reaches the program as a key press; A finger on the touch keyboard reaches the program as a typed character

LP-MSPM0G5187 103 of 278 tests measured

Could work NPU time against the same graph as NNC CPU code, or against hand-written C; Signal to error ratio of the 8-bit graph against the float reference

Not for this chip ML-DSA-44 sign (FIPS 204), mean over 100 messages; openssl speed, EVP, 8 KB blocks

n/a DAC to ADC loop error; Dominant frequency the pack microphone hears from a DAC8311 test tone (1000 Hz expected; about 60 Hz is the silent mains floor)

Energy per block on the NPU against the CPU; Streaming with the taps-1 overlap carried between blocks works; Share of CPU time left free while the NPU runs a block; The same graph on the AM13E230; Repeat runs without a reflash; The Lisp interpreter reads an ADC channel; The Lisp interpreter saves its workspace to flash and restores it

Not for this chip Zephyr upstream support

Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Active current per MHz running CoreMark (slope between two clocks); Lightest sleep, CPU off, clocks on (SLEEP); STOP-class mode, low-speed clock kept; Deepest mode, RAM lost, pin wake; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; Days on a CR2032 (225 mAh) at the reference duty cycle: wake once a minute, 50 ms active at top clock, deepest sleep with RAM kept between; The part works on this board from this repo's driver

LP-EM-CC2340R5 80 of 305 tests measured

Could work Loop rate with 20 percent of the period free (computed, superseded)

n/a Hardware CRC against a table; DAC to ADC loop error

Fastest clean UART loopback; Vendor AI example, one inference; Packet loss over one hour; Vendor examples that build from a terminal

Not for this chip The part works on this board from this repo's driver

Time on air for one packet; Charge per radio event at 3.0 V; Average current at one event per 1000 ms, 3.0 V; Distance at which packet error rate stays under 1%; Deepest sleep with RAM kept (STANDBY-class); Energy per BLE connection (BlueJoule job); Energy per CoreMark iteration; Energy per occupancy inference; Active current per MHz running CoreMark (slope between two clocks); Lightest sleep, CPU off, clocks on (SLEEP); STOP-class mode, low-speed clock kept; Deepest mode, RAM lost, pin wake; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; Energy per ADC sample at 1 ksps from sleep; Energy per 1 kB flash write; Energy per radio transmit event (one advertisement or one packet); Days on a CR2032 (225 mAh) at the reference duty cycle: wake once a minute, 50 ms active at top clock, deepest sleep with RAM kept between; Button press on a Bluetooth LE HID device to the host's action (a media key changing the player), median of 20

LP-CC1354P10-1 65 of 305 tests measured

n/a Hardware CRC against a table

Blink example build time; Blink flash time; DAC to ADC loop error; Fastest clean UART loopback; Vendor AI example, one inference; Payload throughput board to board; BLE echo round trip, 20 bytes; Over-the-air update completed; Examples run with no crash; Loop rate with 20 percent of the period free (computed, superseded); ECDH P-256 shared secret (the classical step ML-KEM replaces)

Not for this chip Zephyr upstream support

LAUNCHXL-F2800137 64 of 261 tests measured

Could work Loop rate with 20 percent of the period free (computed, superseded)

n/a Hardware AES against software (ratio at one message size); Hardware CRC against a table

Box or plug-in to a changed blink

Not for this chip Zephyr upstream support; RTOS task switch round trip; Minimal tasks that fit

DAC to ADC loop error

NUCLEO-G071RB 0 of 278 tests measured

NUCLEO-U083RC 44 of 278 tests measured

Not for this chip DAC to ADC loop error; Fastest clean UART loopback

n/a Energy per BLE connection (BlueJoule job); Energy per radio transmit event (one advertisement or one packet)

BitNetMCU handwritten digit, one inference; Vendor AI example, one inference; Examples that build after moving to another chip

Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Energy per occupancy inference; Active current per MHz running CoreMark (slope between two clocks); Lightest sleep, CPU off, clocks on (SLEEP); STOP-class mode, low-speed clock kept; Deepest mode, RAM lost, pin wake; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; Energy per ADC sample at 1 ksps from sleep; Energy per 1 kB flash write; Days on a CR2032 (225 mAh) at the reference duty cycle: wake once a minute, 50 ms active at top clock, deepest sleep with RAM kept between; Digital-power PID loop, trigger to duty write (mean); Flash used by a hobby part's driver plus its demo; RAM used by a hobby part's driver plus its demo; The part works on this board from this repo's driver; SSD1306 OLED full-frame time (128x64, I2C); Highest control-loop trigger rate with no missed period in 1,000; Control-loop CPU load, interrupt entry to exit, as a percent of a 100 kHz period; Digital-power loop worst case, trigger to duty write, as a percent of a 100 kHz period; Digital-power loop jitter, worst minus best trigger to duty write; Highest digital-power loop rate with no missed period in 1,000; Digital-power loop CPU load, interrupt entry to exit, as a percent of a 100 kHz period; ML-KEM-768 key generation (FIPS 203); ML-KEM-768 encapsulation; ML-KEM-768 decapsulation; ML-DSA-44 key generation (FIPS 204); ML-DSA-44 sign (FIPS 204), mean over 100 messages; ML-DSA-44 verify; Peak stack of one post-quantum call, by painting; SHA-256 of 1 kB, cycles per byte

FRDM-MCXA153 26 of 278 tests measured

n/a Hardware AES against software (ratio at one message size)

PWM or timer trigger to first line of ISR; Trigger, ADC, control step, new PWM value (mean)

DAC to ADC loop error; Fastest clean UART loopback; Deepest sleep with RAM kept (STANDBY-class); Examples that build after moving to another chip; Active current per MHz running CoreMark (slope between two clocks)

Raspberry Pi Pico 2 H 24 of 278 tests measured

Not for this chip Hardware AES against software (ratio at one message size)

DAC to ADC loop error; Fastest clean UART loopback; Examples run with no crash

ESP32-C6-DevKitC-1-N8 36 of 290 tests measured

n/a Hardware CRC against a table

Flash used by the ported UART DMA ring-buffer echo (MaJerle stm32-usart-uart-dma-rx-tx); Static RAM (data + bss) of the ported UART DMA ring-buffer echo; Packets per second answered for 500 back-to-back 5-byte packets, console UART; Replies lost out of 500 back-to-back packets; Median host-measured round trip of one packet to its 4-byte reply

LP-MSPM33C321A 138 of 301 tests measured

Could work WebAssembly interpreter slowdown against the same C (fib 20, instruction counts); WebAssembly interpreter slowdown against the same C (20k-iteration loop, instruction counts); CPU cycles per instruction of a whole program, from its time and an emulator instruction count

Not for this chip openssl speed, EVP, 8 KB blocks

Dominant frequency the pack microphone hears from a DAC8311 test tone (1000 Hz expected; about 60 Hz is the silent mains floor); Cycles per frame of an audio algorithm on the chip (DWT counter); An audio algorithm runs live on the BoosterPack (heard or read back); The S2 button starts a new story on the screen; The decoded clip plays out of a DAC or speaker

DAC to ADC loop error; Fastest clean UART loopback; Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Energy per occupancy inference; Power-up to SPI screen initialized and on; Active current per MHz running CoreMark (slope between two clocks); Lightest sleep, CPU off, clocks on (SLEEP); STOP-class mode, low-speed clock kept; Deepest mode, RAM lost, pin wake; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; Days on a CR2032 (225 mAh) at the reference duty cycle: wake once a minute, 50 ms active at top clock, deepest sleep with RAM kept between

LP-AM13E230 111 of 278 tests measured

Interrupt pend to first line of handler; GPIO toggle; Fastest ADC with DMA; Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Energy per occupancy inference; Active current per MHz running CoreMark (slope between two clocks); Deepest mode, RAM lost, pin wake; Cycles per frame of an audio algorithm on the chip (DWT counter); An audio algorithm runs live on the BoosterPack (heard or read back); LED pin toggles under the Rust build (probe GPIO read, changes counted); Restoring a board's saved flash image after a test (done, or waiting with the reason)

LP-EM-CC2755P10 66 of 305 tests measured

Not for this chip openssl speed, EVP, 8 KB blocks

n/a Hardware CRC against a table; DAC to ADC loop error

Digital-power PID loop, trigger to duty write (mean); Digital-power loop worst case, trigger to duty write, as a percent of a 100 kHz period; Digital-power loop jitter, worst minus best trigger to duty write; Highest digital-power loop rate with no missed period in 1,000; Digital-power loop CPU load, interrupt entry to exit, as a percent of a 100 kHz period

Not for this chip Vendor AI example, one inference

LP-EM-CC35X1 64 of 305 tests measured

Not for this chip openssl speed, EVP, 8 KB blocks; RAM the software H.264 decode needs: decoder heap peak plus input buffer plus RGB565 strip

Not for this chip Embench programs that fit and pass; PWM or timer trigger to first line of ISR; Trigger, ADC, control step, new PWM value (mean)

Data-cache hit and miss counters readable from the application; External flash read mode or clock changeable from the application

Not for this chip Hardware CRC against a table; Zephyr upstream support

Interrupt pend to first line of handler; GPIO toggle; 256-point complex FFT; sin plus cos; Divide; Hardware AES against software (ratio at one message size); RTOS task switch round trip; Minimal tasks that fit; DAC to ADC loop error; One field-oriented control step; Setpoint or state change still works with the internet cable pulled; Controllers that hold one device at once and all still control it; Setup again after a reflash and factory reset, per platform; Deepest sleep with RAM kept (STANDBY-class); Active current per MHz running CoreMark (slope between two clocks); Lightest sleep, CPU off, clocks on (SLEEP); STOP-class mode, low-speed clock kept; Deepest sleep with a 1 s timer wake and nothing to do; Deepest sleep with RAM kept to first instruction; AES-128 CBC encrypt of 1 kB in software (tiny-AES-c); Control loop worst case, trigger to new PWM value, as a percent of a 100 kHz period; Control loop jitter, worst minus best trigger to new PWM value; Highest control-loop trigger rate with no missed period in 1,000; Control-loop CPU load, interrupt entry to exit, as a percent of a 100 kHz period; AES-128 in hardware, cycles per byte; AES-128 in software (tiny-AES-c), cycles per byte; ML-KEM-768 key generation (FIPS 203); ML-KEM-768 encapsulation; ML-KEM-768 decapsulation; ML-DSA-44 key generation (FIPS 204); ML-DSA-44 sign (FIPS 204), mean over 100 messages; ML-DSA-44 verify; Peak stack of one post-quantum call, by painting; Flash used by the post-quantum code in a linked image; SHA-256 of 1 kB, cycles per byte; ECDSA P-256 sign of a 32-byte digest; ECDSA P-256 verify; ECDH P-256 shared secret (the classical step ML-KEM replaces); Software H.264 decode (baseline profile, no video hardware): frames per second; Stored count survives a software reset storm; Whether a ported display driver brought the panel up (yes or no)

LAUNCHXL-F28P65X 12 of 138 tests measured

Box or plug-in to a changed blink; Blink flash time

IWRL6432BOOST 7 of 99 tests measured

Not for this chip Box or plug-in to a changed blink; Zephyr upstream support

CoreMark; Host-side code for a radio or sensor: flash and RAM on this host; Box open to the first valid radar frame parsed on the Mac (TI image, a custom parser); Average power against update rate; Deep-sleep power with memory retained; Reset release to the first frame, application booting from flash; Raw ADC samples streamed over SPI to a host, frames lost; Hardware accelerator FFT against the Cortex-M4F software FFT; Zone on-events in an empty room; Cue to a zone lighting (detect) or clearing (clear), from a timed walk-through; Radar frame to Home Assistant state change

LP-AM263P 28 of 168 tests measured

n/a Blink image flash

Minimal tasks that fit

Fastest ADC with DMA; DAC to ADC loop error; Fastest clean UART loopback; Deepest sleep with RAM kept (STANDBY-class); Energy per CoreMark iteration; Active current per MHz running CoreMark (slope between two clocks)

SK-TDA4VM 31 of 212 tests measured

n/a Vendor examples that build from a terminal

Box or plug-in to a changed blink; Embench programs that fit and pass; OpenGL ES on the GPU: kmscube or glmark2 frames per second

TMDS62LEVM (AM62L EVM) 37 of 85 tests measured

Dhrystone 2.1

Not for this chip AES-128 in hardware, cycles per byte

The rows as CSV, one file per board: AM13E230, AM263P, CC1354P10-1, CC2340R5, CC2755P10, CC35X1, ESP32-C6 (Espressif), F2800137, F28P65X, FRDM-MCXA153 (NXP), IWRL6432BOOST, MSPM0C1104, MSPM0G3507, MSPM0G3519, MSPM0G5187, MSPM0L1306, MSPM33C321A, NUCLEO-U083RC (ST), Pico 2 (RPi), SK-TDA4VM, TMDS62LEVM (AM62L EVM).

Method

Each test is a program that reports its own cycle count or time, and a script reads it back from the chip. A benchmark that checks its own answer, such as CoreMark's CRC, the AES known answer or a model's test vectors, gives no number when the check fails. Three runs give a median. A row marked 1 run is a single run, which can be off by whatever the Mac, the supply or the radio did that minute: cycle counts from a self-checking benchmark barely move, wall-clock times and radio latencies move a lot. Every number is a row in the results store with its source file, date, clock and build. The full rules are in the method above.

Calculated numbers. A number read from a board is measured. These are worked out from other numbers, and each says so where it appears:

Interrupt response in nanoseconds: the best-case cycle count divided by the clock. The chart note says so; the cycles are measured.
The outlined bars in the radio RAM chart, marked "calculated" in its legend: the chip's RAM plus the bytes the linker said a stack that did not link was short.
Table cells marked "calc.", computed from other rows:
- CPU cycles per instruction of a whole program, from its time and an emulator instruction count
- Days on a CR2032 (225 mAh) at the reference duty cycle: wake once a minute, 50 ms active at top clock, deepest sleep with RAM kept between
- Driver current-sense output, cached ADC sample (not a calibrated phase current)
- Emulated machine speed as per mille of the real machine, 159 meaning 15.9 percent
- Flash used by an outside project ported to the board, with its shim
- Host-side code for a radio or sensor: flash and RAM on this host
- Interpreter speed divided by the pace a typical program needs
- LVGL demo screen refreshes per second
- Loop rate with 20 percent of the period free (computed, superseded)
- Mean CPU cycles to decode one 10 ms LC3 frame
- Mean CPU cycles to encode one 10 ms LC3 frame (value is the mean; the note holds the worst frame)
- NPU time against the same graph as NNC CPU code, or against hand-written C
- RAM left after linking a stack
- RAM used by an outside project ported to the board, with its shim
- Signal to error ratio of the 8-bit graph against the float reference
- Tokens per second of a small language model
- Touch to changed pixel (computed from measured parts, or measured with a real finger)
- WebAssembly interpreter slowdown against the same C (20k-iteration loop, instruction counts)
- WebAssembly interpreter slowdown against the same C (fib 20, instruction counts)

Ranks. Control loop and interrupt ranks use the times drawn in the charts. Boards whose times match at the digits shown share a rank: two boards tied for third both read 3 of 11, and the next is 5.

Probe floors. Current figures come from the kit's own EnergyTrace probe. It reads about 1.5 mA on the MSPM0 L1306 with the chip shut down, so every chip figure on that board is the raw reading minus that floor, and one probe count is about 0.1 mA. Very small currents sit at the edge of what the probe can resolve.

Model names. The AI models carry plain names on this site. The ids behind them stay in the data and in the hover text:

Name on this site	Model id in the data
Room occupancy: tiny	`occ_cnn_500`
Room occupancy: small	`occ_cnn_2k`
Room occupancy: medium	`occ_cnn_8k`
Room occupancy: large	`occ_cnn_30k`
Object detection, small (lite)	`yolox-nano-lite-416`
Object detection, small	`yolox-nano-416`
Face detection	`yolox-s-lite-widerface-640`
Object detection, medium	`yolox-s-lite-640`
Pose estimation	`yolox-s-pose-640`
Image classification (RegNetX 200MF)	`regnetx-200mf-224`
Object detection (SSD-Lite MobileNetV2)	`ssd-lite-mobv2-512`
Segmentation (DeepLabV3-Lite)	`deeplabv3lite-mobv2-ade20k-512`
Segmentation (DeepLabV3-Lite, 768x432)	`deeplabv3lite-mobv2-robokit-768x432`

Coverage: tests measured per board

LP-MSPM0G350755 of 301

LP-MSPM0L130684 of 265

LP-MSPM0C110431 of 213

LP-MSPM0G3519114 of 301

LP-MSPM0G518791 of 278

LP-EM-CC2340R576 of 305

LP-CC1354P10-161 of 305

LAUNCHXL-F280013750 of 261

NUCLEO-G071RB (ST)0 of 278

NUCLEO-U083RC (ST)43 of 278

FRDM-MCXA153 (NXP)26 of 278

Raspberry Pi Pico 2 H24 of 278

ESP32-C6-DevKitC-1-N8 (Espressif)36 of 290

LP-MSPM33C321A132 of 301

LP-AM13E23078 of 278

LP-EM-CC2755P1065 of 305

LP-EM-CC35X163 of 305

LAUNCHXL-F28P65X8 of 138

IWRL6432BOOST7 of 99

LP-AM263P26 of 168

SK-TDA4VM31 of 212

TMDS62LEVM (AM62L EVM)29 of 85

Questions this page raises

Which numbers can I compare straight across boards?: Cycle counts and per-MHz scores from the same build at the same compiler level: CoreMark per MHz with ticlang-bare-o3, interrupt latency bare, the model cells in the "fastest plain-C" chart. Milliseconds and absolute CoreMark depend on the clock, so divide by the MHz first. Rows built a different way on each board are not like for like.
Why is the MSPM0 L1306's sleep current 6.2 µA when the data sheet says 1.0?: The board's own probe reads 1.500 mA with the chip shut down, and that is subtracted from every figure. The 6.2 µA is a mean of 36k samples, at the edge of what the probe can resolve, so the probe is the likely cause of the gap.
Why is the MSPM0 L1306's control step so much slower than the LAUNCHXL-F2800137's?: In clock cycles it is about 30 to 45 times (a mean of 12,851 against 286 to 439). In time it is about 108 to 167 times (401.6 µs against 2.4 to 3.7 µs), because the L1306 runs at 32 MHz and the F2800137 at 120 MHz. The Cortex-M0+ has no floating-point unit, so the same C runs in software float there, while the C28x has an FPU. The F2800137's own software-against-TMU rows show its trig unit adds another 2 times on top.
What does "does not fit" mean, and why does it have a number?: The program linked but needed more RAM than the chip has; the number is how many bytes short, from the linker. It is a build result for one configuration built with local changes, not a measurement on the board.
Why are some radio rows "failed"?: The CC2340R5's Zigbee update from Home Assistant starts and then stops: the best of nine transfers reached 88.74%. Every long transfer ended on lost block replies, which points at the radio environment. Its Bluetooth on TI's Zephyr fork boots but sent nothing on the air. Each has a write-up on its project page.
Why do the two Zigbee "command latency" cells differ by 16 times?: They measure different things: a light being switched (CC2340R5, 95 ms) and a sleepy sensor being read, which waits for the sensor's 3 s poll (CC1354P10-1, 1,574 ms). The difference is in what each row times, not in the chips.
Why is the STOP0 current not the data sheet's 151 µA?: The data sheet measures STOP0 with the main oscillator slowed to 4 MHz. The program left it at 32 MHz, so the 329 µA is a different configuration, not a worse chip.
Why is the LAUNCHXL-F2800137's CoreMark not a score?: CoreMark assumes an 8-bit byte; the C28x core's smallest addressable unit is 16 bits, so three of CoreMark's four self-checks fail unpatched. The patched build runs the same work but is not a certified score. Details in c2000-launchpad/bench/general-benchmarks.md.
Why does CoreMark per MHz on the MSPM0 G3507 drop from 32 MHz to 80 MHz?: Flash cannot answer at 80 MHz without the core waiting; it needs two wait states per flash read, so each iteration takes about 7% more cycles. Code copied to RAM would not pay that.

Sources

data/results/<board>.csv: one row per number, with its source file, date, clock and conditions
data/benchmark-suite.json: the test list, units and which way is better
data/boards.csv: the boards, their cores, clocks and RAM
Toolchain for most rows: Intel Mac, macOS 15.7, CCStudio 21.0.1, TI Arm Clang 5.1.1, Zephyr 4.4.2. The variant on each row names anything different.
Last measurement 2026-11-02. 1 kB = 1,024 bytes.