ImportBot: Import: mmapeak and torch-performance-test-data, @jimtjames, @rh3dai; repos ReinForce-II/mmapeak, cachenetics/170tune, zzc0721/torch-performance-test-data

2026-10-30T08:38:28Z

Import: mmapeak and torch-performance-test-data, @jimtjames, @rh3dai; repos ReinForce-II/mmapeak, cachenetics/170tune, zzc0721/torch-performance-test-data

← Older revision		Revision as of 08:38, 30 October 2026
Line 1:		Line 1:
			{{DISPLAYTITLE:mmapeak}}
	{{Status\|Reported}} {{Cards\|both}}		{{Status\|Reported}} {{Cards\|both}}

ImportBot: Import: mmapeak and torch-performance-test-data, @jimtjames, @rh3dai; repos ReinForce-II/mmapeak, cachenetics/170tune, zzc0721/torch-performance-test-data

2026-10-30T08:04:45Z

Import: mmapeak and torch-performance-test-data, @jimtjames, @rh3dai; repos ReinForce-II/mmapeak, cachenetics/170tune, zzc0721/torch-performance-test-data

New page

{{Status|Reported}} {{Cards|both}}

'''mmapeak''' is a small CUDA benchmark that measures peak tensor-core throughput for each data type. In the Discord it is the tool to ask a seller to run before you buy, to show the card's tensor cores work at full rate; [[#torch-performance-test-data|a PyTorch GEMM benchmark]] is the second option. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902}} Author: GitHub ReinForce-II. Repository: https://github.com/ReinForce-II/mmapeak

== Build and run ==
<syntaxhighlight lang="bash">
git clone https://github.com/ReinForce-II/mmapeak && cd mmapeak
mkdir build && cd build && cmake .. && make
./mmapeak # default target 3 s per kernel
./mmapeak -t 10 # longer per kernel
</syntaxhighlight>
Each kernel prints <code><nowiki>run: <ms> <T(fl)ops></nowiki></code> or <code><nowiki>not supported</nowiki></code>. The CMake file builds for sm_75 to sm_120; CUDA 12.8.1 or later is recommended. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/ReinForce-II/mmapeak}}.

== Reading it on a CMP 170HX ==
The 170HX is sm_80 (GA100). Only the INT4, INT8, FP16, BF16 and TF32 rows run; FP8, FP6, FP4 and MX rows print <code><nowiki>not supported</nowiki></code>. The rows that matter: <code><nowiki>mma_s4s4s32_8_8_32</nowiki></code>, <code><nowiki>mma_s8s8s32_*</nowiki></code>, <code><nowiki>mma_f16f16f16_*</nowiki></code>, <code><nowiki>mma_f16f16f32_*</nowiki></code>, <code><nowiki>mma_bf16bf16f32_*</nowiki></code>, <code><nowiki>mma_tf32tf32f32_16_16_8</nowiki></code>. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/ReinForce-II/mmapeak}}.

Expected: 150 to 200 TFLOPS on BF16 and FP16. About 83 TFLOPS on TF32 is normal; that is also what <code><nowiki>gpu-burn -tc</nowiki></code> reports. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902}} mmapeak loops matrix instructions on zero-filled data, so it measures peak issue rate and reads higher than a real GEMM. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/ReinForce-II/mmapeak}}. For comparison, the [[170tune]] reference 8 GB card measured BF16 188 to 193, FP16 159 to 160 and TF32 89 to 92 TFLOPS in GEMM at stock. {{Source|handle=rh3dai|date=2026-10-04|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1556287917775655053|repo=https://github.com/cachenetics/170tune}}.

== torch-performance-test-data ==
A PyTorch square-GEMM benchmark with a public results table (Chinese README). Its table has one row labelled CMP170HX, contributed by GitHub user bendy2: FP32 12.15, TF32 77.09, FP16 166.48, BF16 177.91 TFLOPS (older v1 protocol, no host details). Two unlabelled "NVIDIA Graphics Device" rows nearby match the same profile but are not confirmed as 170HX. For scale, the same table lists an A100 SXM4 80GB at 258.74 FP16 and 264.53 BF16, and an RTX 3090 at 75.64 FP16. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/zzc0721/torch-performance-test-data}}.
<syntaxhighlight lang="bash">
git clone https://github.com/zzc0721/torch-performance-test-data && cd torch-performance-test-data
uv sync && uv run test.py # or: python test.py
uv run test.py --device cuda:1 --sizes 2048 4096 8192
</syntaxhighlight>
Needs torch 2.7 or newer and Python 3.10 or newer. Version 2 of the benchmark reports the median TFLOPS and its spread per matrix size; do not submit results with a spread (CV) above 5 percent. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/zzc0721/torch-performance-test-data}}.

== Caveats ==
* Neither tool is CMP-specific. mmapeak has no 170HX output in its repository and no licence file; its last commit is 2025-12-09. torch-performance-test-data has no licence file either. {{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/ReinForce-II/mmapeak}}.<br />{{Source|handle=jimtjames|date=2026-10-11|link=https://discord.com/channels/1518634379604136178/1518695289278566450/1558871079395663902|repo=https://github.com/zzc0721/torch-performance-test-data}}.
* Throughput says the tensor cores run at full rate, not that they compute correct results. Check correctness separately; see [[Testing and burn-in]].

== See also ==
* [[Testing and burn-in]]
* [[Buying]]
* [[Tools]]

{{ImportFooter}}
[[Category:Tools]]
[[Category:8 GB]]
[[Category:10 GB]]

Mmapeak - Revision history

ImportBot: Import: mmapeak and torch-performance-test-data, @jimtjames, @rh3dai; repos ReinForce-II/mmapeak, cachenetics/170tune, zzc0721/torch-performance-test-data

ImportBot: Import: mmapeak and torch-performance-test-data, @jimtjames, @rh3dai; repos ReinForce-II/mmapeak, cachenetics/170tune, zzc0721/torch-performance-test-data