You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Test Method:
NCCL_ALGO=[ALGO] ./all_reduce_perf -b 2G -e 4G -f 2 -g [gpus], where all_reduce_perf is compiled from nccl-test github repo
Result:
when using ALGO=RING, busbw is about 365-370 GBps when gpus from 2 to 8, about 80% of 450GBps(NVLink)
when using ALGO=NVLS, things become interest. Busbw remains about 365-370 GBps when gpus from 2 to 4. However, Busbw increase from 370 to 420, 440, 460, and 480 GBps when gpus increase from 4 to 5, 6, 7, and 8.
Besides, when gpus from 5 to 6, 7, and 8 at ALGO=NVLS, algbw also increases from 258 to about 265, 270, and 273.
Question:
how does 480GBps been calculated? SHArP ALU @nvswitch3 has 400GFLOPS@float32. Is 400GFLOPS any relationship with 480GBps?
each H100 should be connected directly with all nvswitches, so how could algbw@5gpus lower than algbw@8gpus? In addition, why gpus=2, 3, and 4 remain 370GBps like Ring allreduce?
Looking forward to you reply
The text was updated successfully, but these errors were encountered:
Hello, @sjeaugey nccl team!
ENV:
8 * H100_SXM, ngctorch enviornment
Test Method:
NCCL_ALGO=[ALGO] ./all_reduce_perf -b 2G -e 4G -f 2 -g [gpus], where all_reduce_perf is compiled from nccl-test github repo
Result:
Question:
Looking forward to you reply
The text was updated successfully, but these errors were encountered: