Skip to content

Instantly share code, notes, and snippets.

@jacobkahn
Created July 24, 2019 06:42
Show Gist options
  • Save jacobkahn/31fc0bedde5d2dd9d084d56b42dc3208 to your computer and use it in GitHub Desktop.
Save jacobkahn/31fc0bedde5d2dd9d084d56b42dc3208 to your computer and use it in GitHub Desktop.
NCCL Tests on 1 node p3dn.24xlarge + ethernet - all_reduce_perf -b 8 -e 1G -f 2 -g 1 -c 1 -n 100
#
# out-of-place in-place
# size count type redop time algbw busbw error time algbw busbw error
# (B) (elements) (us) (GB/s) (GB/s) (us) (GB/s) (GB/s)
ip-172-31-39-141:50926:50926 [0] NCCL INFO Launch mode Parallel
8 2 float sum 13.88 0.00 0.00 1e-07 13.97 0.00 0.00 1e-07
16 4 float sum 14.03 0.00 0.00 1e-07 14.04 0.00 0.00 1e-07
32 8 float sum 13.93 0.00 0.00 6e-08 13.86 0.00 0.00 6e-08
64 16 float sum 14.11 0.00 0.01 6e-08 13.88 0.00 0.01 6e-08
128 32 float sum 14.23 0.01 0.02 6e-08 14.11 0.01 0.02 6e-08
256 64 float sum 14.29 0.02 0.03 6e-08 14.26 0.02 0.03 6e-08
512 128 float sum 14.60 0.04 0.06 6e-08 14.55 0.04 0.06 6e-08
1024 256 float sum 15.41 0.07 0.12 5e-07 15.26 0.07 0.12 5e-07
2048 512 float sum 17.07 0.12 0.21 5e-07 16.67 0.12 0.21 5e-07
4096 1024 float sum 20.61 0.20 0.35 5e-07 19.75 0.21 0.36 5e-07
8192 2048 float sum 21.62 0.38 0.66 5e-07 20.72 0.40 0.69 5e-07
16384 4096 float sum 26.13 0.63 1.10 5e-07 25.15 0.65 1.14 5e-07
32768 8192 float sum 28.28 1.16 2.03 5e-07 27.43 1.19 2.09 5e-07
65536 16384 float sum 29.04 2.26 3.95 5e-07 28.58 2.29 4.01 5e-07
131072 32768 float sum 29.23 4.48 7.85 5e-07 29.09 4.51 7.89 5e-07
262144 65536 float sum 41.78 6.27 10.98 5e-07 41.59 6.30 11.03 5e-07
524288 131072 float sum 53.93 9.72 17.01 5e-07 53.40 9.82 17.18 5e-07
1048576 262144 float sum 89.44 11.72 20.52 5e-07 88.43 11.86 20.75 5e-07
2097152 524288 float sum 135.2 15.51 27.14 5e-07 132.9 15.78 27.62 5e-07
4194304 1048576 float sum 154.0 27.24 47.67 5e-07 148.0 28.35 49.61 5e-07
8388608 2097152 float sum 204.3 41.06 71.86 5e-07 203.7 41.19 72.08 5e-07
16777216 4194304 float sum 327.8 51.18 89.56 5e-07 325.7 51.51 90.14 5e-07
33554432 8388608 float sum 575.5 58.31 102.04 5e-07 574.9 58.37 102.14 5e-07
67108864 16777216 float sum 1040.5 64.50 112.87 5e-07 1042.7 64.36 112.63 5e-07
134217728 33554432 float sum 1900.5 70.62 123.59 5e-07 1904.7 70.47 123.32 5e-07
268435456 67108864 float sum 3883.5 69.12 120.96 5e-07 3891.2 68.99 120.72 5e-07
536870912 134217728 float sum 7467.8 71.89 125.81 5e-07 7485.9 71.72 125.51 5e-07
1073741824 268435456 float sum 14948 71.83 125.71 5e-07 14980 71.68 125.44 5e-07
# Out of bounds values : 0 OK
# Avg bus bandwidth : 36.1951
#
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment