Skip to content

Instantly share code, notes, and snippets.

@jacobkahn
Created July 24, 2019 06:41
Show Gist options
  • Save jacobkahn/4e43bfccd79bf864970a98c191cd6c9b to your computer and use it in GitHub Desktop.
Save jacobkahn/4e43bfccd79bf864970a98c191cd6c9b to your computer and use it in GitHub Desktop.
NCCL Tests on 1 node p3dn.24xlarge + efa - 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-172-31-39-141:50814:50814 [0] NCCL INFO Launch mode Parallel
8 2 float sum 13.84 0.00 0.00 1e-07 13.75 0.00 0.00 1e-07
16 4 float sum 13.82 0.00 0.00 1e-07 13.96 0.00 0.00 1e-07
32 8 float sum 13.79 0.00 0.00 6e-08 13.79 0.00 0.00 6e-08
64 16 float sum 13.88 0.00 0.01 6e-08 13.73 0.00 0.01 6e-08
128 32 float sum 14.04 0.01 0.02 6e-08 13.94 0.01 0.02 6e-08
256 64 float sum 14.11 0.02 0.03 6e-08 14.05 0.02 0.03 6e-08
512 128 float sum 14.30 0.04 0.06 6e-08 14.24 0.04 0.06 6e-08
1024 256 float sum 15.17 0.07 0.12 5e-07 15.17 0.07 0.12 5e-07
2048 512 float sum 16.96 0.12 0.21 5e-07 16.54 0.12 0.22 5e-07
4096 1024 float sum 20.60 0.20 0.35 5e-07 19.73 0.21 0.36 5e-07
8192 2048 float sum 21.55 0.38 0.67 5e-07 20.74 0.39 0.69 5e-07
16384 4096 float sum 26.10 0.63 1.10 5e-07 25.09 0.65 1.14 5e-07
32768 8192 float sum 28.07 1.17 2.04 5e-07 27.27 1.20 2.10 5e-07
65536 16384 float sum 29.11 2.25 3.94 5e-07 28.52 2.30 4.02 5e-07
131072 32768 float sum 29.32 4.47 7.82 5e-07 29.02 4.52 7.90 5e-07
262144 65536 float sum 41.38 6.33 11.09 5e-07 41.21 6.36 11.13 5e-07
524288 131072 float sum 53.62 9.78 17.11 5e-07 52.87 9.92 17.35 5e-07
1048576 262144 float sum 88.96 11.79 20.63 5e-07 88.06 11.91 20.84 5e-07
2097152 524288 float sum 129.4 16.21 28.37 5e-07 122.8 17.08 29.89 5e-07
4194304 1048576 float sum 148.4 28.26 49.45 5e-07 147.9 28.37 49.64 5e-07
8388608 2097152 float sum 203.5 41.23 72.15 5e-07 202.9 41.33 72.33 5e-07
16777216 4194304 float sum 327.8 51.18 89.57 5e-07 325.5 51.55 90.21 5e-07
33554432 8388608 float sum 572.5 58.61 102.57 5e-07 573.2 58.54 102.45 5e-07
67108864 16777216 float sum 1039.1 64.59 113.02 5e-07 1041.9 64.41 112.72 5e-07
134217728 33554432 float sum 1900.5 70.62 123.59 5e-07 1904.8 70.46 123.31 5e-07
268435456 67108864 float sum 3878.9 69.20 121.11 5e-07 3889.7 69.01 120.77 5e-07
536870912 134217728 float sum 7467.8 71.89 125.81 5e-07 7485.3 71.72 125.52 5e-07
1073741824 268435456 float sum 14946 71.84 125.72 5e-07 14977 71.69 125.46 5e-07
# Out of bounds values : 0 OK
# Avg bus bandwidth : 36.3371
#
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment