Myrinet logotype

Sockets-MX Performance Results for Myri-10G

Operating System: Windows

Performance

-- LATENCY:
NetPIPE is a well known benchmark which shows latency and
bandwidth results.
The following output was measured between two Windows x64
Compute Cluster Sever boxes with Myri-10G connected with a Myrinet switch.

NetPIPE_x64.exe -h wind01 -l 1 -u 4194304 -n 100 Send and receive buffers are 8192 and 8192 bytes Now starting the main loop 0: 1 bytes 100 times --> 1.31 Mbps in 6.10 usec 1: 2 bytes 100 times --> 2.44 Mbps in 6.25 usec 2: 3 bytes 100 times --> 3.68 Mbps in 6.22 usec 3: 4 bytes 100 times --> 4.88 Mbps in 6.26 usec 4: 6 bytes 100 times --> 7.45 Mbps in 6.14 usec 5: 8 bytes 100 times --> 9.69 Mbps in 6.30 usec 6: 12 bytes 100 times --> 14.68 Mbps in 6.23 usec 7: 13 bytes 100 times --> 15.83 Mbps in 6.26 usec 8: 16 bytes 100 times --> 19.69 Mbps in 6.20 usec 9: 19 bytes 100 times --> 23.53 Mbps in 6.16 usec 10: 21 bytes 100 times --> 25.27 Mbps in 6.34 usec 11: 24 bytes 100 times --> 29.63 Mbps in 6.18 usec 12: 27 bytes 100 times --> 32.47 Mbps in 6.35 usec 13: 29 bytes 100 times --> 35.80 Mbps in 6.18 usec 14: 32 bytes 100 times --> 40.09 Mbps in 6.09 usec 15: 35 bytes 100 times --> 38.28 Mbps in 6.97 usec 16: 45 bytes 100 times --> 50.05 Mbps in 6.86 usec 17: 48 bytes 100 times --> 53.04 Mbps in 6.90 usec 18: 51 bytes 100 times --> 55.70 Mbps in 6.98 usec 19: 61 bytes 100 times --> 67.30 Mbps in 6.91 usec 20: 64 bytes 100 times --> 69.61 Mbps in 7.01 usec 21: 67 bytes 100 times --> 73.92 Mbps in 6.92 usec 22: 93 bytes 100 times --> 97.40 Mbps in 7.28 usec 23: 96 bytes 100 times --> 104.04 Mbps in 7.04 usec 24: 99 bytes 100 times --> 106.98 Mbps in 7.06 usec 25: 125 bytes 100 times --> 132.82 Mbps in 7.18 usec 26: 128 bytes 100 times --> 137.35 Mbps in 7.11 usec 27: 131 bytes 100 times --> 113.70 Mbps in 8.79 usec 28: 189 bytes 100 times --> 168.35 Mbps in 8.56 usec 29: 192 bytes 100 times --> 170.43 Mbps in 8.59 usec 30: 195 bytes 100 times --> 169.06 Mbps in 8.80 usec 31: 253 bytes 100 times --> 220.10 Mbps in 8.77 usec 32: 256 bytes 100 times --> 223.34 Mbps in 8.75 usec 33: 259 bytes 100 times --> 225.32 Mbps in 8.77 usec 34: 381 bytes 100 times --> 321.90 Mbps in 9.03 usec 35: 384 bytes 100 times --> 322.65 Mbps in 9.08 usec 36: 387 bytes 100 times --> 323.75 Mbps in 9.12 usec 37: 509 bytes 100 times --> 403.05 Mbps in 9.64 usec 38: 512 bytes 100 times --> 417.78 Mbps in 9.35 usec 39: 515 bytes 100 times --> 415.12 Mbps in 9.46 usec 40: 765 bytes 100 times --> 567.48 Mbps in 10.29 usec 41: 768 bytes 100 times --> 575.01 Mbps in 10.19 usec 42: 771 bytes 100 times --> 571.93 Mbps in 10.28 usec 43: 1021 bytes 100 times --> 704.94 Mbps in 11.05 usec 44: 1024 bytes 100 times --> 716.74 Mbps in 10.90 usec 45: 1027 bytes 100 times --> 699.90 Mbps in 11.20 usec 46: 1533 bytes 100 times --> 1007.83 Mbps in 11.60 usec 47: 1536 bytes 100 times --> 982.29 Mbps in 11.93 usec 48: 1539 bytes 100 times --> 1004.42 Mbps in 11.69 usec 49: 2045 bytes 100 times --> 1266.92 Mbps in 12.31 usec 50: 2048 bytes 100 times --> 1281.26 Mbps in 12.19 usec 51: 2051 bytes 100 times --> 1263.45 Mbps in 12.38 usec 52: 3069 bytes 100 times --> 1755.88 Mbps in 13.34 usec 53: 3072 bytes 100 times --> 1753.65 Mbps in 13.37 usec 54: 3075 bytes 100 times --> 1642.31 Mbps in 14.29 usec 55: 4093 bytes 100 times --> 2037.66 Mbps in 15.32 usec 56: 4096 bytes 100 times --> 2003.21 Mbps in 15.60 usec 57: 4099 bytes 100 times --> 1879.38 Mbps in 16.64 usec 58: 6141 bytes 100 times --> 2587.09 Mbps in 18.11 usec 59: 6144 bytes 100 times --> 2580.51 Mbps in 18.16 usec 60: 6147 bytes 100 times --> 2522.07 Mbps in 18.59 usec 61: 8189 bytes 100 times --> 3100.60 Mbps in 20.15 usec 62: 8192 bytes 100 times --> 3097.89 Mbps in 20.18 usec 63: 8195 bytes 100 times --> 2936.73 Mbps in 21.29 usec 64: 12285 bytes 100 times --> 3683.52 Mbps in 25.44 usec 65: 12288 bytes 100 times --> 3701.14 Mbps in 25.33 usec 66: 12291 bytes 100 times --> 3599.73 Mbps in 26.05 usec 67: 16381 bytes 100 times --> 4173.56 Mbps in 29.94 usec 68: 16384 bytes 100 times --> 4193.22 Mbps in 29.81 usec 69: 16387 bytes 100 times --> 4119.37 Mbps in 30.35 usec 70: 24573 bytes 100 times --> 5032.94 Mbps in 37.25 usec 71: 24576 bytes 100 times --> 5016.05 Mbps in 37.38 usec 72: 24579 bytes 100 times --> 4927.03 Mbps in 38.06 usec 73: 32765 bytes 100 times --> 5483.16 Mbps in 45.59 usec 74: 32768 bytes 100 times --> 5461.50 Mbps in 45.78 usec 75: 32771 bytes 100 times --> 3894.44 Mbps in 64.20 usec 76: 49149 bytes 100 times --> 4648.86 Mbps in 80.66 usec 77: 49152 bytes 100 times --> 4500.72 Mbps in 83.32 usec 78: 49155 bytes 100 times --> 4531.72 Mbps in 82.75 usec 79: 65533 bytes 100 times --> 4860.28 Mbps in 102.87 usec 80: 65536 bytes 100 times --> 5356.76 Mbps in 93.34 usec 81: 65539 bytes 100 times --> 5086.96 Mbps in 98.29 usec 82: 98301 bytes 100 times --> 5769.72 Mbps in 129.99 usec 83: 98304 bytes 100 times --> 5984.92 Mbps in 125.32 usec 84: 98307 bytes 100 times --> 5742.90 Mbps in 130.60 usec 85: 131069 bytes 100 times --> 5914.40 Mbps in 169.07 usec 86: 131072 bytes 100 times --> 5742.51 Mbps in 174.14 usec 87: 131075 bytes 100 times --> 5914.32 Mbps in 169.08 usec 88: 196605 bytes 100 times --> 6001.11 Mbps in 249.95 usec 89: 196608 bytes 100 times --> 6002.64 Mbps in 249.89 usec 90: 196611 bytes 100 times --> 6015.73 Mbps in 249.35 usec 91: 262141 bytes 100 times --> 6533.84 Mbps in 306.09 usec 92: 262144 bytes 100 times --> 6533.92 Mbps in 306.09 usec 93: 262147 bytes 100 times --> 6525.47 Mbps in 306.50 usec 94: 393213 bytes 100 times --> 7156.09 Mbps in 419.22 usec 95: 393216 bytes 100 times --> 7152.31 Mbps in 419.44 usec 96: 393219 bytes 100 times --> 7162.78 Mbps in 418.84 usec 97: 524285 bytes 100 times --> 7522.71 Mbps in 531.72 usec 98: 524288 bytes 100 times --> 7531.33 Mbps in 531.11 usec 99: 524291 bytes 100 times --> 7517.00 Mbps in 532.13 usec 100: 786429 bytes 100 times --> 7908.08 Mbps in 758.72 usec 101: 786432 bytes 100 times --> 7928.12 Mbps in 756.80 usec 102: 786435 bytes 100 times --> 7918.16 Mbps in 757.76 usec 103: 1048573 bytes 100 times --> 8098.13 Mbps in 987.88 usec 104: 1048576 bytes 100 times --> 8150.96 Mbps in 981.48 usec 105: 1048579 bytes 100 times --> 8146.87 Mbps in 981.97 usec 106: 1572861 bytes 100 times --> 8342.93 Mbps in 1438.34 usec 107: 1572864 bytes 100 times --> 8368.29 Mbps in 1433.98 usec 108: 1572867 bytes 100 times --> 8350.37 Mbps in 1437.07 usec 109: 2097149 bytes 100 times --> 8486.90 Mbps in 1885.25 usec 110: 2097152 bytes 100 times --> 8485.61 Mbps in 1885.54 usec 111: 2097155 bytes 100 times --> 8486.30 Mbps in 1885.40 usec 112: 3145725 bytes 100 times --> 8611.60 Mbps in 2786.94 usec 113: 3145728 bytes 100 times --> 8603.65 Mbps in 2789.52 usec 114: 3145731 bytes 100 times --> 8598.94 Mbps in 2791.05 usec 115: 4194307 bytes 100 times --> 8666.03 Mbps in 3692.58 usec

-- Intel Pallas Benchmark
Sockets-MX can also speed up HPC applications in binary format which use TCP/IP.
For the following test the IMB benchmark was compiled and run under MS-MPI.
Please note that also native MPI versions are available for Windows offering increased performance.
#
# List of Benchmarks to run:

# PingPong

#---------------------------------------------------
# Benchmarking PingPong
# ( #processes = 2 )
#---------------------------------------------------
       #bytes #repetitions      t[usec]   Mbytes/sec
            0         1000        14.32         0.00
            1         1000        14.59         0.06
            2         1000        14.78         0.13
            4         1000        14.84         0.26
            8         1000        14.72         0.52
           16         1000        14.79         1.03
           32         1000        14.87         2.05
           64         1000        16.42         3.72
          128         1000        16.72         7.30
          256         1000        17.00        14.36
          512         1000        17.85        27.35
         1024         1000        19.82        49.28
         2048         1000        22.03        88.66
         4096         1000        28.42       137.44
         8192         1000        34.98       223.33
        16384         1000        54.19       288.35
        32768         1000       100.16       312.00
        65536          640       161.82       386.22
       131072          320       237.87       525.49
       262144          160       403.00       620.34
       524288           80       624.68       800.41
      1048576           40      1057.07       946.01
      2097152           20      1954.48      1023.29
      4194304           10      3974.00      1006.54
#=====================================================