LoginSignup
2
1

More than 3 years have passed since last update.

A6000, RTX 3090 の倍精度性能のメモ

Last updated at Posted at 2021-04-13

mixbench opencl で測定しました.

A6000 は 165W に powerlimit しています.

おおむね 0.55 Tflops でした(ECC off). 3090 も大体同等の値でした.
従って両方とも fp64 性能は fp32 の 1/64 になります.

Ampere 世代で倍精度を求めるのであれば A100(10 TF), A30(5 TF)にしましょう.

------------------------ Device specifications ------------------------
Platform:            NVIDIA CUDA
Device:              RTX A6000/NVIDIA Corporation
Driver version:      460.32.03
Address bits:        64
GPU clock rate:      1800 MHz
Total global mem:    48677 MB
Max allowed buffer:  12169 MB
OpenCL version:      OpenCL 1.2 CUDA
Total CUs:           84
-----------------------------------------------------------------------
Buffer size:            64MB
Workgroup size:         256
Workitem stride:        NDRange
Buffer allocation:      Device allocated
Timer:                  CL event based
Warning:                Half precision computations are not supported
Loading kernel source file...
Precompilation of kernels... [>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>]
----------------------------------------------------------------------------- CSV data -----------------------------------------------------------------------------
Experiment ID, Single Precision ops,,,,              Double precision ops,,,,              Half precision ops,,,,                Integer operations,,,
Compute iters, Flops/byte, ex.time,  GFLOPS, GB/sec, Flops/byte, ex.time,  GFLOPS, GB/sec, Flops/byte, ex.time,  GFLOPS, GB/sec, Iops/byte, ex.time,   GIOPS, GB/sec
            0,      0.000,   10.86,    0.00, 790.86,      0.000,   24.21,    0.00, 709.73,      0.000,    0.00,    -nan,    inf,     0.000,   10.14,    0.00, 847.08
            1,      0.129,    9.25,  116.03, 899.25,      0.065,   23.27,   46.15, 715.36,      0.258,    0.00,     inf,    inf,     0.129,    8.80,  122.00, 945.49
            2,      0.267,    9.01,  238.34, 893.78,      0.133,   22.21,   96.69, 725.16,      0.533,    0.00,     inf,    inf,     0.267,    8.86,  242.45, 909.17
            3,      0.414,    7.72,  417.26,1008.38,      0.207,   21.65,  148.79, 719.16,      0.828,    0.00,     inf,    inf,     0.414,    8.28,  389.03, 940.17
            4,      0.571,    7.87,  545.64, 954.86,      0.286,   20.60,  208.49, 729.70,      1.143,    0.00,     inf,    inf,     0.571,    7.79,  551.59, 965.29
            5,      0.741,    6.71,  799.59,1079.44,      0.370,   23.47,  228.72, 617.54,      1.481,    0.00,     inf,    inf,     0.741,    6.71,  800.56,1080.76
            6,      0.923,    8.08,  797.70, 864.17,      0.462,   19.96,  322.75, 699.30,      1.846,    0.00,     inf,    inf,     0.923,    6.90,  933.74,1011.55
            7,      1.120,    5.68, 1323.72,1181.89,      0.560,   18.34,  409.92, 732.00,      2.240,    0.00,     inf,    inf,     1.120,    5.68, 1322.53,1180.83
            8,      1.333,    5.29, 1623.81,1217.86,      0.667,   18.19,  472.20, 708.30,      2.667,    0.00,     inf,    inf,     1.333,    5.43, 1580.97,1185.72
            9,      1.565,    4.72, 2047.56,1308.16,      0.783,   19.13,  505.15, 645.47,      3.130,    0.00,     inf,    inf,     1.565,    4.80, 2011.76,1285.29
           10,      1.818,    4.23, 2535.85,1394.72,      0.909,   20.95,  512.58, 563.83,      3.636,    0.00,     inf,    inf,     1.818,    4.34, 2474.81,1361.14
           11,      2.095,    3.92, 3015.51,1439.22,      1.048,   22.34,  528.66, 504.63,      4.190,    0.00,     inf,    inf,     2.095,    3.80, 3111.50,1485.03
           12,      2.400,    3.73, 3456.84,1440.35,      1.200,   24.44,  527.23, 439.36,      4.800,    0.00,     inf,    inf,     2.400,    3.60, 3583.85,1493.27
           13,      2.737,    3.21, 4344.01,1587.23,      1.368,   25.93,  538.26, 393.35,      5.474,    0.00,     inf,    inf,     2.737,    3.40, 4108.46,1501.17
           14,      3.111,    2.94, 5113.28,1643.55,      1.556,   27.60,  544.70, 350.16,      6.222,    0.00,     inf,    inf,     3.111,    3.18, 4727.88,1519.68
           15,      3.529,    2.65, 6080.02,1722.67,      1.765,   29.12,  553.09, 313.42,      7.059,    0.00,     inf,    inf,     3.529,    3.26, 4941.45,1400.08
           16,      4.000,    5.49, 3127.74, 781.94,      2.000,   30.98,  554.56, 277.28,      8.000,    0.00,     inf,    inf,     4.000,    5.60, 3068.25, 767.06
           17,      4.533,    4.48, 4075.40, 898.98,      2.267,   32.82,  556.20, 245.38,      9.067,    0.00,     inf,    inf,     4.533,    4.62, 3948.15, 870.92
           18,      5.143,    4.45, 4340.93, 844.07,      2.571,   40.83,  473.40, 184.10,     10.286,    0.00,     inf,    inf,     5.143,    4.48, 4313.16, 838.67
           19,      5.846,    4.24, 4815.76, 823.75,      2.923,   36.69,  556.06, 190.23,     11.692,    0.00,     inf,    inf,     5.846,    3.97, 5137.43, 878.77
           20,      6.667,    3.72, 5774.10, 866.11,      3.333,   38.61,  556.24, 166.87,     13.333,    0.00,     inf,    inf,     6.667,    3.87, 5546.55, 831.98
           21,      7.636,    3.12, 7233.93, 947.30,      3.818,   40.52,  556.47, 145.74,     15.273,    0.00,     inf,    inf,     7.636,    3.41, 6610.66, 865.68
           22,      8.800,    3.70, 6377.85, 724.76,      4.400,   45.50,  519.14, 117.99,     17.600,    0.00,     inf,    inf,     8.800,    5.17, 4573.43, 519.71
           23,     10.222,    2.91, 8474.09, 828.99,      5.111,   46.15,  535.17, 104.71,     20.444,    0.00,     inf,    inf,    10.222,    3.10, 7972.64, 779.93
           24,     12.000,    3.08, 8363.52, 696.96,      6.000,   46.24,  557.26,  92.88,     24.000,    0.00,     inf,    inf,    12.000,    3.13, 8226.81, 685.57
           25,     14.286,    2.43,11051.60, 773.61,      7.143,   48.17,  557.23,  78.01,     28.571,    0.00,     inf,    inf,    14.286,    2.48,10841.36, 758.89
           26,     17.333,    2.43,11474.32, 661.98,      8.667,   50.11,  557.11,  64.28,     34.667,    0.00,     inf,    inf,    17.333,    2.45,11397.42, 657.54
           27,     21.600,    1.73,16762.32, 776.03,     10.800,   52.03,  557.16,  51.59,     43.200,    0.00,     inf,    inf,    21.600,    1.73,16782.19, 776.95
           28,     28.000,    1.78,16922.26, 604.37,     14.000,   53.91,  557.69,  39.83,     56.000,    0.00,     inf,    inf,    28.000,    1.76,17129.60, 611.77
           29,     38.667,    1.00,31092.74, 804.12,     19.333,   55.82,  557.86,  28.85,     77.333,    0.00,     inf,    inf,    38.667,    1.80,17297.33, 447.34
           30,     60.000,    1.00,32160.80, 536.01,     30.000,   57.72,  558.09,  18.60,    120.000,    0.00,     inf,    inf,    60.000,    1.85,17418.81, 290.31
           31,    124.000,    1.00,33203.12, 267.77,     62.000,   59.65,  558.04,   9.00,    248.000,    0.00,     inf,    inf,   124.000,    1.89,17656.63, 142.39
           32,        inf,    0.99,34592.20,   0.00,        inf,   61.57,  558.09,   0.00,        inf,    0.00,     inf,   -nan,       inf,    1.93,17800.76,   0.00
--------------------------------------------------------------------------------------------------------------------------------------------------------------------



2
1
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
2
1