Debian Codegraph @

results/NVIDIA_CUDA/Quadro_GV100.log

raw
1
2 Platform: NVIDIA CUDA
3 Device: Quadro GV100
4 Driver version : 455.23.05 (Linux x64)
5 Compute units : 80
6 Clock frequency : 1627 MHz
7
8 Global memory bandwidth (GBPS)
9 float : 554.70
10 float2 : 575.69
11 float4 : 258.14
12 float8 : 537.39
13 float16 : 552.97
14
15 Single-precision compute (GFLOPS)
16 float : 6216.30
17 float2 : 11449.58
18 float4 : 14290.00
19 float8 : 7261.11
20 float16 : 7262.70
21
22 No half precision support! Skipped
23
24 Double-precision compute (GFLOPS)
25 double : 7212.01
26 double2 : 7191.88
27 double4 : 7168.91
28 double8 : 4505.54
29 double16 : 3124.18
30
31 Integer compute (GIOPS)
32 int : 6219.71
33 int2 : 9340.54
34 int4 : 14371.73
35 int8 : 14373.41
36 int16 : 14342.93
37
38 Integer compute Fast 24bit (GIOPS)
39 int : 9037.88
40 int2 : 6214.77
41 int4 : 6216.47
42 int8 : 9337.87
43 int16 : 14352.44
44
45 Transfer bandwidth (GBPS)
46 enqueueWriteBuffer : 9.75
47 enqueueReadBuffer : 12.11
48 enqueueWriteBuffer non-blocking : 9.32
49 enqueueReadBuffer non-blocking : 11.31
50 enqueueMapBuffer(for read) : 11.25
51 memcpy from mapped ptr : 9.54
52 enqueueUnmap(after write) : 11.26
53 memcpy to mapped ptr : 9.84
54
55 Kernel launch latency : 19.29 us
56