| 1 |
|
| 2 |
Platform: NVIDIA CUDA |
| 3 |
Device: Tesla V100-PCIE-32GB |
| 4 |
Driver version : 455.23.05 (Linux x64) |
| 5 |
Compute units : 80 |
| 6 |
Clock frequency : 1380 MHz |
| 7 |
|
| 8 |
Global memory bandwidth (GBPS) |
| 9 |
float : 716.38 |
| 10 |
float2 : 765.67 |
| 11 |
float4 : 810.35 |
| 12 |
float8 : 723.85 |
| 13 |
float16 : 750.17 |
| 14 |
|
| 15 |
Single-precision compute (GFLOPS) |
| 16 |
float : 14098.15 |
| 17 |
float2 : 14135.97 |
| 18 |
float4 : 14095.57 |
| 19 |
float8 : 14049.00 |
| 20 |
float16 : 13934.45 |
| 21 |
|
| 22 |
No half precision support! Skipped |
| 23 |
|
| 24 |
Double-precision compute (GFLOPS) |
| 25 |
double : 7075.81 |
| 26 |
double2 : 7065.56 |
| 27 |
double4 : 7046.01 |
| 28 |
double8 : 7013.68 |
| 29 |
double16 : 6951.51 |
| 30 |
|
| 31 |
Integer compute (GIOPS) |
| 32 |
int : 14069.94 |
| 33 |
int2 : 14118.04 |
| 34 |
int4 : 14121.60 |
| 35 |
int8 : 14124.16 |
| 36 |
int16 : 14099.04 |
| 37 |
|
| 38 |
Integer compute Fast 24bit (GIOPS) |
| 39 |
int : 14077.32 |
| 40 |
int2 : 14119.12 |
| 41 |
int4 : 14122.14 |
| 42 |
int8 : 14113.63 |
| 43 |
int16 : 14104.60 |
| 44 |
|
| 45 |
Transfer bandwidth (GBPS) |
| 46 |
enqueueWriteBuffer : 12.06 |
| 47 |
enqueueReadBuffer : 10.64 |
| 48 |
enqueueWriteBuffer non-blocking : 10.72 |
| 49 |
enqueueReadBuffer non-blocking : 8.13 |
| 50 |
enqueueMapBuffer(for read) : 10.25 |
| 51 |
memcpy from mapped ptr : 17.55 |
| 52 |
enqueueUnmap(after write) : 12.59 |
| 53 |
memcpy to mapped ptr : 18.20 |
| 54 |
|
| 55 |
Kernel launch latency : 7.88 us |
| 56 |
|