| 1 |
Platform: NVIDIA CUDA |
| 2 |
Device: TITAN V |
| 3 |
Driver version : 430.14 (Linux x64) |
| 4 |
Compute units : 80 |
| 5 |
Clock frequency : 1455 MHz |
| 6 |
|
| 7 |
Global memory bandwidth (GBPS) |
| 8 |
float : 561.37 |
| 9 |
float2 : 591.37 |
| 10 |
float4 : 607.71 |
| 11 |
float8 : 516.49 |
| 12 |
float16 : 466.27 |
| 13 |
|
| 14 |
Single-precision compute (GFLOPS) |
| 15 |
float : 13651.32 |
| 16 |
float2 : 13688.23 |
| 17 |
float4 : 13648.46 |
| 18 |
float8 : 13606.27 |
| 19 |
float16 : 13502.08 |
| 20 |
|
| 21 |
No half precision support! Skipped |
| 22 |
|
| 23 |
Double-precision compute (GFLOPS) |
| 24 |
double : 6858.92 |
| 25 |
double2 : 6846.90 |
| 26 |
double4 : 6822.64 |
| 27 |
double8 : 6797.12 |
| 28 |
double16 : 6737.34 |
| 29 |
|
| 30 |
Integer compute (GIOPS) |
| 31 |
int : 13622.13 |
| 32 |
int2 : 13661.56 |
| 33 |
int4 : 13666.12 |
| 34 |
int8 : 13663.23 |
| 35 |
int16 : 13640.81 |
| 36 |
|
| 37 |
Integer compute Fast 24bit (GIOPS) |
| 38 |
int : 13622.35 |
| 39 |
int2 : 13662.14 |
| 40 |
int4 : 13666.63 |
| 41 |
int8 : 13658.38 |
| 42 |
int16 : 13647.09 |
| 43 |
|
| 44 |
Transfer bandwidth (GBPS) |
| 45 |
enqueueWriteBuffer : 6.09 |
| 46 |
enqueueReadBuffer : 6.45 |
| 47 |
enqueueWriteBuffer non-blocking : 4.58 |
| 48 |
enqueueReadBuffer non-blocking : 4.93 |
| 49 |
enqueueMapBuffer(for read) : 6.05 |
| 50 |
memcpy from mapped ptr : 9.09 |
| 51 |
enqueueUnmap(after write) : 6.26 |
| 52 |
memcpy to mapped ptr : 9.42 |
| 53 |
|
| 54 |
Kernel launch latency : 6.51 us |