| 1 |
|
| 2 |
Platform: NVIDIA CUDA |
| 3 |
Device: GeForce GTX 1080 |
| 4 |
Driver version : 367.27 (Linux x64) |
| 5 |
Compute units : 20 |
| 6 |
Clock frequency : 1733 MHz |
| 7 |
|
| 8 |
Global memory bandwidth (GBPS) |
| 9 |
float : 224.25 |
| 10 |
float2 : 227.78 |
| 11 |
float4 : 236.81 |
| 12 |
float8 : 216.52 |
| 13 |
float16 : 179.27 |
| 14 |
|
| 15 |
Single-precision compute (GFLOPS) |
| 16 |
float : 8549.33 |
| 17 |
float2 : 9216.67 |
| 18 |
float4 : 9262.55 |
| 19 |
float8 : 9164.55 |
| 20 |
float16 : 9158.85 |
| 21 |
|
| 22 |
Double-precision compute (GFLOPS) |
| 23 |
double : 303.79 |
| 24 |
double2 : 303.89 |
| 25 |
double4 : 303.46 |
| 26 |
double8 : 302.27 |
| 27 |
double16 : 299.86 |
| 28 |
|
| 29 |
Integer compute (GIOPS) |
| 30 |
int : 2458.08 |
| 31 |
int2 : 2620.93 |
| 32 |
int4 : 2582.49 |
| 33 |
int8 : 2621.57 |
| 34 |
int16 : 2602.94 |
| 35 |
|
| 36 |
Transfer bandwidth (GBPS) |
| 37 |
enqueueWriteBuffer : 1.77 |
| 38 |
enqueueReadBuffer : 12.85 |
| 39 |
enqueueMapBuffer(for read) : 10.95 |
| 40 |
memcpy from mapped ptr : 10.30 |
| 41 |
enqueueUnmap(after write) : 12.15 |
| 42 |
memcpy to mapped ptr : 10.21 |
| 43 |
|
| 44 |
Kernel launch latency : 4.35 us |
| 45 |
|