| 1 |
|
| 2 |
Platform: NVIDIA CUDA |
| 3 |
Device: GeForce GT 650M |
| 4 |
Driver version: 331.82 (Win32) |
| 5 |
|
| 6 |
Global memory bandwidth (GBPS) |
| 7 |
float : 25.16 |
| 8 |
float2 : 26.84 |
| 9 |
float4 : 26.84 |
| 10 |
float8 : 13.42 |
| 11 |
float16 : 12.39 |
| 12 |
|
| 13 |
Single-precision compute (GFLOPS) |
| 14 |
float : 486.09 |
| 15 |
float2 : 644.06 |
| 16 |
float4 : 613.40 |
| 17 |
float8 : 613.40 |
| 18 |
float16 : 613.40 |
| 19 |
|
| 20 |
Double-precision compute (GFLOPS) |
| 21 |
double : 30.38 |
| 22 |
double2 : 30.34 |
| 23 |
double4 : 30.27 |
| 24 |
double8 : 30.13 |
| 25 |
double16 : 29.92 |
| 26 |
|
| 27 |
Transfer bandwidth (GBPS) |
| 28 |
enqueueWriteBuffer : 4.36 |
| 29 |
enqueueReadBuffer : 4.29 |
| 30 |
enqueueMapBuffer(for read) : 2.95 |
| 31 |
memcpy from mapped ptr : 8.01 |
| 32 |
enqueueUnmap(after write) : 4.33 |
| 33 |
memcpy to mapped ptr : 8.26 |
| 34 |
|
| 35 |
Kernel launch latency : 25.48 us |
| 36 |
|