| 1 |
Platform: NVIDIA CUDA |
| 2 |
Device: Quadro K620 |
| 3 |
Driver version : 515.76 (Linux x64) |
| 4 |
Compute units : 3 |
| 5 |
Clock frequency : 1124 MHz |
| 6 |
|
| 7 |
Global memory bandwidth (GBPS) |
| 8 |
float : 25.41 |
| 9 |
float2 : 26.21 |
| 10 |
float4 : 26.69 |
| 11 |
float8 : 25.73 |
| 12 |
float16 : 22.42 |
| 13 |
|
| 14 |
Single-precision compute (GFLOPS) |
| 15 |
float : 569.12 |
| 16 |
float2 : 839.43 |
| 17 |
float4 : 856.16 |
| 18 |
float8 : 851.06 |
| 19 |
float16 : 848.14 |
| 20 |
|
| 21 |
No half precision support! Skipped |
| 22 |
|
| 23 |
Double-precision compute (GFLOPS) |
| 24 |
double : 27.47 |
| 25 |
double2 : 27.48 |
| 26 |
double4 : 27.42 |
| 27 |
double8 : 27.32 |
| 28 |
double16 : 27.11 |
| 29 |
|
| 30 |
Integer compute (GIOPS) |
| 31 |
int : 258.66 |
| 32 |
int2 : 287.30 |
| 33 |
int4 : 289.72 |
| 34 |
int8 : 275.19 |
| 35 |
int16 : 264.43 |
| 36 |
|
| 37 |
Integer compute Fast 24bit (GIOPS) |
| 38 |
int : 258.66 |
| 39 |
int2 : 287.06 |
| 40 |
int4 : 289.57 |
| 41 |
int8 : 287.89 |
| 42 |
int16 : 286.48 |
| 43 |
|
| 44 |
Transfer bandwidth (GBPS) |
| 45 |
enqueueWriteBuffer : 5.16 |
| 46 |
enqueueReadBuffer : 5.13 |
| 47 |
enqueueWriteBuffer non-blocking : 2.17 |
| 48 |
enqueueReadBuffer non-blocking : 2.73 |
| 49 |
enqueueMapBuffer(for read) : 5.43 |
| 50 |
memcpy from mapped ptr : 4.48 |
| 51 |
enqueueUnmap(after write) : 6.14 |
| 52 |
memcpy to mapped ptr : 4.49 |
| 53 |
|
| 54 |
Kernel launch latency : 6.87 us |