| 1 |
Platform: NVIDIA CUDA |
| 2 |
Device: NVIDIA GeForce GTX 1660 Ti |
| 3 |
Driver version : 565.57.01 (Linux x64) |
| 4 |
Compute units : 24 |
| 5 |
Clock frequency : 1590 MHz |
| 6 |
|
| 7 |
Global memory bandwidth (GBPS) |
| 8 |
float : 235.92 |
| 9 |
float2 : 247.28 |
| 10 |
float4 : 260.64 |
| 11 |
float8 : 254.10 |
| 12 |
float16 : 217.35 |
| 13 |
|
| 14 |
Single-precision compute (GFLOPS) |
| 15 |
float : 5692.43 |
| 16 |
float2 : 5705.85 |
| 17 |
float4 : 5697.71 |
| 18 |
float8 : 5497.52 |
| 19 |
float16 : 4822.71 |
| 20 |
|
| 21 |
No half precision support! Skipped |
| 22 |
|
| 23 |
Double-precision compute (GFLOPS) |
| 24 |
double : 166.56 |
| 25 |
double2 : 169.71 |
| 26 |
double4 : 151.43 |
| 27 |
double8 : 152.88 |
| 28 |
double16 : 163.43 |
| 29 |
|
| 30 |
Integer compute (GIOPS) |
| 31 |
int : 5009.23 |
| 32 |
int2 : 5025.67 |
| 33 |
int4 : 4511.78 |
| 34 |
int8 : 4535.21 |
| 35 |
int16 : 4828.46 |
| 36 |
|
| 37 |
Integer compute Fast 24bit (GIOPS) |
| 38 |
int : 5030.41 |
| 39 |
int2 : 5000.83 |
| 40 |
int4 : 5002.84 |
| 41 |
int8 : 4461.20 |
| 42 |
int16 : 4415.56 |
| 43 |
|
| 44 |
Integer char (8bit) compute (GIOPS) |
| 45 |
char : 4137.69 |
| 46 |
char2 : 4238.38 |
| 47 |
char4 : 4174.55 |
| 48 |
char8 : 4234.00 |
| 49 |
char16 : 3432.68 |
| 50 |
|
| 51 |
Integer short (16bit) compute (GIOPS) |
| 52 |
short : 4185.20 |
| 53 |
short2 : 4014.07 |
| 54 |
short4 : 4125.94 |
| 55 |
short8 : 3622.42 |
| 56 |
short16 : 3496.44 |
| 57 |
|
| 58 |
Transfer bandwidth (GBPS) |
| 59 |
enqueueWriteBuffer : 6.85 |
| 60 |
enqueueReadBuffer : 6.92 |
| 61 |
enqueueWriteBuffer non-blocking : 6.14 |
| 62 |
enqueueReadBuffer non-blocking : 6.08 |
| 63 |
enqueueMapBuffer(for read) : 9.77 |
| 64 |
memcpy from mapped ptr : 11.68 |
| 65 |
enqueueUnmap(after write) : 12.33 |
| 66 |
memcpy to mapped ptr : 11.99 |
| 67 |
|
| 68 |
Kernel launch latency : 4.14 us |