| 1 |
|
| 2 |
Platform: AMD Accelerated Parallel Processing |
| 3 |
Device: Ellesmere (RX 570) |
| 4 |
Driver version : 2639.3 (Linux x64) |
| 5 |
Compute units : 32 |
| 6 |
Clock frequency : 1268 MHz |
| 7 |
|
| 8 |
Global memory bandwidth (GBPS) |
| 9 |
float : 182.67 |
| 10 |
float2 : 191.18 |
| 11 |
float4 : 193.05 |
| 12 |
float8 : 163.17 |
| 13 |
float16 : 139.53 |
| 14 |
|
| 15 |
Single-precision compute (GFLOPS) |
| 16 |
float : 5133.96 |
| 17 |
float2 : 5129.58 |
| 18 |
float4 : 5109.60 |
| 19 |
float8 : 5087.33 |
| 20 |
float16 : 5018.78 |
| 21 |
|
| 22 |
Half-precision compute (GFLOPS) |
| 23 |
half : 5119.05 |
| 24 |
half2 : 5115.60 |
| 25 |
half4 : 5104.28 |
| 26 |
half8 : 5089.02 |
| 27 |
half16 : 5045.35 |
| 28 |
|
| 29 |
Double-precision compute (GFLOPS) |
| 30 |
double : 323.55 |
| 31 |
double2 : 323.39 |
| 32 |
double4 : 322.84 |
| 33 |
double8 : 321.66 |
| 34 |
double16 : 320.53 |
| 35 |
|
| 36 |
Integer compute (GIOPS) |
| 37 |
int : 1032.21 |
| 38 |
int2 : 1032.00 |
| 39 |
int4 : 1031.56 |
| 40 |
int8 : 1030.70 |
| 41 |
int16 : 1029.31 |
| 42 |
|
| 43 |
Transfer bandwidth (GBPS) |
| 44 |
enqueueWriteBuffer : 38.47 |
| 45 |
enqueueReadBuffer : 14.87 |
| 46 |
enqueueMapBuffer(for read) : 220429.70 |
| 47 |
memcpy from mapped ptr : 15.33 |
| 48 |
enqueueUnmap(after write) : 604140.69 |
| 49 |
memcpy to mapped ptr : 15.09 |
| 50 |
|
| 51 |
Kernel launch latency : 42.83 us |
| 52 |
|