| 1 |
|
| 2 |
Platform: AMD Accelerated Parallel Processing |
| 3 |
Device: gfx906 |
| 4 |
Driver version : 3204.0 (HSA1.1,LC) (Linux x64) |
| 5 |
Compute units : 60 |
| 6 |
Clock frequency : 1700 MHz |
| 7 |
|
| 8 |
Global memory bandwidth (GBPS) |
| 9 |
float : 783.04 |
| 10 |
float2 : 741.34 |
| 11 |
float4 : 723.88 |
| 12 |
float8 : 732.36 |
| 13 |
float16 : 679.49 |
| 14 |
|
| 15 |
Single-precision compute (GFLOPS) |
| 16 |
float : 12727.97 |
| 17 |
float2 : 12632.55 |
| 18 |
float4 : 12403.68 |
| 19 |
float8 : 12147.13 |
| 20 |
float16 : 11960.99 |
| 21 |
|
| 22 |
Half-precision compute (GFLOPS) |
| 23 |
half : 6425.83 |
| 24 |
half2 : 24459.28 |
| 25 |
half4 : 24278.00 |
| 26 |
half8 : 23921.18 |
| 27 |
half16 : 23455.81 |
| 28 |
|
| 29 |
Double-precision compute (GFLOPS) |
| 30 |
double : 6206.76 |
| 31 |
double2 : 6176.21 |
| 32 |
double4 : 6135.32 |
| 33 |
double8 : 6107.36 |
| 34 |
double16 : 5924.13 |
| 35 |
|
| 36 |
Integer compute (GIOPS) |
| 37 |
int : 4186.51 |
| 38 |
int2 : 4019.41 |
| 39 |
int4 : 4003.08 |
| 40 |
int8 : 4029.69 |
| 41 |
int16 : 3976.25 |
| 42 |
|
| 43 |
Integer compute Fast 24bit (GIOPS) |
| 44 |
int : 11493.50 |
| 45 |
int2 : 10816.38 |
| 46 |
int4 : 10109.61 |
| 47 |
int8 : 10421.03 |
| 48 |
int16 : 10354.31 |
| 49 |
|
| 50 |
Transfer bandwidth (GBPS) |
| 51 |
enqueueWriteBuffer : 16.91 |
| 52 |
enqueueReadBuffer : 16.85 |
| 53 |
enqueueWriteBuffer non-blocking : 16.91 |
| 54 |
enqueueReadBuffer non-blocking : 16.83 |
| 55 |
enqueueMapBuffer(for read) : 128591.83 |
| 56 |
memcpy from mapped ptr : 16.77 |
| 57 |
enqueueUnmap(after write) : 238609.30 |
| 58 |
memcpy to mapped ptr : 16.91 |
| 59 |
|
| 60 |
Kernel launch latency : 14.06 us |
| 61 |
|