Posting my 4070 results!

[oscarbg]

EDIT: posting here, but would be nice if some thread or wiki can be added with people posting results.. EDIT 2: tested on Linux also on 525.47.35 driver same results would be nice to compare to RDNA3 results (specially with fp16-matrix included)!

vkpeak-20230730-windows (using driver 532.32)

device       = NVIDIA GeForce RTX 4070

fp32-scalar  = 16493.71 GFLOPS
fp32-vec4    = 21843.69 GFLOPS

fp16-scalar  = 16476.93 GFLOPS
fp16-vec4    = 32665.59 GFLOPS
fp16-matrix  = 132483.00 GFLOPS

fp64-scalar  = 520.07 GFLOPS
fp64-vec4    = 521.40 GFLOPS

int32-scalar = 16524.10 GIOPS
int32-vec4   = 16448.31 GIOPS

int16-scalar = 10989.81 GIOPS
int16-vec4   = 14632.48 GIOPS

[allanmac]

Ditto!

device       = Intel(R) Arc(tm) A770 Graphics (DG2)

fp32-scalar  = 19196.22 GFLOPS
fp32-vec4    = 14152.80 GFLOPS

fp16-scalar  = 21674.14 GFLOPS
fp16-vec4    = 25906.79 GFLOPS
fp16-matrix  = 0.00 GFLOPS

fp64-scalar  = 0.00 GFLOPS
fp64-vec4    = 0.00 GFLOPS

int32-scalar = 4715.93 GIOPS
int32-vec4   = 4864.83 GIOPS

int16-scalar = 9187.25 GIOPS
int16-vec4   = 9653.79 GIOPS

[jeqka24]

@allanmac, can you post updated results (due to the new intel driver capabilities mentioned in https://github.com/nihui/vkpeak/issues/11)?