News for hackers. Stuff that matters.

A comment on Which GPUs to get for deep learning. Parent comment

Re:Which GPUs to get for deep learning by tbenst on Wednesday September 09, 2020 @02:20AM (#24417834)
> NVIDIA GPUs offer up to 8x more half precision arithmetic throughput when compared to single-precision, thus speeding up math-limited layers

Right but I’ve benchmarked the best case scenario, ie a large GEMM call in C++, and still not seen anywhere close to 8x. I’ve never seen a code example, no matter how limited, showing a 8x speed up.

[ Parent | on HN ]
Re:Which GPUs to get for deep learning by llukas on Wednesday September 09, 2020 @03:14AM (#24418204)
https://developer.nvidia.com/blog/programming-tensor-cores-c...

See cuBLAS mixed-precision GEMM.

[ Parent | on HN ]