4 ms·
This only really holds true if you are running OpenCL 1.2 under CUDA. If you run native OpneCL 2.0 on NVIDIA GPUs you aren't getting a major penalty, there wil
by dogma1138 10y ago
This only really holds true if you are running OpenCL 1.2 under CUDA.
If you run native OpneCL 2.0 on NVIDIA GPUs you aren't getting a major penalty, there will still be things that CUDA is going to be faster at but that has nothing to do with intentional gimping.