I’m getting different performance between these two tests.
I am getting 2GB/s using the CUDA test and 3GB/s on the ocl test. This is using pinned memory and the rest of the default settings.
My setup is a bit sketchy as I have a GT460M (mobile GPU) attached to a server class Intel workstation, but I don’t believe that should really have an impact of which language I use.
BTW, I’m expecting ~6GB/s and only getting 3 even in the opencl test…
Maybe someone else has experienced the same?