currently i have tried to get my fft running and it works fine until i reach 512512512 cubes. then some error occurs : cufft.cu error cufft_invalid_value. but if i try the same procedure with 256256256 there is no problem. i am using a Quadro FX 5600 so the memory should not be the problem, i think. is there some other restriction in the system?
512^3 is 128 million elements, so just for the float2 (complex) source data that’s 1GB. That doesn’t leave a lot of room for anything else.
CUFFT probably uses some temporary storage also.
mmm then i have to try something else. ty