I ran the deviceQuery program on my RHEL 4 32-bit linux box. I get the following result:

There is 1 device supporting CUDA

Device 0: “Quadro FX 4600”

Maximum number of threads per block: 512
Maximum sizes of each dimension of a block: 512 x 512 x 64

The above two lines seem to be contradictory. My impression was the block dimensions should not be more than 512, since the number of threads is 512. Am I correct ?


Did you start X before running this test?

Which display driver version are you using?

Why are they contradictory? 512x1 = 512 so the maximum num threads in each dimension is 512. You can flexibly choose any block config as long as the sum of the threads is <512.

What is a bit boring however is that the max config isn’t 512^3. Discrimination of the 3rd dimension :fear: … the world is (almost) a disc :fear:


Just to clarify:
Max Nx=512
Max Ny=512
Max Nz=64

with the limitation that Nx *Ny *Nx <512

For example, you could run a block with dimension 888 or 51211.
If you try a configuration 11512, you will get a failure.

That makes sense… Thanks for the replies…