I plan to buy a new machine for running my algorithm implemented in GPU under CUDA. I currently have a GTX 460 graphics card, and may get a GTX 590 a few months later.

I am not sure which CPU, mainboard, memory, case, power, etc. I should use, so as to make the machine run well.

Could you give me some advice?

link|improve this question
feedback

migrated from stackoverflow.com Sep 13 '11 at 12:42

This question came from our site for professional and enthusiast programmers.

1 Answer

up vote 0 down vote accepted

Here are some advices based on my experience (I've tested several configurations for our university HPC center):

There is no difference which CPU to use (Intel or AMD), while it have at least (N + 1) cores, where N is numbers of GPUs you have.

Memory -- not less than your GPUs have.

Motherboard: must have PCI-Express 2.0 x16 slot for each card. Bus bandwidth is often limiting factor. No SLI needed.

Case and power --- just general considerarions (for reference, typical GPU consumes ~200W).

Of course, all common rules apply, I've mentioned only GPU-specific ones.

link|improve this answer
feedback

Your Answer

 
or
required, but never shown

Not the answer you're looking for? Browse other questions tagged or ask your own question.