I apologize for the newbie question, but I’m trying to get into GPU computing and need some help!
I have a Dell XPS system (Intel Core i7-920, 64-bit, 8GB RAM) and am thinking getting a CUDA GPU to speed up my Matlab scripts. Which of the CUDA-enabled cards (CUDA GPUs - Compute Capability | NVIDIA Developer) will work on my system?
Also, after I figure out which ones work, what recommendations do you have in picking a good-valued card at a reasonable price (on a $300-$600 budget). Any help would be appreciated.
I have a very similar Core i7 machine next to my desk. The Core i7 platform is excellent for CUDA, very high transfer rates to/from the card, etc. I assume this type of machine has a PCI-Express x16 slot, so the only question is what the power supply can handle. Look that up to make sure.
I’m very much enjoying the value-for-money of the newish GTX-275 card, and the GTX-285 is worth the purchase if you have the cash.
If you want to get into more complicated CUDA programming, the “sandwich” cards like the GTX-295 are also pretty cool - you get two devices in one card, and all the fun / headaches of utilising it completely. It’s a good practice environment for multi-GPU programming. Not standard fare though.
Hi, thanks! What’s the difference between the BFG, EVGA, PNY, XFX variants? I also see that the 275 has memory options of 896MB and 1792MB. I’m betting that the 1792 would be significantly better and worth the $50 extra. Would you agree?
Hmm, it depends on the problem really. In my work, when I go over 300MB of device memory, I’ve broken something … but that’s because that code also has to run on the “little” cards. If you have the $50, I think I can recommend you take the big one, UNLESS it’s got slower ram, then take the smaller and quicker card.