Hardware recommendations

Hi. I’m a graduate student in Electrical Engineering, specifically Computer Engineering, and I’m interested in getting started with CUDA. I’ve taken some Computer Vision classes and used openCV in some projects and will probably be using CUDA for some of those applications. I have a couple hundred dollars in my budget to get a GeForce card. My machine has a 400W power supply which lead me to the GTX 550i or GTS 450.

When trying to narrow down the cards should I look for more cores or more memory? Any recommendations on hardware?

Which hardware you buy (optimized for memory or speed) is going to depend on what you want to run. If it’s your first time developing with CUDA though, I’d say go with something with more memory – perhaps the 2GB GTX 560 Ti or the 2.5GB GTX 570 Ti. Even with a “lower-end” card, you should still see some fantastic speedups, but you’ll want to make sure that any data you’re working with fits within the memory of the card – transferring data back and forth between the host because you don’t have enough memory on the card is likely to hurt your performance figures more than just having fewer cores.