So last summer I worked on a research project concerned with mathematical computation on GPU’s using CUDA. This summer, I’ve been invited back to perform similar work, and I’m excited about our shiny new equipment budget.
I’ve been tasked with purchasing a new graphic card to augment our existing machine, which has a single GTX 295 card. This machine has a second PCIe 2.0 16x slot, so I’d like to fit something new and cool. Sadly, I’ve fallen out of the loop in terms of nVidia’s offerings. Right now, I’m looking at a 480 Fermi based card.
Next up, I’d like to ask about the changes in the SDK. What’s changes have been introduced since September of last year? How hard is it to migrate code from CUDA to OpenCL? How feasible would it be to rework my current codebase as OpenCL for portability?
Thanks for any suggestions!