I am just retired and am about to set up a small consulting business. I am looking to upgrade my home computer to an engineering work station class device. I do laser and physics simulations in LabView and am looking for compatible CPUs and GPUs to run repetitive 2-d FFTs in my codes. What open software is out there for using the GPU cards? etc. I would like to hear your views on what hardware and software would make a device capable of fast simulations.
Any modern processor will work great. If your code has potential for high parallelization, high core counts help. (This currently means AMD. Have a look at my benchmark page.)
Properly written LabVIEW code works well without any special toolkits. How big are your 2D arrays? What is your definition of "fast"? For the same algorithm better hardware can give you maybe a factor of two. Better code can give you orders of magnitude better performance over poorly written code. So focus on code first! We can help. Be careful, parallelization has some overhead and might slow you down. Always try both.