learning parallel computing by CUDA

[kevin_24]

should I prefer Tesla or quadro graphics card for learning CUDA parallel computing ?

 

[kanewolf]

Any new NVIDIA graphics card can support CUDA. Get a 7xx or newer card. Much cheaper than a Quadro. Tesla cards are specialized and are for "production" use rather than learning.

 

[itmoba]

Both Quadro and Tesla are overkill for someone learning, though if you can get them for free then that's another story entirely. If you don't know how to use threads yet, then, I strongly suggest that you begin by learning pthreads (if Linux/Unix/Unix-variants) or threads (Windows). Afterwards, you can ease yourself into CUDA by learning a little bit of OpenCL (EasyCL is a good starting point). By the time you understand OpenCL, you'll have no problem jumping into CUDA. Make sure the GPU you get has support for, at least, version 3.0 compute capability; technically speaking, you can settle for 2.1, though it lacks some of the features that are now commonplace in CUDA development.