Why is everyone saying CUDA is not open source? It IS open source and free to use, just as much as OpenCL. If it wasn't, then my company (and my alma mater) would be paying NVIDIA out the nose, which we aren't (except in hardware). We're not "sponsored" by them either, unless you count developer e-mails and forums posts as "sponsorship". The only thing people are complaining about is the fact that CUDA is "locked" to NVIDIA cards only, which I wholeheartedly agree with. Personally, it's the only reason why I have a GTX 460 768MB along side my Crossfire setup. What everyone is failing to understand is that optimization is already existing for NVIDIA's implementation of OpenCL (they have 100% compatibility with OpenCL 1.1 as much as AMD has), it's just that CUDA is more in use because of the wide array of functions and support. (e.g. optimizations, direct video memory usage, static code analysis, etc.) Again, usage of CUDA is free, just like using *nix. A lot of open source (and commercial) developers would not be using it if it wasn't.