Originally Posted by mushroomboy;13433963
Have you ever thought that OpenCL wasn't designed for this? CUDA was, but it's not using strictly CUDA but being based off CUDA. If AMD had the hardware to support the operations it could easily be ported and done, since they don't offer it currently then it's not as easy as OpenCL.
 So apparently OpenCL was an option, but I'm wondering if they would have had to do extra work to get OpenCL to work. I'm guessing CUDA has the work done already which is why they chose it.
I'm guessing the extra work for OpenCL isn't worth it as nVidia cards aren't as efficient in OpenCL as they are in CUDA and on top of that, fglrx still isn't anywhere near the best driver for Linux.
Originally Posted by TFL Replica;13436529
It's funded research. Some of the responses in this thread are at a level of absurdity I didn't even know existed!
True enough, it's exactly the same as TWIMTBP, but fglrx would probably stop it working well anyway with OpenCL.
Originally Posted by DuckieHo;13436582
Just because you don't know about them, it does not make them worthless.
For example, NVIDIA GPGPUs power some Wall Street High-frequency trading. This is worth more than you can imagine....
I should have added "That I know of", sorry.