It's a known fact that branching is bad for CUDA. The question I have is a little more specific. I have a (static) loop. Static here means that it follows the exact same path every time, and executes ...
What would be a good GPU to start learning CUDA on? Are any GPUs unsuitable?