If you are CPU-bound, extra threads won't help in Python due to the GIL (it limits an entire Python process to one core, even with multiple threads).
Robert
Yes, and if you're doing any kind of parallel processing and you want it sped up "as much as possible", then you need to use the GPU. The easiest way is to use OpenCL or CUDA or something. Best way is pure OpenGL, with floating point FBOs or the transform feedback buffer.
Ian