Does every GPU need its own dedicated CPU core?

Found this posted in a folding@HOME forum a year ago:

Is there any truth to this, or is this something folding-specific? (And what’s core17?) I was actually thinking about building a system with more GPUs than CPU cores.

Found an answer, after much googling. Looks like this is only the case for OpenCL on Nvidia GPUs, but not CUDA or OpenCL in general. (Core17 is a version of Folding@HOME on GPUs that uses OpenCL)