The GP part stands for "something else than graphics". CUDA is not general purpose in the sense that a lot of types of computation and data movement results in

The GP part stands for "something else than graphics". CUDA is not general purpose in the sense that a lot of types of computation and data movement results in terrible performance. You buy GPU accelerators for performance reasons. Case in point, GPUs are standard (for now) for training DNNs, but the big server farms that actually do the scoring run on Xeons.