Was reading through the turing and volta architecture whitepapers, and was curious if a GPU could simultaneously perform multiple different operations if they were of different precisions.
A GPU certainly can, given the overwhelming parallelism on multiple levels in its architecture.