The alternatives seem to be:
i) If you are limited by one or the other, you are going to run into a problem when the scene taxes the constraint hardware. So you are constrained by one of the CPU or GPU and 95% of the time you have redundant (wasted?) capacity in the less taxed hardware. Is that significantly better?
ii) Limit framerate so that nothing is constrained. So you sacrifice framerate for 95% of the time (for argument’s sake) to have (redundant) headroom for just in case.
Personal choice I suppose.