Saying the GPU is overpowered compared to the CPU kind of loses any meaning when GPGPU w/ unified memory is in the picture, doesn't it?
XBone's CU shortfall exists because they are using ESRAM, which in best (limited) circumstances pretty much matches PS4.
The ESRAM is displacing the PS4's increased CU's, ROPs, and ACE.
MS' choice between upclocking and +2CU is based on their existing design, which is ROP bound, and thus upclocking helps everything including ROP.
That a 5% clock bump is more helpful than a +16% increase in CU's is in fact revealing of the limits of their design.
Otherwise, think if there is a reason that high end PC GPUs tend to be very big with many CUs (more than either XBone/PS4),
instead of just aiming for high clock speeds of lower CU designs, with either high failure/bin rates or exotic materials to enable that.
Or for that matter, why GPGPU using GPUs for non graphics code instead of using conventional CPUs is actually a major thing, with GPUs of course having MANY MANY shaders/etc vs. CPUs relatively fewer execution cores at high clockspeeds.
