I have always wondered if the approach of tuning a large amount of cores to fit within a low power window would be comparable tuning a lesser amount to the same window, or if diminishing returns scales both ways (there really is only one efficiency region for a design)
Also I can see a sticker has been applied to the exhaust part of the block which surprise, surprise has lost its adhesion and is peeling off. I know this can't be the final product, but surely they can do better.
I have always wondered if the approach of tuning a large amount of cores to fit within a low power window would be comparable tuning a lesser amount to the same window, or if diminishing returns scales both ways (there really is only one efficiency region for a design)
Depends on the architecture, process, and even batch. Many designs have peak perf/watt at much lower clocks than their peak clock, making larger lower clocked designs desirable in many mobile designs, however some chips on a given wafer or even cores on a given chip can have defects that alter their clock/power curves, more enabled cores increases the likelihood of this, making these chips quite rare on less mature nodes.
The Fury Nano was a fully-enabled core, mostly just a Fury X with a tweaked clock curve and better-binned dies for lower power consumption. That was on a super-mature 28nm node though.