Aren't these things connected? In the case of M4, increasing the peak performance by a very small amount appears to invoke tremendous costs in power consumption. To me, this indicates that even if the chip were stable at a much higher wattage, the performance win would be, at best, negligible. Wouldn't this mean that the peak is effectively capped?
In my penulutimate post I was distinguishing between whether the barrier was practical (the power vs. clock curve is too steep), or absolute (chips won't run stably at higher voltage). And I was suggesting it was the latter, since even with a steep curve, I thought a Studio or MP should be able to accommodate the power and thermals for a turbo boost on just one or two cores.
To put some acutual numbers to this, I'd like to know what the %power increase would be to boost a M4 core from 4.4 GHz (the current max clock) to, say, 5.0 GHz on a hypothetical M4 Ultra.
For the purposes of this back-of-the envelope calcuation, let's say the M4 Ultra will have 30 performance cores, i.e., 10x the 3 performance cores on the 9-core M4 iPad*. [M2 Max = 8 perf cores; M3 Max = 12 perf cores; so extrapolate that M4 Max would have 15 perf cores ⇒ M4 Ultra would have 30 perf cores.]
Thus we want:
(extra TDP to increase a single M4 core from 4.4 GHz to 5.0 GHz)/(10 x current all-core Max TDP on 9-core M4 iPad)
≈ est. % increase in TDP to add that turbo boost to an M4 Ultra.
Would you be able to supply those two numbers?
For instance, even if it's an extra 5W to do that SC boost, if the est max CPU TDP on an M4 Ultra w/o the boost is, say, 100 W, that's only a 5% increase in power.
*It's better to use the 9-core model, since it's more likely to tell us what the thermally unconstrained (or minimally constrained) per-core TDP (which is what we'd expect to see on the Ultra) would be than the 10-core model.