And you missed my point. M2 is going to have 4nm version of the performance cores from A15. So is A16. Same cores, on the same node.
Reading the Mac Pro whitepaper, the whole reasoning for Mac Pro is "no limits". Cover the edge use cases completely.
The requirements seem to me, even more extreme processing power, higher memory limits (>1TB), greater peripheral expansion (PCI/Thunderbolt/MPX/?).
Take as few givens:
1. 3rd party GPU cards are off the table.
2. Unified memory programming model is here to stay (at least at the 128GB level)
3. No separate dies from the Pro/Max/Ultra model (given low volume).
If one were going to do a Mac Pro w/M1 Ultra, could they stack or add more addressable memory in a new package? would PCI expansion slots be workable?
If one were going to wait for the next process node, would it make more sense to do the next M2/3 Pro/Max/Ultra on N3 instead of N4?
What's the tradeoff in leading with an M-series chip on a new process instead of an A-Series? Would the low yields with a new process would be economically too costly with these larger dies? Too aggressive?
How big could chip packages get? 4 Max Chips? 8?