Make no mistake: Rembrandt-H will be a chiplet design consisting of 2 5nm CCD and an 6nm IOD. It will basically be the same as a 7950X.
With this in mind there are IMHO 2 options:
- They keep IFOP as the interconnect. Energy efficiency and consumption will be horrible from a mobile perspective. This will basically be a desktop replacement - nothing else.
- They do something like EFB bringing energy efficiency almost on par with a monolithic solution.
Sadly ATM I think option 1 is much more likely. We know that they keep IFOP for Genoa. And I simply cannot believe that they are doing a CCD being able to connect to both. OTOH. This could explain why the die size of CCD and IOD did not shrink that much given the process advantage. Maybe it gave them some spare area for an iGPU due to pad limitation in the first place? Anyone with more knowledge?