chiplets

If you integrate two CPU chiplets and two GPU chiplets, you're looking at 36 CPU cores, 80 GPU cores, and 1228 GB/s of memory bandwidth.

If you can connect two CPU chiplets via the same interface, making an M5 Ultra is doable by joining two CPU chiplets, each with a GPU chiplet attached.

Multiple chiplets on an interposer is even better.

But 24 cores per die is small enough that even bad yields give you enough good chiplets.

A chip was always a single piece of silicon (aka die), so chiplets used to just be called chips.

Modern AMD processors are basically a bunch of smaller processors (chiplets) glued together with an interconnect.

You’ll get two more cores and more memory but they moved to chiplets which could hit performance on this first try.

Or it could be the same CPU as in pro/max with more GPU chiplets.

With chiplets, you accept that some units of space will be completely unusable.

In hindsight, this shouldn't be any surprise because these chips are now chiplets.

(It used to be called multi-chip modules (MCM) back in the day but now it's chiplets.) You can see the chiplets pretty clearly in photos of e.g.

On AMD the L3 cache is partitioned between the 2 chiplets.

Today's 'chiplets' term implies that the pieces aren't fully-functioning by themselves, they're more like individual "organs".

An assembly of multiple chiplets would not itself be "a chiplet", but a multi-chip module.

Apple says that M5 Pro and M5 Max use an “all-new Fusion Architecture” that welds two silicon chiplets into a single processor.

So parent was 'wrong' that "chiplets" were ever called MCM's.

Chiplets are different physical chips, potentially from different nodes or processes.

Chiplets are separate dies that come from different wafers.