Instead of one giant slab of silicon, modern chips are increasingly built from smaller dies stitched together. The reasons are mostly about cost and yield.
For most of the industry's history, a processor was one piece of silicon, a single die cut from a wafer. The trend now, especially in high-end CPUs and accelerators, is to build a processor out of several smaller dies, called chiplets, connected together inside one package. This is not a gimmick. It is a response to hard economic and physical limits.
The yield problem#
Chips are made on circular silicon wafers, and the manufacturing process is never perfect. Defects land randomly across the wafer. Any die unlucky enough to contain a fatal defect is scrap.
The key insight is how defects interact with die size. A large die presents a bigger target, so it is statistically more likely to catch a defect, and when it does, you throw away a large, expensive piece of silicon. Smaller dies catch fewer defects each, and when one is bad, you lose only a small piece. So for the same defect rate, cutting a design into several small dies yields more good silicon per wafer than one big die.
That is the core argument. Smaller dies mean better yield, and better yield means lower cost.
The reticle limit#
There is also a hard ceiling on how big a single die can be. Lithography prints a chip through a mask, and the machine can only expose a limited area in one shot, known as the reticle limit. You cannot simply keep growing a monolithic die past that boundary.
Chiplets sidestep this. If you need more total compute than one reticle-sized die can hold, you build several dies and connect them. The package as a whole can deliver far more transistors than any single die could.
Discussion