Hacker News Reader

> Second, the recourse to two M1 Max chips fused into a M1 Ultra means TSMC’s 5 nm process has reached its upper limit.

M1 Max is ~20x22 mm (~430 mm²), double this, even without some of the interconnect die space, doesn't fit into the reticle anyway.

View on HN

Cerebras is 46225 mm²

That is made in a multi-step process, as far as I understand its design, they made the point of designing it from a completely repetitive structure which is the size of one step.

Among the issues that Cerebras had to overcome was how to connect tiles on a wafer where each tile has to be smaller than a reticle. In normal chips you have a gap on all sides where the diamond saw will cut the wafer into individual dies. Having wires cross that gap requires non standard processing. And the tiles themselves would still be limited to a single reticle (under 35mm on a side), so a multi-reticle M1 would not be easy to design.

Reply on news.ycombinator.com