Chip makers moved beyond simple transistor shrinking to keep improving

In short: Chip makers did not hit a hard wall on making chips better, but they had to move beyond the old habit of simply shrinking transistors.

What's going on

In the early 2000s, many engineers and executives worried that the usual way of improving computer chips was running out of road. The usual method was to shrink transistors, which are tiny on and off switches that make computers work.

As transistors moved below about 10 nanometers (a nanometer is a billionth of a meter), new problems became harder to ignore. One was quantum tunneling, which is when electricity can leak through barriers that are supposed to block it, like water seeping through a wall. Manufacturing also got more complicated, and the cost of each new generation of chips climbed sharply.

Research from that period suggests the “limit” was not only about physics. It was also about money and practicality, meaning it might still be possible to go smaller in theory, but not worth the cost in real factories.

Instead of stopping, the industry adapted. Companies leaned more on 3D stacking, which is like building a taller apartment building instead of spreading out on one floor. They also used chiplets, which means building a processor out of several smaller chips that work together, like a team of specialists rather than one all purpose worker. Better materials and newer ways to print patterns onto silicon also helped.

What to watch

The big question now is not whether chips can keep improving, but how. Watch for more designs that focus on stacking, mixing chiplets, and smarter ways of connecting parts, since the easy gains from simple shrinking are harder to get.

Source: NYTimes