◆ CIRCUITS & CHIPS, DECODED FOR THE AI ERA 10 FREE CALCULATORS NO LOGIN · NO ADS · NO TRACKING BUILT FOR EE STUDENTS WORLDWIDE v0.2 LIVE
◆ CIRCUITS & CHIPS, DECODED FOR THE AI ERA 10 FREE CALCULATORS NO LOGIN · NO ADS · NO TRACKING BUILT FOR EE STUDENTS WORLDWIDE v0.2 LIVE

For 30 years, the process node number told you the transistor gate length. Then around 2014, physics stopped cooperating — and the numbers became marketing.

The historical meaning

Until roughly the 90nm node (mid-2000s), the "node" was the gate length of a transistor — the physical distance electrons had to travel between source and drain. Shrink it, and you get faster switching, lower voltage, more transistors per area. This was Moore's Law working exactly as planned.

Where it broke

Below ~22nm, current started leaking through the gate, and shrinking further stopped delivering the same benefits. Foundries kept the node numbers going for marketing continuity, but the actual feature sizes stopped tracking them.

Today: TSMC's "3nm" node has transistor features closer to ~24nm for the smallest metal pitch, and gate lengths around 16-18nm. The "3" is a naming convention, not a measurement.

What actually changed at each recent node

The real metric you should track

Instead of node names, look at transistor density (MTr/mm²) — millions of transistors per square millimeter. TSMC 3nm: ~215 MTr/mm². TSMC 5nm: ~140. This actually tells you how much stuff fits on a chip.

Why the marketing continues

Nobody wants to say "our next chip is on the '22-nanometer-equivalent-density' node." The number is a proxy for progress. As long as everyone agrees to keep the naming consistent, it works — even if it's technically fiction.

why this matters
Whenever you see a headline about "3nm chips" or "2nm coming next year," remember: the number is roughly a generational label, not a physical dimension. It correlates with progress but doesn't measure anything specific.
← back to glossary