Explore / Foundry Processes / What a Node Name Means

What a Node Name Means

Foundry Processes foundational foundry marketing; industry roadmaps 7 nm is not 7 nm

The idea

Through the 1990s a node name was the drawn gate length: a 350 nm process had 350 nm gates. Once gate length stopped shrinking in step with density (around 22 nm), the names kept counting down as marketing labels for successive generations. Nothing in a modern "3 nm" process measures 3 nm. What defines a node is a set of pitches and the density they permit:

cell area ≈ contacted gate pitch × cell height ; cell height = tracks × metal pitch

contacted gate pitch (CPP) the distance from one transistor gate to the next including the contact between them; sets cell width per transistor
metal pitch (MP) the pitch of the finest routing layer; with the track count it sets cell height
tracks how many routing tracks tall a standard cell is (from 12 in older nodes down to 5 or 6 in the newest); fewer tracks per cell is where much recent density came from
fin / sheet pitch for FinFET and gate-all-around devices, how many fins or sheets fit under the cell height, which bounds drive strength

Two processes with the same name can differ by 30 percent in density, and one foundry's "7 nm" can match another's "10 nm". Public comparisons therefore use CPP (Contacted Poly Pitch), MP (Metal Pitch), and a density figure such as transistors per square millimetre for a standard cell mix.

Why it matters to a physical-design engineer

Design rules follow pitches, not names: the router's grid, the cell library's height, and the multi-patterning constraints all come from CPP (Contacted Poly Pitch) and MP (Metal Pitch).
Track height sets library architecture: a 6-track cell has room for fewer fins and fewer pin-access points than a 9-track cell, so the same node offers different libraries with different density-versus-performance points.
Cross-foundry comparisons by name are meaningless; compare the numbers.

Pros, cons, and the trade-off

What names still do

Order generations within one foundry: N5 is newer and denser than N7.
Anchor roadmaps and public discussion, even if loosely.

What they no longer do

Measure anything: no drawn feature matches the number.
Compare foundries: the same name means different pitches at different fabs.
Predict cost or performance: those depend on the pitches, the device, and the library, not the label.

The trade-off

The name is a convenient fiction that survived because it is convenient. A physical-design engineer works in pitches, track heights, and rule decks; the number on the slide is for the slide.

Where to look it up

The IRDS (International Roadmap for Devices and Systems) tabulates the pitches behind each generation.
Foundry technology pages give CPP (Contacted Poly Pitch), MP (Metal Pitch), and density figures; independent teardown analyses measure them.
For open kits, the pitches are in the technology LEF (Library Exchange Format): SKY130's finest metal pitch and ASAP7 (Arizona State University 7 nm predictive PDK (Process Design Kit))'s can be read directly.