Skip to content

IN DEVELOPMENT · NO CHIP HAS BEEN TAPED OUT

What a mature process node is

A mature process node is a chip manufacturing process that has been in production for a long time: well understood, well characterized, and offered by many manufacturers. An explainer without figures.

· 3 min read

IN SHORT

A mature process node is a generation of chip manufacturing that has been in production for a long time. Its behavior is well understood, its design rules are settled, and many manufacturers offer it. It fits fewer transistors into a given area than the leading edge does. BramHA is being designed for mature nodes; the specific process has not been announced.

BramHA is being designed for mature process nodes. The phrase carries a good part of the program's argument, so it is worth explaining plainly.

This article does so without figures. The program has not announced a process or a foundry, and no chip exists.

A process, and a generation of it

A chip is made by a manufacturing process: a long sequence of steps that builds transistors and the wiring between them on a wafer of silicon. A process node is one generation of such a process.

Each newer generation makes its features smaller. Smaller features fit more transistors into the same area, which means more computing in the same space and usually less power for each unit of work.

The newest generations in production are called the leading edge. They are where the densest chips are made, including the processors that run large AI models.

What makes a node mature

A node becomes mature with time and use. Several things follow.

It has a long production history. A great many chips have been made on it, so its behavior is well understood: how its transistors perform, how they vary, and how they age.

Its design rules are settled. The models that designers use to predict how a circuit will behave have been checked against years of real results.

It is widely offered. Many manufacturers run long-established processes, where only a few run the leading edge.

It costs less to design for. The work of preparing a design for manufacture is smaller, and the price of a mistake is lower.

And it tends to stay in production. Long-established processes are kept running for years because so many products depend on them.

What mature does not mean

Mature does not mean obsolete. Much of the world's electronics for control, power, and sensing is built on mature nodes, and new designs are started on them all the time.

It does not mean the design is old either. The node is the manufacturing process, not the chip. A new design can be built on a long-established process.

It does mean less density. A mature node fits fewer transistors into a given area, so it takes more area and more power for the same amount of computing. That rules out heavy AI work. What mature nodes cost you is the full account of that trade.

Why it suits a robot's own chip

The jobs BramHA is being designed for are control, motion, sensing, and identity. They need certainty: predictable timing, robust behavior, and a part that can be obtained for as long as a machine is made and serviced.

Those are the qualities a mature node is strong in. A well-understood process gives fewer surprises. A widely offered process lowers the risk of depending on one source. A process that stays in production suits machines with long working lives.

The heavy AI work that would need the leading edge is not on the machine at all. It is served from inside the building by an inference pool. Why robots do not need the leading edge makes that case.

What the program has not said

The program has said that the silicon is being designed for mature process nodes. It has not said which one.

The process and the foundry have not been announced. The specification of the first design is not published. There are no performance, power, or price figures, and no chip has been taped out.

So the general qualities described here are qualities of mature nodes as a class. They are the reason for the program's choice. They are not measurements of a BramHA chip, because there is none to measure.

Questions this raises

Is a mature node the same as an old chip?
No. The node is the manufacturing process, not the design. A new design can be built on a long-established process, and many are.
Why are no figures given here?
The program has not announced a process or a foundry, so naming a figure would imply a decision that has not been made. The distinction that matters is between mature nodes and the leading edge.
What does a mature node give up?
Density. A mature node fits fewer transistors into a given area, so the same computing takes more area and more power. That rules out heavy AI work, which is served by a pool in the building instead.

KEEP READING