Nanometer race: Intel introduces new node names such as “Intel 7” and “Intel 4”

0
248

Intel gives its manufacturing processes new names: 10 nm Enhanced SuperFin becomes “Intel 7”, 7 nm becomes “Intel 4” and later “Intel 20A” for 2 nm is to follow. Intel is aligning its marketing to Samsung and TSMC, smaller means faster and better. The change is important for the new foundry business.

Intel draws level with Samsung and TSMC

Progress in production, that means officially a new production stage almost every year at Intel, no matter how big the changes were. There were suffixes such as “+++”, “SuperFin” or “Enhanced SuperFin” – the last term to identify the revised “10 nm production” does not even survive a year with today's adaptation.

Because as Intel announces, the current and all future nodes will be renamed again. This time, the marketing department will primarily take over, as was already indicated in the first rumors at the end of March this year. The new names are based on Samsung and TSMC, which manufacture chips according to their own classification in 7 and 5 nm and are on the way to 3 nm. Intel itself wants to leave out the term nanometer and only use names like “Intel 4”. Because according to Intel, the term has not been correct with regard to the gate length, which the original name was aimed at, since 1997.

New designations for Intel's production stages (Image: Intel)

Power per watt is the new reference point

With all the new names, Intel only wants to refer to power per watt, from now on there are no longer any pure reference values ​​such as transistor density or pure nanometers. As is well known, the competitors also proceed in a similar manner, where performance improvements or energy-saving potential are always primarily spoken of.

The industry and TSMCs look at the nanometer race ( Image: Intel)

And that's why Intel today calls 10 to 15 percent more performance with the same energy requirement in the “Intel 7” production stage, with “Intel 4” it should be even 20 percent, while “Intel 3” should add another 18 percent. All values ​​are thus within the scope of what the other foundries always reveal in their optimizations and new production steps, such as recently for example Samsung Foundry.

Performance gains in the production stages at Samsung Foundry (Image: Baidu)

Product overview until 2025

Alder Lake will be the first product in “Intel 7”, which is to be released this fall. Series production of the chips is already underway. Sapphire Rapids for servers also uses this manufacturing process. Meteor Lake will be manufactured in “Intel 4” the following year, but will not be launched until 2023. These CPUs should fall into the same TDP range as the current products, Intel today confirms for the first time 5 to 125 watts – these values ​​are already set at Alder Lake. In addition to client CPUs, the server segment is also addressed: Granite Rapids is planned with the “Intel 4” manufacturing process for the same period as Meteor Lake.

The “Intel 3” process now stands for something that was once thought to be 5 nm production. Intel does not officially name products yet, but they should be manufactured by the end of 2023 and be available from 2024. The improvements are effective in all areas, which is partly thanks to the increased use of EUV, without which this would hardly be possible.

Intel's successes in the production stages (Image: Intel)

Gate- All-around FETs in the Ångström node

From 2024 Intel wants to usher in a new era. Instead of calling the production “Intel 2” based on 2 nm, one goes on the Ångström path and increases the power of ten accordingly, because 1 Å = 0.1 nm. Intel 20A means the first stage, in which new gate-all-around FETs and massive adjustments to the interconnect will take place. Intel calls them “RibbonFET” and “PowerVia” in marketing. “Intel 20A” is to be followed by “Intel 18A” in 2025. At this point, Intel has already revealed so much that High-NA-EUV will be used by ASML for the first time. ComputerBase will provide further details in a separate report.

Big announcements must be followed by action

New levels of production, names and more have been coming from Intel every year. And every year they were thrown overboard, including legendary announcements and prospects far into the future. In the future, nothing at all should come from the previously planned “+” and “++”: no more LP, HP, LPE or anything else, but always one new step after the other. The fact that this can be sustained is to be classified as an ambitious goal, because currently no competing foundry can manage it.

Intel roadmap with backport option from 2019 (picture: Intel)

Roadmap shows self-confidence

Today, the starting position for Intel is therefore no different than before. Laying out a roadmap for three or four years in public shows only one thing: self-confidence. But that is exactly what the company has to deliver – and not change the names again within a year. Intel hopes that competitors will sooner follow this path again. Purely in terms of marketing, TSMC with 2 nm could perhaps sell better than “Intel 20A”, smaller only by name is currently known to be “better” – customers were ultimately vaccinated on this.

ComputerBase received information about this article from Intel under NDA. The only requirement was the earliest possible publication time.