Why Is Intel Killing the Nanometer Nomenclature in 2022?

Intel is not only eliminating the concept of nanometer with the technologies it announced, but also ushering in the golden age of silicon processors, the angstrom era.

Intel, which AMD has put into a difficult situation with its Ryzen processors in the last few years, is eager to introduce the technologies it has developed in order to retake the market. However, we can say that these innovations are not so edible and open the door to a milestone development for silicon processors nanometer.

If you are one of the few who follow the technology agenda, you must have seen that many companies, including AMD, which produces processors other than Intel, turn to outsourcing production facilities in order to reduce costs. Intel, on the other hand, was not very keen on turning to outsourcing production facilities, despite all the criticism made against it.

As a result, we have all witnessed that while Intel continues to develop its own production technologies and invests in its factories, AMD produces more efficient processors compared to Intel by using TSMC’s advanced lithography technologies. The market started to gradually shift towards AMD. The name of Intel, which uses different units as a measure in processors with TSMC, has started to be badly remembered.

Why is the nanometer being abandoned?

While AMD made the switch to 7nm TSMC production, Intel continued with 14 nm+++. For this reason, the following perception began to emerge among the buyers in the processor market: “While AMD was 7nm, Intel still remained at 14nm, in fact, it was only able to produce a 10nm processor nanometer in such a short time”. However, this did not have any reality, the 10 nm production technology that Intel plans to pass in the next generation was already close to TSMC’s 7nm technology. Because of this, Intel continued to lose market share, with long discussions of nanometers raging in forum gossip.

That’s why Intel is finally updating its production terminology for its nodes to match TSMC production. The company also announced some new technologies. After this summary of the background of the event, we can start talking about the new terms introduced and the technologies Intel has developed for the new production center it plans to establish.

Intel Refreshing

At Intel’s Accelerated event , Intel CEO Pat Gelsinger unveiled a detailed roadmap for future production nodes. It also announced that they have decided to accelerate their innovation-driven production maps to put the leadership in performance-oriented processors on more permanent ground by 2025. Gelsinger also did not bypass Moore’s law , coined by Intel’s founder and physicist George Moore , with the words: “In our quest to perpetuate Moore’s law, we will be relentless in innovating with silicon magic”.

In fact, it is necessary to be happy that Intel’s CEO talks about the company in this way. “Will Intel pull out of the market or focus on AI computers like IBM?” We can say that Intel has seriously announced that it will do what it knows best. It should not be forgotten that a good competition always returns a plus to the users.

This strategy of accelerating and remarketing company targets is also a very important move for Intel’s IDM 2.0 project , which was announced on its website on March 23, 2021. With IDM 2.0, Intel said it would open two more factories in Arizona with an investment of $20 billion and set up a major manufacturing center in the US and Europe to serve customers around the world . Based on this, we can say that Intel plans to become a rival of Asian TSMC with new standards. This is exactly why Intel has determined its measurement method to be compatible with each other. Thus, they will be able to better analyze the competition that will arise between TSMC and themselves and make moves.

New Terminology

It was obvious that the industry needed a new terminology to understandably mean smaller sizes as time went on, rather than the nanometer unit already used to measure the gate length of transistors. Since the time when transistors became three-dimensional, that is, with the transition to FinFET (Tri-Gate in Intel terminology) in 2011, one-dimensional measurement had already become irrelevant from processor manufacturing technologies.

We can say that Intel is still a little late to fix this. While AMD, along with its manufacturing partner TSMC, is pushing its marketing strategy by launching nominally 7nm processors, Intel seemed to be rotting at 14nm nodes. But in reality, Intel’s 10nm nodes were much closer to TSMC’s 7nm nodes when it comes to transistor density.

What is Intel 7?

Intel will have realized this because it’s drawing a line on all this nonsense and completely eliminating the concept of nanometers. Based on the Enhanced SuperFin nodes forming the upcoming new Alder Lake processors , it now announces a specific metrics rule will be used. Initially, the nominal 10nm Enhanced SuperFin nodes will be named Intel 7 . We can say that this is a production process directly equivalent to TSMC’s N7 production .

It is known that in the mass production process with Intel 7 , these nodes provide 10 % to 15% performance gain per watt compared to the previous classic 10nm SuperFin . That’s why Intel thinks it’s right to give this manufacturing technology a new name. This is a normal change to be expected during the renewal process. Intel will no doubt put this to good use in the markets.

Then we see Intel 4 , which would be referred to as 7nm if nanometers had not disappeared and would come out as a competitor to TSMC’s N4 production process . With Intel 3 and Intel 20A, the limits of silicon seem to be pushed.

According to the image shown at the Intel Accelerated event, the following performance gains and technologies will come to the silicon world with these new processes. If we refer to the ones written in the image in order,

Intel 7

Developed based on the development of FinFET transistors, this architecture will provide approximately 10% to 15% performance gain per watt compared to Intel 10nm SuperFin. It will be used in the Alder Lake architecture in 2021 and in processors expected to go into production for Sapphire Rapids data center processors in the first quarter of 2022.

Intel 4

EUV lithography will be fully utilized to embed incredibly small technologies inside processors using ultra-short wavelength lights. In addition to the improvements in space, there will be a 20% performance increase per watt. Intel 4, Meteor Lake and Granite Rapids architecture data center processors for end users will be released in 2023. It will be ready for production in the second half of 2022.

Intel 3

It will take advantage of FinFET and enhanced EUV to deliver an 18% performance increase per watt along with additional space improvements compared to Intel 4. It will be ready for production in the second half of 2023.

Intel 20A: Angstrom Era

With the 20A, Intel ushers in the angstrom era of processors with two new and groundbreaking RibbonFETs and PowerVia. RibbonFET, a versatile transistor developed by Intel, will be the company’s first new transistor architecture since it introduced FinFET in 2011. With this technology, it will provide faster transistors while adding drivers with the help of multiple fins in a smaller space. With PowerVia, it is an innovation that Intel has made in terms of power in the industry. PowerVia, a rear power distribution technique, eliminates the need for power routing at the front plate of the chip, further optimizing the transmission of signals. The Intel 20A is expected to be in production in 2024.

2025 and beyond

Beyond the Intel 20A, the Intel 18A will already begin development in early 2025, with optimizations to the new transistor technology RibbonFET. Intel will also work to develop the next-generation High NA EUV technology. For this reason, it cooperates closely with ASML to purchase the first production vehicle in this field in the sector.

Why Is Intel Going To Rename?

Undoubtedly, it is not all about innovation. There is also the marketing of the production center to be established in the business. It shouldn’t be surprising that big manufacturers also start to turn to Intel’s foundries if they get a response from the user side. However, we hope that the new techniques will become more detailed as we continue to talk about chips from different manufacturers.

After that, beyond the Intel 3, which will be delivered at the end of 2023, where things get very interesting is the renaming to the Intel 20A, which will be released in 2024. Naming the first node of the angstrom era of semiconductors , Intel 20A, indicates that nominal production below 1nm is targeted.

What is Angstrom?

This may also be where Intel mixed up the simplified node naming again . 1 angstrom is literally a unit of measurement below 1nm, which equals 0.1 nanometer . However, Intel has difficulty pointing out that although the letter A in Intel 20A technology stands for angstrom, it is just a name that refers to architecture, not measurement. In other words, Intel 20A should not be thought of as a process that measures 2 nm transistors, because it will not be.

These new angstrom chips will not only come with the new naming conventions that Intel hopes to catch up with, they will also come with RibbonFET , the first new Intel transistor developed since FinFET, which debuted at 22 nm in 2011 . RibbonFET, also known as NanoSheet or GAA, will launch with the Intel 20A in 2024.

Intel 20A will also bring with it a new power distribution model called PowerVia . Signal transmission will become more optimized and will greatly contribute to the production process.

All of these are quite technical issues for the end user, but they are also a bit of a prediction for the future. Intel already plans to be a leader in processing performance per watt with 18A by 2025. While Intel insists it’s something more than a performance race, the only thing end users care about is performance and power consumption, of course.

It seems that users will find what they really want, namely performance leadership, in Intel in the coming years, similar to the past, unless AMD makes a move.