In early May, we talked about the fact that IBM was the first in the world to create a 2nm processor by making test samples. IBM executives said at the time that “50 billion transistors were placed on a chip about the size of a fingernail.” A little later, it was clarified that “size with a nail” is 150 mm2… Honest at IBM 2 nm or marketing is a subject of discussion, but the company itself announced just such a technical process.
Now TSMC has announced its own plans to launch production of 2nm chips. Moreover, she also spoke about a number of technological advances, which, perhaps, will give her the opportunity to start producing processors using 1nm technology in a few years.
What’s TSMC got now?
It should be said right away that all these statements about 2nm and 1nm do not mean the possibility of launching the production of the corresponding chips in the near foreseeable future. In one case, the company is just beginning to develop a technological process, in the other it is about preliminary research.
What TSMC has now is an improved version of 5nm technology called N5P. It makes it possible to increase the frequency of the chip cores by 5% or reduce power consumption by 10% with the same crystal complexity as in the case of a conventional 5-nm process technology.
N5P gives you the opportunity to get started with a 4nm process technology. Moreover, in this case, we are talking about the beginning of the production of processors at the end of this year (5-nm chips made according to the improved technical process are starting to be produced just a few days ago). 4nm technology offers little benefit in power consumption and frequencies, based on the design and manufacturing infrastructure of 5nm chips. There are no fundamental differences here – everything is about the same.
The start of production of 4-nm chips is scheduled for the end of this year, but they will be supplied in large quantities next year, 2022. “N4 is a direct transition from N5 with a compatible design, providing further performance, power and transistor density enhancements for the next wave of 5nm chips. Production of N4 is scheduled for the second half of this year, and full production of this type of elements is scheduled for 2022 “, – stated representative of the company.
Next year, TSMC is set to begin production of 3nm chips, during which it plans to use FinFet multi-gate transistors. There will already be a choice: a significant increase in performance or a drop in power consumption (depending on what exactly the customer needs).
According to company representatives, the increase in performance of 3nm processors compared to 5nm process technology will be 10-15%. The new chips will consume a third less energy if the complexity is the same as that of a 5nm chip. The density of the transistors is increased by a factor of 1.1-1.7.
The company will increase the number of layers, but the processors will still be produced by deep ultraviolet (DUV) lithography… Well, since we are talking about FinFET, fundamental changes in infrastructure and tools will not be required. Now the company’s management announced that the development of the 3-nm technical process is actively moving forward, the progress is significant. By the way, the company is going to show the first 3nm chips at the end of this year, as well as 4nm. The lines are expected to start operating at full capacity in 2022.
What about 2nm and 1nm technologies?
As for 2-nm chips, transistors of a different type, GAAFET (gate-all-around FET), will be created for them. By the way, Samsung is switching to GAAFET already in 3nm chips, which will require a change in electronic design automation tools, the creation of a new infrastructure, etc. To speed up the work, TSMC leaves everything for 4- and 3-nm, but changes in 2-nm. Now the Fab 12 factory is being re-equipped to carry out work in this direction.
The company’s management last year announced its intention to to release the first 2nm chips “no later than 2024”. Then the timeline shifted to 2023 – thanks to the fact that Apple joined the development. The Apple company is interested in becoming the first recipient of 2nm processors, so it does its best for this. Most likely, such chips will be produced in China, at the TSMC factory in the city of Baoshan, located near Hsinchu.
Well, more recently TSMC stated about a “technological breakthrough” in the development of 1-nm microcircuits. The company together with a group of scientists from the National Taiwan University (NUT, National Taiwan University) and the Massachusetts Institute of Technology (MIT, Massachusetts Institute of Technology, USA) was able to develop new technology.
The developers used bismuth as a contact electrode, which reduced the resistance of the material and increased the current. After this discovery, made at MIT, all participants began to refine the technology. Including TSMC and NTU. Scientists from the last university offered a method for reducing the component channel using a helium ion beam lithography system.
Of course, before the manufacture of the first processors using 1nm technology, it is several years, so far we are talking only about a scientific achievement, which is not so easy to implement in conditions of mass production. The authors of the technology so far, even purely theoretically, do not compare chips that can be manufactured using a new process technology with existing processors.
Experts argue that 1nm is the technology limit for current processor manufacturing systems. What will happen next is not yet clear.