TSMC to Initiate 1.4nm Chip Fabrication Process Technology R&D

[BiG StroOnZ]

Summary

If a new rumor is to be believed, TSMC is set to formally announce its 1.4nm-class technology in June. The initiation of R&D on TSMC's 1.4nm chip fabrication process / pathfinding of TSMC's 1.4nm node is almost underway. 

 

 

Quotes

Quote

TSMC has once again sparked technological competition by announcing the initiation to develop a 1.4nm chip fabrication process.

 

TSMC will reportedly convert its 3-nm process R&D team into a 1.4nm process R&D team in June. TSMC plans to reassign the team that developed its N3 (3 nm-class) node to development of its 1.4nm-class fabrication process in June, reports Business Korea. Typically, foundries and chip designers never formally announce R&D milestones, so we are unlikely going to see a TSMC press release saying that development of its 1.4nm technology had been started.

 

Standard process technology design flow includes pathfinding, research and development phases. Pathfinding involves things like fundamental exploration of materials and physics and in many cases, it is performed simultaneously for numerous nodes. By now, pathfinding for TSMC's N2 has probably been concluded, so appropriate teams specializing in fundamental physics and chemistry are working on a successors for N2, which may well be called 1.4nm. Keeping in mind that TSMC's N2 enters HVM in late 2025 (so expect the first 2nm chips from the company to be delivered around 2026) and TSMC's two-and-a-half to three-year node introduction cadence, we can potentially expect TSMC's 1.4nm process to be used for commercial products starting in 2028.

 

At the Samsung Foundry Forum 2021, Samsung Electronics announced its plan to mass-produce chips using a 2nm process in 2025. TSMC has gone a step further by initiating the development of a 1.4nm process ahead of Samsung.

 

Intel has recently joined the technology race. It announced a plan to develop a sub-2nm process ahead of TSMC and Samsung. Intel announced that it would re-enter the foundry industry in 2021. It also vowed to mass-produce 1.8-nm products in the second half of 2024.

 

My thoughts

It's crazy to think of a 1.4nm process coming into play while we are currently on 7nm for AMD, 8nm for NVIDIA, and 10nm for Intel (Intel 7). Next-Gen NVIDIA will be 5nm, next-Gen AMD GPUs will be on a mix of 5nm and 6nm, next-Gen AMD CPUs will be on 5nm and next-Gen Intel (Raptor Lake) will still be on 10nm (Intel 7). While the first 2nm chips should come around 2026, they are saying expect the 1.4nm products to come around 2028. While that appears like quite a while, six years from now to be exact, it seems like only yesterday that the Sandy Bridge i7-2600k and Bulldozer FX-8150 was launching on 32nm, and the Kepler GTX 680 / Tahiti HD 7970 was launching on 28nm. The big question is what happens after 1.4nm? Do we start transitioning to different materials other than Silicon (Graphene, Carbon Nanotubes, Indium gallium nitride, Gallium-arsenide) or will progress stagnate (and we see the true end to Moore's Law)? 

 

Sources

https://www.tomshardware.com/news/tsmc-to-initiate-14nm-process-technology-rd

http://www.businesskorea.co.kr/news/articleView.html?idxno=92813 

[TVwazhere]

10 minutes ago, BiG StroOnZ said:

The big question is what happens after 1.4nm? Do we start transitioning to different materials other than Silicon (Graphene, Carbon Nanotubes, Indium gallium nitride, Gallium-arsenide)

IIRC from years ago, 7nm was supposed to be the hard limit for silicon, but I guess they've gotten around that?

[BiG StroOnZ]

5 minutes ago, TVwazhere said:

IIRC from years ago, 7nm was supposed to be the hard limit for silicon, but I guess they've gotten around that?

 

Yeah, I remember when alternatives to Silicon were the big talk not too long ago because of these discussions (hard limits for Silicon). But generally speaking it seemed alternative materials were not even close to mass production. Some of the talk was that they can use combinations of these alternative materials in conjunction with Silicon as workarounds. 

[PeachGr]

30 minutes ago, TVwazhere said:

7nm was supposed to be the hard limit for silicon

That came to my mind as well. If I ve had a dollar for every impossible I ve heard over the years.  

 

32 minutes ago, TVwazhere said:

Graphene

Also a dollar for every time I ve heard that graphene is the solution! It's been years that I hear of it, and as years are passing, I start to believe that this product has a lot of problems they won't say

[Lurick]

2 minutes ago, PeachGr said:

That came to my mind as well. If I ve had a dollar for every impossible I ve heard over the years.  

 

Also a dollar for every time I ve heard that graphene is the solution! It's been years that I hear of it, and as years are passing, I start to believe that this product has a lot of problems they won't say

Nah, you just need to add some graphene to your thought process and it becomes clear!

[porina]

Please keep up to date on marketing nm. Current Intel process in production is called 7, formerly known as 10 SFE until they aligned it with the other players in industry. The 7 is as 7 as TSMC's. Intel are still on their process leadership recovery path which isn't expected to complete until 2025 where they expect to overtake TSMC. I forgot the exact reason but they expect to deploy some new ASML kit in production before TSMC around that time.

[Quackers101]

sus, also it really depends on what they mean.

like is 2nm even giving the benefit we expect from going 5nm to 2nm? and whatever else they might do to reach that and how stable it is.

As intel have done a lot to try and expand on the notion of nm, fitting more in the 3D space, as we hear more about. like the AMD 3D x5800

[Mark Kaine]

3 hours ago, TVwazhere said:

IIRC from years ago, 7nm was supposed to be the hard limit for silicon, but I guess they've gotten around that?

honestly I'm pretty sure it was 2nm... (like 10 years ago)

 

this video explains it a bit (i have no idea why i cant find it on veritasium btw,  either google is stupid or idk lol)

 

but yeah, they must have found some tricks, even 5nm was thought to be nigh impossible a couple of years ago

 

 

2 hours ago, porina said:

Please keep up to date on marketing nm. Current Intel process in production is called 7, formerly known as 10 SFE until they aligned it with the other players in industry. The 7 is as 7 as TSMC's. Intel are still on their process leadership recovery path which isn't expected to complete until 2025 where they expect to overtake TSMC. I forgot the exact reason but they expect to deploy some new ASML kit in production before TSMC around that time.

good point,  it could also just be some name trickery, after all it depends what you measure exactly?  maybe they're still stuck between ~5-10nm in reality  

 

As for Intel overtaking anyone Im not sure, they seem to be behind for a very long time now, its hard to imagine.

[leadeater]

2 hours ago, Lurick said:

Nah, you just need to add some graphene to your thought process and it becomes clear!

Still waiting for my IBM 50Ghz room temperature Graphene CPU....

[porina]

7 hours ago, Mark Kaine said:

good point,  it could also just be some name trickery, after all it depends what you measure exactly?  maybe they're still stuck between ~5-10nm in reality  

That's the problem. It doesn't really measure anything now. Pick a marketing number. It just happened TSMC and Samsung kinda aligned on their scale, and Intel's looked a step out, which is what the renaming corrected. If they wanted to deceive, they could have picked an even lower number. In practice Intel 7 is close enough to TSMC 7nm. Of course, TSMC 7 was in use for longer, with AMD using it from mid-2019, and Intel's equivalent only being used from late 2021.

 

7 hours ago, Mark Kaine said:

As for Intel overtaking anyone Im not sure, they seem to be behind for a very long time now, its hard to imagine.

We can only wait and see. Delays are always a possibility, on all sides.

 

Under the old name system, Intel 10 and 7 are delayed, but processes are not wholly dependant on past nodes to progress. Future nodes can be worked in parallel. Since 14nm family, we've had 10, 10SF, 7 (formerly 10SFE). Through to the end of 2025, they plan to offer 4, 3, 20A, 18A. Note this is for the process, not for product, which can lag. Like AMD will first offer TSMC 5nm based product later this year even though Apple have been for a while. TSMC 4 and 3 are entering production this year. but AMD offerings using either may be way out. Intel 4 is to be used as a part of Meteor Lake possibly towards end of next year. in a similar way AMD CCD and IOD are not made on same process.

 

The way Intel chose to frame their recovery plan does make sense. Whatever is going on in the background, they're getting the next gen gear from ASML before TSMC and aim to deploy it in production before TSMC. It could be a short term lead before TSMC catch up. The sweet spot is around 2025 time frame.

 

 

[leadeater]

9 hours ago, Mark Kaine said:

As for Intel overtaking anyone Im not sure, they seem to be behind for a very long time now, its hard to imagine.

Time scale wise it's not been long at all. Node names and numbers are just well, silly names and numbers. Intel 14nm was still leading edge and better than other market options in multiple aspects for quite some time. Personally I would say only TSMC 5nm is outright better in every way (and only with EUV and new transistor types which are part of TSMC 5nm anyway), TSMC 7nm was not. Intel nodes are designed for completely different purposes than that of TSMC etc and suit Intel's products very well.

 

Intel has been the leading silicon fab company far and away time scale wise, TSMC being a little ahead right now is the exception not the normal. Anything can change of course but 5 years is not 40+ years.

 

And remember everyone is actually ASML, it's all how well one can use the same tool as the other.

[captain_to_fire]

15 hours ago, BiG StroOnZ said:

It's crazy to think of a 1.4nm process coming into play

I think Apple has probably called dibs on the majority of TSMC's 1.4 nm chips before any other company can. This is how Apple managed to release new Macs and iPhones in huge volumes despite the chip shortage in 2021.

[Forbidden Wafer]

People just report it wrongly.

It is not that is unheard of or impossible to build tinier features, let alone use exotic processes and materials.

The real problem is building them at scale and economically.
Much like battery research. We have super advanced designs, including solid batteries, but can't scale up their production significantly.

[Mark Kaine]

4 hours ago, leadeater said:

Time scale wise it's not been long at all.

Ah, ok i guess that makes sense, Ive only been following this only since a few years tbh, so my impression of hopeless intel might be wrong  

 

 

[Doobeedoo]

Nice. Going so small that I really wonder how long silicon can be scaled, when will new materials need to be a replacement.

 

Oh I also just saw this:
https://www.tomshardware.com/news/imecs-sub-1nm-process-node-and-transistor-roadmap-until-2036-from-nanometers-to-the-angstrom-era

[BiG StroOnZ]

4 hours ago, captain_to_fire said:

I think Apple has probably called dibs on the majority of TSMC's 1.4 nm chips before any other company can. This is how Apple managed to release new Macs and iPhones in huge volumes despite the chip shortage in 2021.

 

I see articles on Apple beginning volume production in Q4/2H 2022 of 3nm chips with TSMC, and them being in line for 2nm chips from TSMC in 2024-2025, but nothing regarding them planning on using 1.4nm. It might be one of those things that's not said yet, but understood as likely.

[hishnash]

18 hours ago, BiG StroOnZ said:

 

I see articles on Apple beginning volume production in Q4/2H 2022 of 3nm chips with TSMC

would be strange for them to start in Q4 more likely end of Q2 start of Q3 for iPhone, I don't see apple starting a new node on a larger Mac chip... unless this is something for AR/VR glasses ...

 

22 hours ago, captain_to_fire said:

I think Apple has probably called dibs on the majority of TSMC's 1.4 nm chips before any other company can. This is how Apple managed to release new Macs and iPhones in huge volumes despite the chip shortage in 2021.

For sure, apple is one of the few vendors out there with without any cashflow concerns at all, they are able to put down payments for chips 5+ years in advance. Even intel cant risk making commitments like this as they have internal fab projects that they do not want to undermine.  Apple basically bankrolled 7nm and 5nm and will have also bankrolled 3nm and lower...  Without the iPhone and the 10+ year long project of the M1 with the money apple has put behind it I expect TSMC would not be anywhere close to were they are now. 

 

[Kronoton]

31 minutes ago, hishnash said:

would be strange for them to start in Q4 more likely end of Q2 start of Q3 for iPhone, I don't see apple starting a new node on a larger Mac chip... unless this is something for AR/VR glasses ...

 

I guess it depends on when TSMC is ready to start volume production and what initial volume that actually is. Might just be enough for the AppleWatch or some new product. Might be that they use in just one Mac or that they want to stockpile chips for some early 2023 launch.

 

Other options would be paying TSMC to NOT use 4nm till they start with the 2024 iPhone or allow them to sell the initial volume to someone else. Neither option is really good.

[leadeater]

44 minutes ago, hishnash said:

would be strange for them to start in Q4 more likely end of Q2 start of Q3 for iPhone, I don't see apple starting a new node on a larger Mac chip... unless this is something for AR/VR glasses ...

As above it's down to when TSMC is ready, not when Apple wants them to be. That said we should remember that Apple is a risk partner so the volume production target date from TSMC may not be strictly applicable to certain apple chips, or at least some early batches of them. I'm only seeing 2H though, not Q4 for volume production.

[BiG StroOnZ]

6 hours ago, hishnash said:

would be strange for them to start in Q4 more likely end of Q2 start of Q3 for iPhone, I don't see apple starting a new node on a larger Mac chip... unless this is something for AR/VR glasses ...

 

5 hours ago, leadeater said:

I'm only seeing 2H though, not Q4 for volume production.

 

It's an older article (dated Dec. 2nd 2021) and they're quoting DigiTimes:

 

Quote

Apple’s chip manufacturing partner, TSMC, is expected to begin volume production of 3-nanometer chips during the fourth quarter of 2022, according to a new report from DigiTimes. This could mean that we start seeing 3nm chips in Apple products as soon as 2023.

 

The DigiTimes report itself is sparse on details, citing anonymous “industry sources.” The report indicates that TSMC has started “pilot production” of 3nm chips, with volume production currently slated for Q4 2022:

 

https://9to5mac.com/2021/12/02/3nm-iphone-ipad-mac-chips-tsmc/