Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel) - [2020 - 2025]

[DisEnchantment]

TSMC's N7 EUV is now in its second year of production and N5 is contributing to revenue for TSMC this quarter. N3 is scheduled for 2022 and I believe they have a good chance to reach that target.


N7 performance is more or less understood.


This year and next year TSMC is mainly increasing capacity to meet demands.

TSMC Radically Boosts CapEx to Expand Production Capacities, To Reach $14B For 2019

www.anandtech.com

For Samsung the nodes are basically the same from 7LPP to 4 LPE, they just add incremental scaling boosters while the bulk of the tech is the same.

Samsung is already shipping 7LPP and will ship 6LPP in H2. Hopefully they fix any issues if at all.
They have two more intermediate nodes in between before going to 3GAE, most likely 5LPE will ship next year but for 4LPE it will probably be back to back with 3GAA since 3GAA is a parallel development with 7LPP enhancements.

Samsung’s Aggressive EUV Plans: 6nm Production in H2, 5nm & 4nm On Track

www.anandtech.com

Samsung's 3GAA will go for HVM in 2022 most likely, similar timeframe to TSMC's N3.
There are major differences in how the transistor will be fabricated due to the GAA but density for sure Samsung will be behind N3.
But there might be advantages for Samsung with regards to power and performance, so it may be better suited for some applications.
But for now we don't know how much of this is true and we can only rely on the marketing material.

This year there should be a lot more available wafers due to lack of demand from Smartphone vendors and increased capacity from TSMC and Samsung.
Lots of SoCs which dont need to be top end will be fabbed with N7 or 7LPP/6LPP instead of N5, so there will be lots of wafers around.

Most of the current 7nm designs are far from the advertized density from TSMC and Samsung. There is still potential for density increase compared to currently shipping products.
N5 is going to be the leading foundry node for the next couple of years.

For a lot of fabless companies out there, the processes and capacity available are quite good.

---------------------------------------------------------------------------------------------------------------------------------------------------


FEEL FREE TO CREATE A NEW THREAD FOR 2025+ OUTLOOK, I WILL LINK IT HERE

 

[johnsonwax]

Authors of the paper worked at Intel when it was developed, but now works at Nvidia/Apple. Doesn't imply that Nvidia/Apple are using Intel tech.

 

[jdubs03]

Authors of the paper worked at Intel when it was developed, but now works at Nvidia/Apple. Doesn't imply that Nvidia/Apple are using Intel tech.

I don’t think that is the case for all the authors tbf.

 

[johnsonwax]

I don’t think that is the case for all the authors tbf.

You think it's more likely that Apple is going to be making an unprecedented reveal at a technology conference that they are now fabbing at Intel rather than they happened to be authors that previous worked at Intel? Same for Nvidia?

 

[RTX]

https://www.reddit.com/r/hardware/comments/1k3tb5z/intel_18a_vs_intel_3_power_and_performance

https://twitter.com/x/status/1910922662316298319

There's 2 variants of N4 in the tweet with a 300+mV difference.

 

[511]

https://www.reddit.com/r/hardware/comments/1k3tb5z/intel_18a_vs_intel_3_power_and_performance

https://twitter.com/x/status/1910922662316298319

There's 2 variants of N4 in the tweet with a 300+mV difference.

The design is a bit different for Mobile SOC(HX370) and Halo die for the Zen 5 core and maybe the RTL is different as well? I can't say for sure but there must be a clear difference.

 

[Doug S]

You think it's more likely that Apple is going to be making an unprecedented reveal at a technology conference that they are now fabbing at Intel rather than they happened to be authors that previous worked at Intel? Same for Nvidia?

Even if it was someone who was an Apple employee all along co-authoring such a paper wouldn't indicate any such thing. I think it is safe to say that Apple has the resources to have a few people work with Intel on 18A and future processes so they know where they stand in comparison with TSMC.

Even if Apple believed that say 14A would be better than TSMC's contemporary processes and that Intel would be able to deliver great yields they couldn't make such a switch because Intel doesn't and in that timeframe couldn't have sufficient capacity for Apple's needs.

If we see Apple using Intel fabs it will be sort of a pick and choose situation initially. If Intel was better for lower power maybe they fab future watch SoCs there. If they were better for analog stuff maybe they fab their modem transceivers there, that sort of thing. If we want to know when Apple is switching wholesale to Intel we would have some pretty obvious clues, like Intel suddenly investing tens of billions getting those Ohio fabs built on an accelerated schedule, and their books suddenly showing huge prepayments for future chip deliveries.

 

[eek2121]

Authors of the paper worked at Intel when it was developed, but now works at Nvidia/Apple. Doesn't imply that Nvidia/Apple are using Intel tech.

I don't know anything about any Apple Plans, however NVIDIA is definitely working on using an Intel process in future products.

 

[johnsonwax]

Even if it was someone who was an Apple employee all along co-authoring such a paper wouldn't indicate any such thing. I think it is safe to say that Apple has the resources to have a few people work with Intel on 18A and future processes so they know where they stand in comparison with TSMC.

Even if Apple believed that say 14A would be better than TSMC's contemporary processes and that Intel would be able to deliver great yields they couldn't make such a switch because Intel doesn't and in that timeframe couldn't have sufficient capacity for Apple's needs.

If we see Apple using Intel fabs it will be sort of a pick and choose situation initially. If Intel was better for lower power maybe they fab future watch SoCs there. If they were better for analog stuff maybe they fab their modem transceivers there, that sort of thing. If we want to know when Apple is switching wholesale to Intel we would have some pretty obvious clues, like Intel suddenly investing tens of billions getting those Ohio fabs built on an accelerated schedule, and their books suddenly showing huge prepayments for future chip deliveries.

I'm of the view that Intels foundry business is only going to be saved by investment by domestic fabless industry, and that would need buy in from the government (which isn't going to happen with this crew), and that Apple would have plenty of uses for such a business commitment even if they kept with TSMC for leading node stuff. So in theory I agree with you completely here.

However, Apple is not going to announce such a change in stance except by Tim Cook on a stage at a PR event they control. Apple investors are not going to learn this kind of information at a tech conference. They simply don't operate that way ever. Lots of other companies do, but Apple controls that message extremely carefully. Even in the case of prepayments, Apple hides that shit. I'm a long time investor and spotting these changes in spending, etc. were a big part of how I worked. Apple paid DMG Mori billions to build secret factories that made mills only for their use and nobody spotted it. We saw the increased capex on Apple's books, but nobody could figure out where it was going. Some folks in the CNC trade spaces observed that something had changed in supply and demand in their market and there was some speculation that it was Apple on one side of it, and some speculation that these new mills from DMG Mori seemed to be designed to solve a set of problems the industry wasn't asking to be solved, so they were speculating that they were the public market variant of the design for Apple, but it took years before a DMG Mori exec admitted that yeah, they had these secret factories.

You could see the change on the balance sheets, but there was no way to connect the two or even know if they were related. It'd be harder with Intel because people will look at such a prepayment/capex change and suspect a handful of companies on the other side, but this will always happen on Apple's terms. Even when the new EMV electronic payments standard changed on which ApplePay is based, EMV didn't roll out until after Apple announced. They controlled the entire thing. If you want their billions, they dictate the PR terms.

 

[DrMrLordX]

Even if Apple believed that say 14A would be better than TSMC's contemporary processes and that Intel would be able to deliver great yields they couldn't make such a switch because Intel doesn't and in that timeframe couldn't have sufficient capacity for Apple's needs.

A few years ago, it seemed that Intel was the first customer getting High NA EUV equipment, and it also seemed like they might be getting more machines early on than anyone else. Has that situation changed? If it hasn't, 14a volume may actually be pretty good (compared to TSMC's equivalent node).

 

[511]

A few years ago, it seemed that Intel was the first customer getting High NA EUV equipment, and it also seemed like they might be getting more machines early on than anyone else. Has that situation changed? If it hasn't, 14a volume may actually be pretty good (compared to TSMC's equivalent node).

The issue with High Na is tariffs it's $400 million per machine and at minimum of 10% tariff and you need 20 of these for a fab that would approx increase the cost by $0.8 Billion and add few another things a fab will be quite a bit more expensive.

 

[dttprofessor]

The issue with High Na is tariffs it's $400 million per machine and at minimum of 10% tariff and you need 20 of these for a fab that would approx increase the cost by $0.8 Billion and add few another things a fab will be quite a bit more expensive.

20？no！

 

[Doug S]

However, Apple is not going to announce such a change in stance except by Tim Cook on a stage at a PR event they control. Apple investors are not going to learn this kind of information at a tech conference.

They wouldn't announce it at any sort of event. Apple didn't go out of their way to say anything about switching from Samsung to TSMC a decade ago, and they only mention TSMC in passing at their big events. They know consumers aren't buying phones or PCs because they have TSMC's latest process inside. Heck Apple doesn't even talk about clock rates and that's arguably of greater importance (at least if you are comparing last year's iPhone/Mac to this year's) to consumers than technobabble about "TSMC N3B" or whatever.

Where Apple would announce something like this would be in their quarterly investor call. Or given Trump's desire to take credit for stuff he has nothing to do with, I could see Cook and the new Intel CEO announcing a deal at some sort of presser at the White House, with both hoping (perhaps vainly, but hoping) that letting him take credit might help them down the road.

 

[johnsonwax]

Where Apple would announce something like this would be in their quarterly investor call. Or given Trump's desire to take credit for stuff he has nothing to do with, I could see Cook and the new Intel CEO announcing a deal at some sort of presser at the White House, with both hoping (perhaps vainly, but hoping) that letting him take credit might help them down the road.

I think we're past the point of most CEOs being gullible enough to fall for the 'maybe if we suck Trumps dick he'll give us a nice pat on the head' form of negotiating. Trumps whole thing is establishing dominance, and if you appease him, he's just going to demand more later. He only backs down if you kick his ass. Cook has been extremely quiet though this whole period, and I think that's the smart move - give Trump nothing to work with. Huang announced their Arizona Blackwell commitment, and did the $1M/plate dinner and Trump still yanked the rug out from under him the next day.

 

[Saylick]

Looks like LBT is going good on his word to make staff cuts at Intel. Would be curious to know how the cuts are distributed since there's foundry and design sides. My understanding is that the foundry side employs more people than the design side.

https://www.investing.com/news/stock-market-news/intel-to-announce-plans-to-cut-over-20-of-staff-this-week-bloomberg-3997205

 

[DrMrLordX]

The issue with High Na is tariffs it's $400 million per machine and at minimum of 10% tariff and you need 20 of these for a fab that would approx increase the cost by $0.8 Billion and add few another things a fab will be quite a bit more expensive.

Tariffs are going to be an issue for anything delivered by ASML.

 

[511]

Tariffs are going to be an issue for anything delivered by ASML.

Yeah it's proportional to the equipment cost and you don't have an option 🤣

 

[coercitiv]

Tariffs are going to be an issue for anything delivered by ASML.

They can always build fabs on Europe

 

[DrMrLordX]

They can always build fabs on Europe

True. Though then tariffs affect the end product (if you're still selling in the US).

 

[jpiniero]

They can always build fabs on Europe

I know you are joking, but given the problems that TSMC had with workers in Arizona... I imagine it'd be 1000x worse in Europe.

 

[igor_kavinski]

Overclocking "Arrow Lake" - 285K, 265k, 245k...

😇 Let the games begin.....

www.overclock.net

Summary: 18A unsuitable for high frequency parts. Nova Lake K SKUs will most likely be TSMC silicon.

 

[RTX]

10% faster than N2P

 

[511]

Btw on ASML with High NA it's going to difficult to determine tarrif value the machine is assembled at

View attachment 122635
10% faster than N2P

This means N2P is only 2-3% better than N2 🤣

 

[Thunder 57]

Btw on ASML with High NA it's going to difficult to determine tarrif value the machine is assembled at

This means N2P is only 2-3% better than N2 🤣

Apparantly it's better than 18A at least for the high end stuff.

 

[511]

Apparantly it's better than 18A at least for the high end stuff.

I don't think it will be better than 18AP if they move forward with the promise of 10% with 18AP cause this is sure that 18A is between N2 and N3P.

 

[511]

Apparantly it's better than 18A at least for the high end stuff.

Since they have lowered N2P perf I think 18AP if they move forward with the promise of 10% over 18A will match N2P cause this is sure that 18A is between N2 and N3P.