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.

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

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.

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

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

 

[ajsdkflsdjfio]

It's not a completely separate thing.
It's proof that Intel didn't even bother to go to the most obvious large purchasers and create something tailored to their needs...

It feels like Intel ran their playbook from the 1990s with AI. They assumed that the target market should be "workstations" (something basically for an individual, paid for by a department). Which meant they were
(a) competing directly with nVidia in terms of Hopper or Blackwell cards
(b) ignoring by far the largest purchasers.

You're creating an completely different discussion here about something mostly foreign to their product side business model. Intel has never really been a one-off contract design company, and like Nvidia/AMD mostly only design products for sale to the general market usually released under their own brand. Even rare custom products like for consoles largely use their existing IPs with small modifications. Maybe you are right and there is an argument that Intel should have tried to compete in the custom per-company contract market but that's a much more complex argument involving an entirely separate market.

This type of random entry is probably best when intel is actually doing well in their core businesses and have money in the bank to take more risks. Also it's probably best to not try and do this shift of business strategy when the market is already pretty saturated and when you are behind by a ton. You compare intel's AI failings in the past couple of years to Intel missing mobile but they are completely different things. When Intel had the many chances to enter the mobile market, there were much less/weaker established players and Intel was was still dominant compared to most other competing companies.

 

[jdubs03]

Little update here:

https://twitter.com/x/status/1864850746514248087

I was thinking about this just now. Could the Galaxy S26 Ultra have N2 before the iPhone 18 Pro? With the timing it seems quite possible.

 

[poke01]

Little update here:

https://twitter.com/x/status/1864850746514248087

I was thinking about this just now. Could the Galaxy S26 Ultra have N2 before the iPhone 18 Pro? With the timing it seems quite possible.

Nope, Qualcomm chips are released in October

 

[desrever]

Little update here:

https://twitter.com/x/status/1864850746514248087

I was thinking about this just now. Could the Galaxy S26 Ultra have N2 before the iPhone 18 Pro? With the timing it seems quite possible.

Qualcomm also hasn't signed up to use N2 in 2025

 

[jdubs03]

Nope, Qualcomm chips are released in October

Ah yeah true. Not happening then.

 

[SiliconFly]

.

 

[igor_kavinski]

Unless Intel Foundry can pull a miracle, an effort must be underway to redesign the chip for whichever TSMC node is considered cutting-edge in 2025.

Is TPU clueless or is it really possible to redesign CWF for a TSMC node and still have it released in Q4 2025 at the latest?

 

[SiliconFly]

.

 

[igor_kavinski]

The key to life is moderation. Don't overdo any one single thing. I guess the lesson for Intel is the same. They should've been moderately cautious and designed their CPUs for both their and TSMC's nodes so in case of any trouble, they could at least have product out to save face and be like old AMD (paper launches and product availability woes). Instead, now they are even worse off than AMD of old.

 

[SiliconFly]

.

 

[DrMrLordX]

Unless they bought Nvidia itself, don't really see how that puts them in a good position to compete. Like if Intel bought Glofo or even Samsung it certainly wouldn't automatically make them a contender for top foundry.

1). I was merely making the point that Intel did, in fact, have a foot into AI, even if their efforts/acquisitions were not successful
2). NV isn't invulnerable. Many have rightly observed that their approach to AI acceleration is flawed, and that there is an opportunity for a focused competitor to challenge them.

 

[igor_kavinski]

It's fake news. Rumors say it may have taiwanese origins.

Well, either TSMC's too scared of 18A and wants to sign up as many customers as possible into long term contracts by spreading disinformation or they are extremely curious and want Intel to share more public information about their progress so they can make necessary adjustments to their own roadmap.

 

[SiliconFly]

.

 

[OneEng2]

IBM paid Globalfoundries some billions of dollars to take their fabs.

I wasn't aware of that. I wonder how the final balance sheet looked by the time they got it offloaded again though?

Actually, I meant TPU. It's fake news. Rumors say it may have taiwanese origins. Both Patrick Moorhead & Dr. Ian Cutress said it's fake. Ian said the yield was >60% during August itself. It should be well above 70% or even 80% by now.

If Intel is getting good yields now with 18A, They would be publishing it to God and his brother to instill confidence with new customers. For a process that is supposed to be in mass production in about 6 months, the silence is deafening IMO.

If I look at what HAS been published, you get things like "PTL is booting". Sure, but how many wafers of chips did it take to get one that worked?

For Intel to secure the confidence of new customers, they will need to released detailed and industry standardized analysis on the production readiness of 18A, not just click bait.

At 6 months prior to published full volume production dates, I would certainly have thought that this would have been done already.

 

[Win2012R2]

It's all under heavy NDAs, but if 18A really worked then TSMC pricing would have been under pressure (which does not appear to be the case), almost all if not all interested parties in 18A were hoping to leverage it in their pricing negotiations with TSMC.

 

[SiliconFly]

.

 

[LightningZ71]

It's all under heavy NDAs, but if 18A really worked then TSMC pricing would have been under pressure (which does not appear to be the case), almost all if not all interested parties in 18A were hoping to leverage it in their pricing negotiations with TSMC.

The market in question is heavily supply constrained. Producers have near total price authority. The only consumer with notable leverage is Apple because they are willing to risk-shate to help finance their leading edge nodes by purchasing substantial at-risk volume when final volume yields aren't known

Don't believe that about Apple? How many leading edge "known good die" deals do you see in the industry for high volume products? I've seen Samsung do them, but their yields are notably poor, meaning that many customers won't use them any other way.

 

[jpiniero]

The market in question is heavily supply constrained.

Was going to say that more or less... even if 18A was good, Intel literally can't afford to properly build it out. It's almost like Intel Foundry should stop at 18A until they can show they can build an actual customer base that isn't Intel Products.

 

[LightningZ71]

You nailed it. TSMC has never had any form of real competition until now. This is the first time their (thousands of smaller) customers are actually going to have an alternative. A leverage like you said. Not that any of their customers is going to run to Intel fabs considering the financial mess Intel is in. But it still gives them leverage. And thats not good for business. We can see more drama/fake news going forward until their products actually reach the market.


I think they'll be sharing the details this Jan or early Feb in one of their events.

Intel is not really true competition yet, especially for the "thousands of smaller buyers" because their industry standard design tools are still in their infancy. One of TSMC 's biggest strengths is their ability to leverage mature design tools and provide their customers with design assistance all along the way. Samsung even has an advantage over Intel there just by the fact that they've been a foundry for a very long time.

Assuming that Intel survives, they will eventually get close enough there for it to be competitive, but, for now, only big, well funded players can use them for highly custom parts, unless Intel is willing to take the financial hit of holding their hand through using Intel's tools.

 

[Win2012R2]

This is the first time their (thousands of smaller) customers are actually going to have an alternative

Small customers can't afford advanced nodes, R&D+masks isn't nothing also.

 

[Doug S]

It's all under heavy NDAs, but if 18A really worked then TSMC pricing would have been under pressure (which does not appear to be the case), almost all if not all interested parties in 18A were hoping to leverage it in their pricing negotiations with TSMC.

Intel won't have very much 18A capacity when compared to what TSMC is bringing online with N2, let alone what they already have online with N3 and N4/N5. So it is quite possible that both are true: TSMC could increase prices AND Intel's 18A works well.

Even if Intel had tons of 18A capacity to sell which was known to work great that would only put pressure on TSMC's future N2 orders, or plans for N3 expansion, not their current N3 and certainly not N5 orders, which are for customers who have already designed or are currently in the process of designing/taping out for TSMC.

 

[SiliconFly]

.

 

[SiliconFly]

.

 

[maddie]

In a surprising turn of events, Pat Gelsinger has finally spoken. Replying to Patrick Moorhead, he confirms that the 18A 10% yield leak is in fact fake news.

He's right, it's 10.4 %.

 

[SiliconFly]

.