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

 

[511]

As expected this is what Pat was fired for - "18A on track, yields great blah blah blah", quiet clearly renaming 20A to 18A does not conjure premium customers out of thin air the moment PDK is "ready". The -P version looks like what 18A should have been from start - this year.

It seems to me also that their gamble on High-NA is a bust too, TSMC made the right decision not to rush it.

It's over.

Gamble on High NA is a TBD don't forget EUV is one of the reasons TSMC is where it is today it may not be viable now but for future nodes it can be.
There was no renaming of 20A to 18A 20A was scrapped what proof do you have they renamed 20A to 18A

 

[Win2012R2]

I REALLY hope they can recover and begin to compete again, but its looking pretty bleak at this time.

The only thing going for Intel is limited N2 capacity at TSMC - but the problem is that AMD does not need to take that much extra market share as long as it's the premium segment that makes most profits, where as Intel might have to end up making Celerons on 18A++++^2

 

[Win2012R2]

There was no renaming of 20A to 18A 20A was scrapped what proof do you have they renamed 20A to 18A

The proof is 18-P, which is what 18A was supposed to be - so yes, clearly it was renamed, which is better market news than saying actually it is delayed by 1 year.

 

[511]

The proof is 18-P, which is what 18A was supposed to be - so yes, clearly it was renamed, which is better market news than saying actually it is delayed by 1 year.

No no no that is not the proof and on what basis are you basing this info on?

 

[Io Magnesso]

I think Intel 18A/18AP and Intel 20A are different things.
Or rather, 18A is a 20A half node or a finished product... Well, it seems like it's essentially the same thing. But it's different

 

[511]

If you look at Intel 3 vs 4 there are clear difference in metals and pitches and RC characteristics would be correct to call two different things as one?

 

[igor_kavinski]

where as Intel might have to end up making Celerons on 18A++++^2

Skymont Celerons FTW!

 

[511]

Skymont Celerons FTW!

It better be Arctic wolf Celerons with AVX-512

 

[igor_kavinski]

It better be Arctic wolf Celerons with AVX-512

Intel should simply give up all pretense of being a tech leader now and start doing seriously desperate stuff like the good old days of dual socket Celerons.

I have an even better idea: quad socket Celerons!

Make the socket universal. Every five years, you drop in a new architecture Celeron and the computing power keeps on increasing. By the time your grand kids are ready to use that computer, they get to experience a 30+ year old Celeron in the first socket

 

[Io Magnesso]

Intel should simply give up all pretense of being a tech leader now and start doing seriously desperate stuff like the good old days of dual socket Celerons.

I have an even better idea: quad socket Celerons!

Make the socket universal. Every five years, you drop in a new architecture Celeron and the computing power keeps on increasing. By the time your grand kids are ready to use that computer, they get to experience a 30+ year old Celeron in the first socket

Dual Celeron Besides, it's not what Intel intended to do... (regular)

 

[DZero]

How is the situation of Intel going? We are at Intel 3? Or what is going on?

 

[511]

How is the situation of Intel going? We are at Intel 3? Or what is going on?

Yes we are on intel 3 rn unless Panther Lake is in consumer hands

 

[DZero]

Yes we are on intel 3 rn unless Panther Lake is in consumer hands

Which products are fully made by Intel by now without using TSMC? CPU and GPUs?

 

[511]

Which products are fully made by Intel by now without using TSMC? CPU and GPUs?

Their whole DC stack except Gaudi(SPR/GNR/EMR/SRF/Clearwater Forest) also Intel is going to continue using TSMC on client even though the compute tile on N2 everything else is on Intel Node except it.
I have seen this posted somewhere but don't quote me on it Intel's TSMC N3 biz is roughly 30K WSPM as of rn. I can't remember where though

 

[LightningZ71]

Just give me a processor with a CPU tile with ten e-cores quads with two quads flexed for higher clocks and the other 8 flexed for lower leakage. Call it the Celeron Pro Series.

Even better, do it with Nova Lake's platform and use two of those tiles. Throughput monster...

 

[DZero]

Their whole DC stack except Gaudi(SPR/GNR/EMR/SRF/Clearwater Forest) also Intel is going to continue using TSMC on client even though the compute tile on N2 everything else is on Intel Node except it.
I have seen this posted somewhere but don't quote me on it Intel's TSMC N3 biz is roughly 30K WSPM as of rn. I can't remember where though

Damn, Intel is on the same boat than Samsung?... is not good the situation then...

 

[511]

Damn, Intel is on the same boat than Samsung?... is not good the situation then...

Not really Samsung is in even worse situation than Intel when it comes to foundry also external utilization is going to drop quite a lot starting next year for Intel.

 

[igor_kavinski]

Just give me a processor with a CPU tile with ten e-cores quads with two quads flexed for higher clocks and the other 8 flexed for lower leakage. Call it the Celeron Pro Series.

Yes. Death to the IDC cores!

 

[Doug S]

I think LBT is thinking of putting the fabs in hibernation mode for a while until they can get their act together. He's probably going to try to run Intel like AMD with outsourced manufacturing. That seems logical because Intel's 2024 losses are mainly fab related.

That's a terrible strategy if that's what they're doing. The less experience they have with recent nodes the more difficult it will be to jump back on the train. They already canceled 20A because they wanted to concentrate on a successful launch of 18A. Oops, guess that didn't work. I suppose the next step will be to cancel 18A entirely and claim it is to concentrate on a successful launch of 14A. If they run into problems there they can always cancel 14A to concentrate on 10A

They are getting dangerously close to becoming a laughingstock like Samsung, announcing nodes that give them "leadership" over TSMC, complete with the rumors of initial poor yields, followed up by rumors of improved yields and customer interest. Then in order to distract everyone from the fact those improved yields and customer interest were a mirage, you announce your next node that will "extend" your imagined lead.

Looks like TSMC may become the only game in town for advanced processes, at least until China eventually finds a better lithography solution than ASML's Rube Goldberg EUV machines and crushes TSMC on per wafer pricing.

 

[LightningZ71]

There is anotherrrrrrr........
(Rapidus began sending test wafers in April, full rate production for "2nm class" products in 2027.)

 

[Doug S]

There is anotherrrrrrr........
(Rapidus began sending test wafers in April, full rate production for "2nm class" products in 2027.)

Yeah I just ignore them because what they're trying to do is 10x harder than what Samsung and Intel are failing to do. That's like someone halfway through college who has never played football trying out for Ohio State and hoping for an NFL career.

 

[marees]

Not really Samsung is in even worse situation than Intel when it comes to foundry also external utilization is going to drop quite a lot starting next year for Intel.

Reportedly the Switch 3 is switching from samsung 8nm to intel 18a

 

[igor_kavinski]

Yeah I just ignore them because what they're trying to do is 10x harder than what Samsung and Intel are failing to do.

Apparently, Japanese like a good challenge

 

[DZero]

Yeah I just ignore them because what they're trying to do is 10x harder than what Samsung and Intel are failing to do. That's like someone halfway through college who has never played football trying out for Ohio State and hoping for an NFL career.

Or someone who is 18 year old trying to be the Captain of Real Madrid.... without playing a match at all.

Reportedly the Switch 3 is switching from samsung 8nm to intel 18a

I don't be surprised if Rapidus is successful, Nintendo leaves Intel too and goes fully national.

Even more... I don't be even be surprised if Nintendo moves to RISC-V because is starting to get optimized into a way that might end viable in some years.

 

[Tuna-Fish]

It's over.

Big ships take a long time to sink.

Pat's attempt at saving Intel failed, but it's in no way over yet. Everything is on fire, but the company can survive for years in this state, stumbling, hopefully towards salvation. Maybe Tan can do better than Pat.