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]

Why would she if there's other Intel executives being more honest about 18A, and she is actively pushing for more products to be external?

If it's not dual faced than what is it hey 18A is good and also we will be doing more products external. If 18A is good why are you doing more products external.

He is doing the reverse while threatening to cut 14A and future advanced nodes permanently?

He is playing 4D chess with companies. Intel can still get decent wafers out of Oregon Dev fab for 14A. He won't be ramping Ohio.

This is literally the bare minimum.
And there is no point to selling the fabs by now. They already sunk all the money into 18A, milking it as much as they can is just getting a good ROI. Even the board would likely agree there's no point now (unless they get a very, very good deal

It's better to not put the clown of board they thought Selling Fab was a good idea before LBT put sense into them.

 

[Geddagod]

If it's not dual faced than what is it hey 18A is good and also we will be doing more products external. If 18A is good why are you doing more products external.

Analysts are already asking that question in their investor-executive technology conferences haha.
MJ just repeats the usual BS
At least Zinsner is honest about 18A not turning out how they originally hoped, MJ just refuses to acknowledge that.

He is playing 4D chess with companies. Intel can still get decent wafers out of Oregon Dev fab for 14A. He won't be ramping Ohio.

Even if Intel does do 14A just internally, then it's not a real threat, is it? Why would external customers care at all?
Do you think LBT has fooled only the rest of the tech CEOs, but somehow we have peered past their bluff? Either LBT isn't bluffing, or other tech CEOs know that LBT is bluffing just as much as you do.

 

[511]

At least Zinsner is honest about 18A not turning out how they originally hoped, MJ just refuses to acknowledge that.

He also said now it is doing fine now in terms of performance and they are yielding it.

Even if Intel does do 14A just internally, then it's not a real threat, is it? Why would external customers care at all?
Do you think LBT has fooled only the rest of the tech CEOs, but somehow we have peered past their bluff? Either LBT isn't bluffing, or other tech CEOs know that LBT is bluffing just as much as you do.

Intel doing Internally is different than doing it for external they can not do lots of stuff required for External. Also nearly Every Semi CEO is looking for alternative except for couple closed ones. It was LBTs way of waking everyone that Intel is not doing a free service.

 

[Geddagod]

He also said now it is doing fine now in terms of performance and they are yielding it.

"fine" and "yielding it'' aren't exactly very confidence inspiring lol. Sure yield may be usable now, but who knows how far they fell short of their initial goals.

Intel doing Internally is different than doing it for external they can not do lots of stuff required for External.

The threat would be that there would no longer be an Intel foundry that other customers can use to potentially get TSMC to lower prices. 14A continuing, even just internally, would still be a threat for that, since apparently there is no internal vs external pdk. And because R&D has been started with external in mind from the start, there should no longer be that sort of a difference.
It's not a 4D chess move at all.

Also nearly Every Semi CEO is looking for alternative except for couple closed ones.

And they clearly don't think Intel is an alternative. Intel so far is only courting 2 customers for 14A.

 

[DrMrLordX]

Intel so far is only courting 2 customers for 14A.

That's two more than they have today for 18A

 

[511]

That's two more than they have today for 18A

We don't count DoD ?cause that one is guaranteed as of rn

 

[511]

And they clearly don't think Intel is an alternative. Intel so far is only courting 2 customers for 14A.

Intel never said anything about customer on 14A.

The threat would be that there would no longer be an Intel foundry that other customers can use to potentially get TSMC to lower prices. 14A continuing, even just internally, would still be a threat for that, since apparently there is no internal vs external pdk. And because R&D has been started with external in mind from the start, there should no longer be that sort of a difference.

They can still cut stuff that is unnecessary that they won't use.

"fine" and "yielding it'' aren't exactly very confidence inspiring lol. Sure yield may be usable now, but who knows how far they fell short of their initial goals.

Well they matched their goals later and who knows what time frame is talking about when they didn't miss their performance target.

 

[DrMrLordX]

We don't count DoD ?cause that one is guaranteed as of rn

I'm surprised they have the DoD on 18A. What projected volume?

 

[511]

I'm surprised they have the DoD on 18A. What projected volume?

Not large enough but it was a $3 Billion Contract.

 

[Geddagod]

Intel never said anything about customer on 14A.

Shot in the dark- Nvidia and Microsoft

They can still cut stuff that is unnecessary that they won't use.

Like what?

Well they matched their goals later

Look at 18A performance claims on their website...

 

[511]

Shot in the dark- Nvidia and Microsoft

Intel and Microsoft has been doing Business sooo long that it's way too easy guess and No for Nvidia.

Like what?

https://www.cadence.com/content/dam/cadence-www/global/en_US/documents/tools/silicon-solutions/protocol-ip/protocol-ip-br.pdf

Look at the IP table if they want to do external they ought to have the IP for it.

Look at 18A performance claims on their website...

15% but in the paper they show 18-25% at a conference

 

[igor_kavinski]

View attachment 130358

"with the two engaging customers" could be interpreted as 18A and 14A both being good enough to appeal to customers.

 

[regen1]

Hmm what is this about then?

https://twitter.com/x/status/1967806574422020287

Qualcomm had test chips at 18A like some others but mobile chips were never a primary target of 18A, at least vanilla 18A.
GAAFET + BSPDN, at least initially the primary target is HPC.
Mobile chips will take some time.
TSMC is transitioning to GAAFET with optional BSPDN.
N2 GAAFET with no BSPDN
A16 = N2 wtih BSPDN and some optimizations
A14 will have versions with and without BSPDN.

Not much reason for Qualcomm being brought to discussion by him when Intel never made such claims on mobile chips unless he means Qualcomm not using 18A for some other stuff

 

[desrever]

mobile chips were never a primary target of 18A, at least vanilla 18A.
GAAFET + BSPDN, at least initially the primary target is HPC.

This is some serious goal post move and cope. If thats the case then why is high powered Nova lake SKUs are on TSMC and only the lower end, low powered SKUs are using 18A?

 

[regen1]

This is some serious goal post move and cope. If thats the case then why is high powered Nova lake SKUs are on TSMC and only the lower end, low powered SKUs are using 18A?

By mobile, I am referring to smartphone chips. You can check all the initial 18A slides and the numerous official quotes by various Intel reps.

FINFET to GAAFET is one thing but FINFET to GAAFET+BSPDN is another thing. More complexity and cost for smartphone chips.

 

[fastandfurious6]

only the lower end, low powered SKUs are using 18A/Intel processes

Bingo

Intel hasn't been able to produce any high-powered chip on self-node since Intel 10nm++++++ (the last 10nm iteration got rebadged to "Intel 7", alder lake to raptor lake).

Can someone translate this into professional terms? That statement will basically be the core technical reason of Intel's demise

 

[DrMrLordX]

Qualcomm had test chips at 18A like some others but mobile chips were never a primary target of 18A, at least vanilla 18A.
GAAFET + BSPDN, at least initially the primary target is HPC.

Why would they target HPC when the volume play for any foundry is almost always mobile? Were they thinking margins or . . . ?

If thats the case then why is high powered Nova lake SKUs are on TSMC and only the lower end, low powered SKUs are using 18A?

18A seems to have problems reaching high enough clocks (parametric yield problems I guess?).

Intel hasn't been able to produce any high-powered chip on self-node since Intel 10nm++++++ (the last 10nm iteration got rebadged to "Intel 7", alder lake to raptor lake).

Not entirely true. Granite Rapids is on Intel 3.

 

[fastandfurious6]

Not entirely true. Granite Rapids is on Intel 3.

the only exception, and server chips are designed for max cores not max perf

if it was that good they could easily use it for high-power client SKU... they didn't, intel 3 has the (useless) meteor lake

 

[511]

Bingo

Intel hasn't been able to produce any high-powered chip on self-node since Intel 10nm++++++ (the last 10nm iteration got rebadged to "Intel 7", alder lake to raptor lake).

Can someone translate this into professional terms? That statement will basically be the core technical reason of Intel's demise

ARL-U aka MTL-U reaching 5.3 Ghz exists and iGPU for PTL as well.
You are missing the part Intel produced more Server CPU on Intel 3 than AMD Turin as well.

 

[511]

Why would they target HPC when the volume play for any foundry is almost always mobile? Were they thinking margins or . . . ?

Intel's Specialty Intel hasn't made a mobile Optimized like ever.

 

[fastandfurious6]

ARL-U aka MTL-U reaching 5.3 Ghz exists and iGPU for PTL as well.

U cpus were first introduced 12+ years ago. they still suck...... literally still no reason to prefer U over 45w+ cpus. that is at least 12+ years after U cpus got introduced.... The only use case for em theoretically is handhelds, and we see Halo 9950x in a handheld running fine.... limited ofc but fine

 

[511]

U cpus were first introduced 12+ years ago. they still suck...... literally still no reason to prefer U over 45w+ cpus. that is at least 12+ years after U cpus got introduced.... The only use case for em theoretically is handhelds, and we see Halo 9950x in a handheld running fine.... limited ofc but fine

Well 2+8 CPUs are perfectly fine for most people who do we browsing and basic office stuff. You are just knowingly denying the fact.
There is not a single Zen 4/5 CPU reaching 5.3 GHz on Mobile except the ones based on Desktop die or the halo as well.

 

[DrMrLordX]

Intel's Specialty Intel hasn't made a mobile Optimized like ever.

Yeah but if you're going to market IDF, you've got to appeal to customers. There aren't that many players outside of Intel that want to fab HPC on 18A.

 

[511]

Yeah but if you're going to market IDF, you've got to appeal to customers. There aren't that many players outside of Intel that want to fab HPC on 18A.

Intel simply added that part to 18AP while I agree with you on they should have added that to 18A.

 

[Geddagod]

You are missing the part Intel produced more Server CPU on Intel 3 than AMD Turin as well.

It's pretty close.
5.72 million GNR+SRF units in, 4.35 million Turin units (Q2 2025). AMD ships more 4/3nm using parts than Intel does in server.