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

 

[DrMrLordX]

Capacities aside, Intel is the only other company out there doing sub 7nm besides TSMC and Samsung.

Commercially, Intel hasn't gotten there yet.

 

[Thibsie]

Capacities aside, Intel is the only other company out there doing sub 7nm besides TSMC and Samsung.

Right now ? Are they ?

 

[JasonLD]

Commercially, Intel hasn't gotten there yet.

Well, hasn't released but will happen in 2023, probably towards the end.

Right now ? Are they ?

They are certainly working on it, aren't they? I am not talking about commercial availiability.

 

[DrMrLordX]

Well, hasn't released but will happen in 2023, probably towards the end.

Right. Very much behind schedule. Even if Intel does get Intel 4 into full production on Meteor Lake, by then, the damage will already be done. Continued node development doesn't really put them on the leading edge anymore . . .

 

[JasonLD]

Right. Very much behind schedule. Even if Intel does get Intel 4 into full production on Meteor Lake, by then, the damage will already be done. Continued node development doesn't really put them on the leading edge anymore . . .

Releasing Meteor Lake by 2023 would be keeping them on node parity(or slight ahead) against AMD on consumer side. Besides, TSMC won’t be moving to 2nm until H2 2025 so it will be interesting times ahead.

 

[DrMrLordX]

Releasing Meteor Lake by 2023 would be keeping them on node parity(or slight ahead) against AMD on consumer side.

AMD hasn't been on leading edge nodes since . . . ever.

Besides, TSMC won’t be moving to 2nm until H2 2025 so it will be interesting times ahead.

That's what Intel is gambling on.

 

[Henry swagger]

https://twitter.com/x/status/1549391632482668547

You MUST provide your own commentary when you post a link or picture as we've told you before.

Daveybrat
AT Moderator

 

[igor_kavinski]

https://twitter.com/x/status/1549391632482668547

Oh nice. At least, some good came out of mining. Too bad we won't be seeing Nvidia GPUs on this process this year. Or maybe never?

 

[Henry swagger]

Oh nice. At least, some good came out of mining. Too bad we won't be seeing Nvidia GPUs on this process this year. Or maybe never?

Maybe for the 5000 series if nvidia does duel supply again like the ampere line

 

[nicalandia]

Not Sure where to put this, but since it's related to Intel's Process troubles..

Presenting ‘Cascade at the Bat’: A Bad Intel CPU Poem

Presenting 'Cascade at the Bat': A Bad Intel CPU Poem

Presenting "Cascade at the Bat", a bit of wretched-but-hopefully-funny poetry I wrote a few years ago. Of all the things I've ever published, this is definitely one of them.

www.extremetech.com

"Cascade At the Bat

The outlook wasn’t brilliant for the Intel chip that day
The score stood 32-18, with no threads left to play
With core counts insufficient and clock speed all but gone,
No one knew what to expect when Cascade stepped on the lawn.

It wasn’t very long ago that Intel ruled the game
Conroe, Penryn, and Westmere all caused Sunnyvale great pain
Then came Sandy, Ivy, Haswell — great hitters, one and all
Collectively they owned the field
And fed AMD the ball.

But lo’, a new inning of the game began in 2017
Once AMD at last debuted a capable machine
After years of piles driven by a CPU disaster
Ryzen routed Kaby Lake, and proved itself much faster

Now, was Intel frightened? No! For this had happened once before
When Prescott launched — and shat the bed — back in 2004
Intel poured a cup of Coffee and admitted no mistake
Instead, it booked itself a stay on scenic Stagnant Lake.

Then Zen 2 debuted on 7, and it gave Team Red the lead
While Intel sat, stuck on 14, despite its growing need
An endless stream of plusses replaced Tick-Tock as a guide
And Intel turned to price cuts as its process scaling died.

Cascade’s workstation bloodline is writ plainly on the chip
Xeon owns this market, despite 10-nanometer’s slip.
Priced to move and thrice-refined, the core takes a batting stance
Its rival is irrelevant, Threadripper scarcely earns a glance.

And here’s the starting benchmark: It’s a render test well-known
But when the scores appear on-screen, the crowd lets out a moan.
Now that the run is over, Sunnyvale has clearly won,
“Cinebench!?” spat Cascade Lake, and the reviewer said: “Strike one!”

A second app is readied: a Qt compiler test
This type of application is where Chipzilla scores the best
Intel’s core stood watching, as the final tallies flew
“That’s not real world!” cried Cascade, while the reviewer called: “Strike two!”

All pretense gone, at ease no more, Cascade rethinks its tack
SIMD math may yet disrupt its rival’s product stack.
There’s one more test still left to score before the end of day
MATLAB is the perfect choice to keep its foe at bay.

The droops are gone from Cascade’s clock, its SIMD offsets all read “None.”
With thermal limits set aside, its heat could melt the sun.
The reviewer queues the MATLAB run, mouse pointer on “Begin”
Cascade leaps for the finish line — this task, it’s sure to win!

Oh, somewhere in this favored land, the sun is beaming down
Somewhere fanboys argue, and somewhere gamers frown
From Movidius to Mobileye, Intel still has clout
But as far as Threadripper’s concerned — mighty Cascade has struck out."


Are we going to see "Sapphire at the Bat’: A Bad Intel CPU Poem" in the next few years?

 

[DrMrLordX]

Intel lands Mediatek as an IFS 2.0 customer!!!!!

Intel to produce Taiwanese company MediaTek's chips

U.S. chipmaker Intel Corp said on Monday it will produce chips for Taiwan’s MediaTek Inc , one of the world's largest chip design firms.

www.reuters.com

Oh wait. Intel 16? Ummmm . . . is that a rebrand of 22FFL? Guess we won't see new Dimensity SoCs coming from Intel.

 

[Doug S]

Intel lands Mediatek as an IFS 2.0 customer!!!!!

Intel to produce Taiwanese company MediaTek's chips

U.S. chipmaker Intel Corp said on Monday it will produce chips for Taiwan’s MediaTek Inc , one of the world's largest chip design firms.

www.reuters.com

Oh wait. Intel 16? Ummmm . . . is that a rebrand of 22FFL? Guess we won't see new Dimensity SoCs coming from Intel.

It says SoCs used "for smart devices" so probably highly cost sensitive IoT stuff. Perhaps getting a better price than capacity constrained TSMC, since those older 22FFL processes are probably fairly idle at Intel these days as even their chipsets will be using a newer process by now.

 

[DrMrLordX]

It says SoCs used "for smart devices" so probably highly cost sensitive IoT stuff. Perhaps getting a better price than capacity constrained TSMC, since those older 22FFL processes are probably fairly idle at Intel these days as even their chipsets will be using a newer process by now.

Intel's running into low-margin territory here, yes. That much should have been obvious once they mentioned the process (which is the first I'm hearing of Intel 16).

 

[Doug S]

Intel's running into low-margin territory here, yes. That much should have been obvious once they mentioned the process (which is the first I'm hearing of Intel 16).

Sure but they have to start somewhere. It isn't reasonable to expect them to be able to sign up a major client for leading edge production when they have no real track record as a foundry, and have been quite publicly failing to deliver processes on schedule for the past five years.

Intel might need to agree to take on some of the risk for early customers to get customers they can promote. Fund the cost for them to port a design made on N4 or 3GAE to Intel 3 so they can dual source and directly compare performance. That way the customer eliminates risk because if Intel can't deliver on their promises they're fine, if Intel does deliver they will get some of those clients going all-in with Intel on subsequent products.

 

[maddie]

Sure but they have to start somewhere. It isn't reasonable to expect them to be able to sign up a major client for leading edge production when they have no real track record as a foundry, and have been quite publicly failing to deliver processes on schedule for the past five years.

Intel might need to agree to take on some of the risk for early customers to get customers they can promote. Fund the cost for them to port a design made on N4 or 3GAE to Intel 3 so they can dual source and directly compare performance. That way the customer eliminates risk because if Intel can't deliver on their promises they're fine, if Intel does deliver they will get some of those clients going all-in with Intel on subsequent products.

One has to smile, knowing the history of Apple being turned away so long ago.

 

[Thibsie]

Sure but they have to start somewhere. It isn't reasonable to expect them to be able to sign up a major client for leading edge production when they have no real track record as a foundry, and have been quite publicly failing to deliver processes on schedule for the past five years.

Intel might need to agree to take on some of the risk for early customers to get customers they can promote. Fund the cost for them to port a design made on N4 or 3GAE to Intel 3 so they can dual source and directly compare performance. That way the customer eliminates risk because if Intel can't deliver on their promises they're fine, if Intel does deliver they will get some of those clients going all-in with Intel on subsequent products.

But we're speaking about Intel, right ?

 

[DisEnchantment]

Samsung Foundry On EUV, Pellicles, Capacity, and Yield

Samsung Foundry talks about EUV, wafer capacity and mid-term plans,pPellicles, and advanced node yield.

fuse.wikichip.org

Some notes about Samsung Foundry Yields and other things like capacity, EUV etc.
7LPE yields started abysmally. But eventually by late last year things stabilized a lot.
5LPE and 4LPE don't suffer as much.
Defect density for 4LPE now reaching < 0.2

 

[dark zero]

It says SoCs used "for smart devices" so probably highly cost sensitive IoT stuff. Perhaps getting a better price than capacity constrained TSMC, since those older 22FFL processes are probably fairly idle at Intel these days as even their chipsets will be using a newer process by now.

Or maybe low tier Kompanio chips? Also they need to make new low tier Helio / Dimensity chips with that process.

 

[Exist50]

It says SoCs used "for smart devices" so probably highly cost sensitive IoT stuff. Perhaps getting a better price than capacity constrained TSMC, since those older 22FFL processes are probably fairly idle at Intel these days as even their chipsets will be using a newer process by now.

Pat Gelsinger:

I’ll make them as many Intel 16 chips as they want

Chip makers to carmakers: time to get out of the semiconductor Stone Age | Fortune

Intel, Qualcomm try to wean auto companies from dependence on obsolete technology

fortune.com

Sure sounds like they have spare capacity.

 

[Markfw]

Pat Gelsinger:

Chip makers to carmakers: time to get out of the semiconductor Stone Age | Fortune

Intel, Qualcomm try to wean auto companies from dependence on obsolete technology

fortune.com

Sure sounds like they have spare capacity.

The way I read it, they have lots of capacity in 16nm and maybe 14nm.

 

[jpiniero]

Samsung Foundry On EUV, Pellicles, Capacity, and Yield

Samsung Foundry talks about EUV, wafer capacity and mid-term plans,pPellicles, and advanced node yield.

fuse.wikichip.org

Some notes about Samsung Foundry Yields and other things like capacity, EUV etc.
7LPE yields started abysmally. But eventually by late last year things stabilized a lot.
5LPE and 4LPE don't suffer as much.
Defect density for 4LPE now reaching < 0.2

Very good. Now how long will it take Intel to get 7 nm down to that level? Will they ever?

 

[maddie]

Very good. Now how long will it take Intel to get 7 nm down to that level? Will they ever?

Intel 7 is worse than .2 defect rate?

 

[jpiniero]

Intel 7 is worse than .2 defect rate?

I believe so but I am talking about the node(s) formerly called 7 nm, ie: Intel 4 and 3.

 

[Doug S]

One has to smile, knowing the history of Apple being turned away so long ago.

Turned away for what? You mean for the original iPhone, or the endless rumors that Intel would fab Apple's chips?

Given the capacity issues Intel experienced simply supplying modem chips to Apple for a few years they're probably glad they never took on the task of fabbing Apple's SoCs.

 

[jpiniero]

Turned away for what? You mean for the original iPhone, or the endless rumors that Intel would fab Apple's chips?

Given the capacity issues Intel experienced simply supplying modem chips to Apple for a few years they're probably glad they never took on the task of fabbing Apple's SoCs.

The modem isn't as sexy but every iPhone needs one. So that's still a lot of chips. Alas it got 10 nm'd.