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

 

[Win2012R2]

They did not go back they had Ampere on 8nm already and A78 as well they simply reused their IP.

It's only released now though, and will be produced for like next 10 years probably.

 

[511]

It's only released now though, and will be produced for like next 10 years probably.

yeah but they wanted it to be cheap so here it is only way to make cheaper products

 

[maddie]

Imagine how good it can be propped up by 18A, with expanded structures (internal caches and buffers of several hundred kilobytes approaching half a megabyte each or even more).

My analogy of Pentium 4 type CPUs and Conroe type CPUs.

Imagine a race car that can go screaming fast on straights but suffers a devastating slowdown on bends. That's P4. Then there's the car that does pretty good on straights but it does comparatively outstanding on bends. That's Conroe.

The thing with P4 type CPUs is that when you are speeding down that straight highway, the sense of speed is just beyond awesome.

My condolences. As a person older than you, I've seen many friends descend into nostalgic episodes similar to this. It's a one way ticket my friend.

 

[Io Magnesso]

My condolences. As a person older than you, I've seen many friends descend into nostalgic episodes similar to this. It's a one way ticket my friend.

To be honest, in my generation, I'm a Core i generation. I'm interested in nostalgic stories
The first generation Core I was the first laptop CPU I touched.

 

[Io Magnesso]

No way, Intel 18A-P's "P" is not the performance P... Is it P as "+"...?
It is written that mobile is also a target, so it may not have just improved performance.
For example Perhaps the IP is quite substantial compared to the original 18A?

 

[Josh128]

No way, Intel 18A-P's "P" is not the performance P... Is it P as "+"...?
It is written that mobile is also a target, so it may not have just improved performance.
For example Perhaps the IP is quite substantial compared to the original 18A?

It means "Amazing-Performance".

 

[Io Magnesso]

It means "Amazing-Performance".

Intel 18 "Ångstrom - Plus"

 

[Josh128]

Intel 18 "Ångstrom - Plus"

So we have the first 18A+ before 18A is even out. Amazing work Intel. I cant wait for 18A-PPP, that ones going to be a doozy!

 

[Io Magnesso]

So we have the first 18A+ before 18A is even out. Amazing work Intel. I cant wait for 18A-PPP, that ones going to be a doozy!

Simply I was guessing the meaning of the "P" part of Intel 18AP.
I didn't say that Intel 18AP would come first
Stop talking like I said what I didn't say
Is it a joke? , not interesting

 

[The Hardcard]

Imagine how good it can be propped up by 18A, with expanded structures (internal caches and buffers of several hundred kilobytes approaching half a megabyte each or even more).

My analogy of Pentium 4 type CPUs and Conroe type CPUs.

Imagine a race car that can go screaming fast on straights but suffers a devastating slowdown on bends. That's P4. Then there's the car that does pretty good on straights but it does comparatively outstanding on bends. That's Conroe.

The thing with P4 type CPUs is that when you are speeding down that straight highway, the sense of speed is just beyond awesome.

The problem with this is that a small percentage of code has long straits. Typically there is a branch every six instructions, which leaves just a few mere nanoseconds to exhilarate in straightway speed.

I think that played as much a role in the collapse of NetBurst as the energy wall. What’s the point of blasting through instructions being your main party trick when long instruction streams are not the common case? Not to mention the significant percentage of long instruction streams that are even faster with some form of parallel ALUs (SIMD, GPU, etc.)

 

[igor_kavinski]

I think that played as much a role in the collapse of NetBurst as the energy wall.

The main problem I remember was the heat that even non-tech users were complaining about and making fun of in that era. 18A should alleviate that. A modern 8C16T P4 working at 6.5 GHz with AVX-512 should still be smaller and less power hungry than an Alder Lake CPU. It would make a great budget CPU.

 

[Thunder 57]

The main problem I remember was the heat that even non-tech users were complaining about and making fun of in that era. 18A should alleviate that. A modern 8C16T P4 working at 6.5 GHz with AVX-512 should still be smaller and less power hungry than an Alder Lake CPU. It would make a great budget CPU.

So throw away the uop cache/L1i cache and go back to a trace cache? We are beating a dead horse here.

 

[poke01]

A modern 8C16T P4 working at 6.5 GHz with AVX-512 should still be smaller and less power hungry than an Alder Lake CPU. It would make a great budget CPU.

yeah thats called Zen6. We don't need P4. We will have great x86 CPUs in 2026

 

[Tigerick]

Does Samsung have any major customers for SF3-family nodes anymore? Who actually uses it?

edit: also . . .

MSN

not sure if anyone posted this already. Meh!

Yep, I am about to post that news which should be related to AMD's cancellation of all projects with SF:

Link:

AMD had reportedly been working closely with Samsung on the SF4X process—not only for EPYC server CPUs, but also for Ryzen APUs and Radeon GPUs—as part of its dual-sourcing strategy. However, that collaboration now appears to be collapsing, the report indicates.

While it remains unclear if the move affects only EPYC chips, AMD’s overall confidence in Samsung Foundry seems to be waning. The company has chosen not to proceed with mass production on Samsung’s 4nm process for GPUs due to concerns over process stability, as suggested by the report.

That's mean Sonoma Valley which is supposed to replace Mendocino is changing from SF to TSMC's N4C process. I really don't understand how hard could 4xZen5c with 2CU SoC be made?

No wonder Samsung is looking to improve their 4nm as well which is supposedly to be considered as mature node.

 

[DrMrLordX]

No wonder Samsung is looking to improve their 4nm as well which is supposedly to be considered as mature node.

Oh dear. N4X was a chance for Samsung to make some inroads against N4P and/or N4C and it looks to have gone sideways.

 

[Win2012R2]

I really don't understand how hard could 4xZen5c with 2CU SoC be made?

Or AMD just used that as leverage to secure better TSMC pricing.

 

[fastandfurious6]

I believe the Switch 2's SoC taped out in 2021.

only Nintendo can afford to use such outdated hardware and still compete in big sales... crazy overuse of brand power won't last long. how many Mario karts and donkey kongs lol

 

[LightningZ71]

Nintendo's targe audience is NOT hardcore gamers! It hasn't been in recorded history. They are aiming for people that are into more casual gameplay, that want long battery life, that want variety without intensity. They cater to those that like to do both by providing a platform for those other games to be ported to the switch ecosystem, but it is well known that performance is often lacking. This SoC upgrade is as much about keeping the hardware current for multi-platform and vendor support as it may be about performance.

 

[511]

So we have the first 18A+ before 18A is even out. Amazing work Intel. I cant wait for 18A-PPP, that ones going to be a doozy!

lol can't read the slides and where do you see 18AP before 18A also your joke need improvements replace P with '+' like 18A+++.
a '+' is a better process meme template for intel than 'P'.

Or AMD just used that as leverage to secure better TSMC pricing.

AMD is a close TSMC Customer they won't be leaving for anything else anytime soon they wanted to increase margin due to Samsung being cheaper but that 4nm from Samsung is not good also talking about leverage is not happening unless Intel/Samsung cooks something good for external.

18A is Intel Internal use and from what i know there is a intel only PDK that has better Power and Performance than the thing shared externally due to design flow difference. 14A is fully Industry standard EDA with none of the Intel customization.

Oh dear. N4X was a chance for Samsung to make some inroads against N4P and/or N4C and it looks to have gone sideways.

TSMC increased price on every node so N4C cost as much as N4P used to cost.

 

[Doug S]

Nintendo's targe audience is NOT hardcore gamers! It hasn't been in recorded history. They are aiming for people that are into more casual gameplay, that want long battery life, that want variety without intensity. They cater to those that like to do both by providing a platform for those other games to be ported to the switch ecosystem, but it is well known that performance is often lacking. This SoC upgrade is as much about keeping the hardware current for multi-platform and vendor support as it may be about performance.

Yep it could be as simple as "we need to upgrade the memory controller because soon we won't be able to buy the RAM it interfaces with". So you have to spin a new SoC but that doesn't mean you have to be aggressive in your choice of process or upgrading the CPU/GPU performance. It is going to continue to play mostly the same kind of games because that's what its audience wants. Its like comfort food for gamers.

 

[johnsonwax]

yeah thats called Zen6. We don't need P4. We will have great x86 CPUs in 2026

Good lord that reads like one of those predictions CEOs make 18 months before they declare bankruptcy.

 

[poke01]

Good lord that reads like one of those predictions CEOs make 18 months before they declare bankruptcy.

Now that I read it again, it totally does.

🤣🤣

 

[johnsonwax]

only Nintendo can afford to use such outdated hardware and still compete in big sales... crazy overuse of brand power won't last long. how many Mario karts and donkey kongs lol

Well, they've been using withered tech since at least the Game Boy in the late 80s, so about 40 years so far. Their IP is some of the strongest in the world.

But this is where a lot of investors often go wrong. The most reliable part of the movie industry isn't the summer blockbusters but the b movies - the Sharknados, etc. They basically never lose money, they're made cheap as hell, and they crank out loads of them. They're very low risk because of the low cost to produce. That's kind of Nintendo. They don't take on a lot of risk with cutting edge tech and focus less on groundbreaking raytraced open world games and on ones with familiar characters and settings that are just really fun to play. Minecraft is the most popular game in the world and it's kinda shit (I play a lot of it) but it's fun. Balatro was a huge hit - single dev card game. Almost everyone has played Stardew Valley - also a single dev game. You want a good ROI - having a one man dev shop make a game in a year and take home $15M after Steams cut is pretty damn good. None of the GTA6 devs are going to walk home with $15M.

 

[johnsonwax]

Now that I read it again, it totally does.

🤣🤣

I'm calling it here - poke01 has inadvertently incited gods wrath to ruin Zen6.

Sorry everyone, these things happen.

 

[igor_kavinski]

AMD launches even their best have some drama anyway. Can't wait for the Zen 6 drama to unfold.