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.

TSMC Radically Boosts CapEx to Expand Production Capacities, To Reach $14B For 2019

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.

Samsung’s Aggressive EUV Plans: 6nm Production in H2, 5nm & 4nm On Track

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.

 

[SiliconFly]

TSMC N3E has been analyzed by TechInsights.
As like you had known, there are SRAM density regression and logic density regression (c.f. N3B)...

A little bit of density regression was kinda expected right? So, I don't think it's a big surprise.

 

[Ghostsonplanets]

There's apparently strong rumors that AMD's next budger line chip will be a samsung product. I hope that it's on an already established node.

Yes. A Navi 5 die and Sonoma Valley on SF4X.

Is Semiaccurate implying that Nvidia's ARM PC SoC will be made at Intel Foundry?

ARM Outs Their New IP lineup for 2024

Last week ARM unveiled their new IP that will be featured in the devices consumers buy in 2025.

semiaccurate.com

They're not implying. They're the ones who are saying Nvidia SoC is fabbed at Intel. The Xpea guy is the one who is saying it's fabbed on TSMC.

Just for clarity, but would that be the 3GAP node which is now . . . SF3 I think?

That's correct. If SFoundry fails once again at delivering an Exynos to the market and with their MCBFET based node, then it's time to seriously discuss if Samsung Foundry will exist at the leading edge at the near future.

 

[LightningZ71]

it's arguable that they aren't at the leading edge. They are only shipping in volume nodes that are N-2 from TSMC if we count N3E and N3B as shipping in volume and N4P/X as effectively leading edge nodes. To my knowledge, Samsung doesn't have a volume node with usable yields that are similar in density/performance/power to any of those nodes. They have advertised nodes and some products on roughly "4nm class" nodes, but, I don't recall them being as good as what's out there.

 

[trivik12]

Samsung going all Snapdragon for S25 means their 3nm class nodes are bust at this point. That is unfortunate. We need competition for TSMC. Otherwise they will keep hiking prices every year like a clockwork.

 

[Henry swagger]

Intel 3 Process Node Detailed: 18% More Performance At Same Power, 10% Higher Density, Shipping With Xeon 6 CPUs Now

Intel detailed its next-gen Intel 3 process node which brings a solid 18% performance uplift over Intel 4 along with density improvements.

wccftech.com

Intel 3 is impressive.. 18A should be even better

 

[SiliconFly]

Intel 3 Process Node Detailed: 18% More Performance At Same Power, 10% Higher Density, Shipping With Xeon 6 CPUs Now

Intel detailed its next-gen Intel 3 process node which brings a solid 18% performance uplift over Intel 4 along with density improvements.

wccftech.com

Intel 3 is impressive.. 18A should be even better

Funny we're talking about Intel 3, when 20A is around the corner. 18% performance figure was published long back and is in line with expectations. Nothing new here.

And 18A (with BSPD) is expected to offer a whopping ~27% increase over Intel 3 putting it in front of the competition.

And to be a successful node, 18A need to hit volume in a decent timeframe. If 20A can do that, we can expect 18A to follow suit. If 20A hits delays, 18A might face the same. Only time will tell!

 

[Ajay]

And to be a successful node, 18A need to hit volume in a decent time frame. If 20A can do that, we can expect 18A to follow suit. If 20A hits delays, 18A might face the same

That's the whole ball of wax. Deliver on time with expected PPA (+/-) and many things will open up for Intel. Don't and we may be saying goodbye to Gelsinger.

 

[Ghostsonplanets]

Intel 3 Process Node Detailed: 18% More Performance At Same Power, 10% Higher Density, Shipping With Xeon 6 CPUs Now

Intel detailed its next-gen Intel 3 process node which brings a solid 18% performance uplift over Intel 4 along with density improvements.

wccftech.com

Intel 3 is impressive.. 18A should be even better

Some very nice improvements. 18% PPW is a nice uplift, alongside a complete node with HD cells.

I'm curious to see how Intel 3 will translate into consumer chips wih ARL-U. And we'll be able to make some nice comparison between Intel 4 MTL-U and Intel 3 ARL-U.

 

[DrMrLordX]

Yes

That's correct. If SFoundry fails once again at delivering an Exynos to the market and with their MCBFET based node, then it's time to seriously discuss if Samsung Foundry will exist at the leading edge at the near future.

Ugh. 3GAE getting mothballed was somewhat forgivable (within the context of their previous nodes being mediocre-to-bad), but 3GAP/SF3 suffering yield issues is completely unacceptable.

 

[controlflow]

Intel Delivers Leading-Edge Foundry Node with Intel 3 Technology; on Path Back to Process Leadership

Editor’s Note: Welcome to the new Intel Foundry blog! We're launching this resource as a new way to peek inside our systems foundry for the AI era with some of the experts making it happen. We're thrilled to launch the blog with this exciting look at the Intel 3 node’s performance gains from...

community.intel.com

Good information and details in the video Intel posted here on Intel 3. Pretty impressive. Intel's execution on process seems to be getting back on track for real now. Should be really interesting to see what 20A and 18A bring.

 

[igor_kavinski]

Pretty impressive.

Lastly, the Intel 3-PT node combines all these advances together into a single process and then adds in even more performance enhancements alongside superior ease-of-use for designers, while including support for finer-pitch 9um TSVs and hybrid bonding options for even higher-density 3D-stacking. We believe that the Intel 3-PT node delivers a unique combination of performance, flexibility, and cost for a wide variety of applications. As the ultimate FinFET-based process node, it will be a mainstay and employed alongside new technologies for many years to come - for both internal and external foundry customers.

This would be impressive if they release client something on it.

Any ideas what internal products would utilize this node going forward (their own comment saying "many years to come")?

 

[FlameTail]

Good information and details in the video Intel posted here on Intel 3. Pretty impressive. Intel's execution on process seems to be getting back on track for real now. Should be really interesting to see what 20A and 18A bring.

Good. TSMC's hiking prices should be put to a stop.

 

[lightisgood]

Intel 3 Process Node Detailed: 18% More Performance At Same Power, 10% Higher Density, Shipping With Xeon 6 CPUs Now

Intel detailed its next-gen Intel 3 process node which brings a solid 18% performance uplift over Intel 4 along with density improvements.

wccftech.com

Intel 3 is impressive.. 18A should be even better

I'm impressed with Intel 3's FinFET performance@1.1V ...
Intel finally breaks the barrier of 5GHz.
I'm sure I3 is outstanding HPC process, today.

BTW, Intel says that I3 will have >10 years life time.
I wonder if Intel will implement UHD library for I3?
If intel planning it, I3's PPA should be evenly matched with N3.

Anyway, it is beginning the fantastic race of Intel vs. TSMC.

 

[DrMrLordX]

Good. TSMC's hiking prices should be put to a stop.

Intel can't compete with TSMC on a volume basis. They can't even come close. Plus they're still taking N3 wafers, and they may order up N2 as well.

 

[Ghostsonplanets]

Good. TSMC's hiking prices should be put to a stop.

Intel neither has the volume nor the proven competitive advantage for it to make any meaningful dent into TSMC share.

It will take some time and Intel Foundry will need to prove themselves as a reliable partner.

 

[lightisgood]

Intel neither has the volume nor the proven competitive advantage for it to make any meaningful dent into TSMC share.

It will take some time and Intel Foundry will need to prove themselves as a reliable partner.

First of all, Intel isn't up to win TSMC as the foundry company.
Intel forecast their 2030's revenue is about $100B and $10~15B of there is the revenue from external customer.

I simply think Intel is up to win TSMC as the IDM.
IDM 2.0 is IDM 2.0, not foundry 2.0.

 

[Doug S]

First of all, Intel isn't up to win TSMC as the foundry company.
Intel forecast their 2030's revenue is about $100B and $10~15B of there is the revenue from external customer.

I simply think Intel is up to win TSMC as the IDM.
IDM 2.0 is IDM 2.0, not foundry 2.0.

Intel built their fabs up until now based on their own planned needs. A revenue mix where 10-15% is external would be consist with their existing plan where they built enough capacity for their own needs, plus a margin for error. They are likely being conservative and only forecasting external revenue they already have commitments for.

The huge Ohio investment is the expansion for foundry capacity (not that it will be exclusive to that, that's just where the main expansion is) and they've apparently cut the initial investment and are now planning 2028 to open up the first fab there. They have kind of a chicken and egg problem, in that it doesn't make sense to make a major investment for new capacity without major external demand, but they can't seek out any major customers to generate that level of demand until those customers know Intel will have the capacity. In addition that type of customer will want to see results, so they aren't even going to consider signing anything until they see production 18A silicon to know Intel is able to execute on their claims.

So if you assume 18A optimistically shipping production chips at the end of 2025, with early 14A perhaps running a few test/risk wafers and timelines for 10A available to interested parties only then could Intel begin to hope to sign up major customers for 10A in 2030. That would give them certainty to spend the billions needed to accelerate buildout of Ohio to have that capacity ready. The other alternative is taking more VC money. That lets them build out more quickly, but they have to share the profits with those outside investors.

 

[Geddagod]

Some very nice improvements. 18% PPW is a nice uplift, alongside a complete node with HD cells.

I'm curious to see how Intel 3 will translate into consumer chips wih ARL-U. And we'll be able to make some nice comparison between Intel 4 MTL-U and Intel 3 ARL-U.

Apparently neither ARL-U or GNR/SRF use the 2-2 HD option 🤣
Maybe PTL iGPU will use it.

If intel planning it, I3's PPA should be evenly matched with N3.

Reeeeeee

So if you assume 18A optimistically shipping production chips at the end of 2025,

Faraday says they will have 18A chips available by 1H 2025. Press X to doubt.

 

[Doug S]

More on Samsung's 3nm woes. Sounds like Exynos isn't cancelled yet but they'll have to if they can't get yields up.

https://www.trendforce.com/news/202...edly-below-20-struggling-for-mass-production/

 

[Ajay]

20%, that’s just awful. What is going on at that fab?

 

[lightisgood]

More on Samsung's 3nm woes. Sounds like Exynos isn't cancelled yet but they'll have to if they can't get yields up.

https://www.trendforce.com/news/202...edly-below-20-struggling-for-mass-production/

So, Intel should have the world first GAA process for commercial on Intel 20A, right?
Intel 20A/18A have good yield curves today.

 

[FlameTail]

So, Intel should have the world first GAA process for commercial on Intel 20A, right?
Intel 20A/18A have good yield curves today.

And Samsung was hyping up their 3nm GAA since 2020 as being the first GAAFET node.

 

[Ghostsonplanets]

20%, that’s just awful. What is going on at that fab?

Mind you, this is the second iteration of Samsung 3nm. The original 3GAE/SF3 was used for a extremely low-yield cryptomining chip and then disappeared. 3GAP/SF3P was then brought forward and renamed as SF3. And they yet are having yields problems...

I'm symphatetic towards Samsung Foundry. But they have been hit with problems after problems ever since the EUV transition. With exception of 5LPE/P and 4LPP+/SF4P, all their EUV nodes have been extremely problematic (4LPE, 7LPE, 3GAE).

Samsung went from having Apple, Nvidia and Qualcomm manufacturing their premier chips in their foundry to being snubbed by everyone. Even Intel doesn't take them seriously when talking about foundry ambitions.

If 3GAP/SF3 fails and Exynos 2500/Dream is a disaster, I see no path for Samsung in the future of leading edge manufacturing. Yet, they're already talking about SF2 and SF1.4...

 

[FlameTail]

"All talk but no action"

Samsung Foundry

 

[adroc_thurston]

So, Intel should have the world first GAA process for commercial on Intel 20A, right?

i20A exists in the most nominal sense of that word (it basically does not).
Now, i18A is both CWF and PTLs and yes it will be the first commercial GAA + BSPDN anything on the market!