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.

 

[moinmoin]

TSMC backing off again. No end to their excuses. They just don't really want a US fab IMHO (even though they have a point about the lack of execution on the CHIPS funding).

TSMC delays 3nm Arizona fab by a year, cites lack of U.S. subsidies and waning demand

TSMC faces another setback in the U.S.

www.tomshardware.com

To be fair this is about phase 2 which is some form of N3, a node not really settled yet. Nobody wants a node that nobody wants. N4 as part of phase 1 is scheduled to go live in 2025.

 

[jpiniero]

TSMC's US efforts have been a disaster so far.

Diversifying out of Taiwan, esp because of the invasion threat, is a good idea. But perhaps they need to find a different country. One more to their liking.

 

[Thunder 57]

TSMC's US efforts have been a disaster so far.

Diversifying out of Taiwan, esp because of the invasion threat, is a good idea. But perhaps they need to find a different country. One more to their liking.

A big problem is the difference in culture. Americans don't want to slave away like they do back at TSMC.

 

[desrever]

A big problem is the difference in culture. Americans don't want to slave away like they do back at TSMC.

they should just go join a culture like China then

 

[Ajay]

they should just go join a culture like China then

They don't want their headquarters there. Though, if the USA didn't suggest we'd take our big stick and walk away - they'd have more fabs there.

 

[Aapje]

TSMC's US efforts have been a disaster so far.

Diversifying out of Taiwan, esp because of the invasion threat, is a good idea. But perhaps they need to find a different country. One more to their liking.

They could go to France.

A big problem is the difference in culture. Americans don't want to slave away like they do back at TSMC.

Americans are actually known for working pretty long hours for a Western nation. It's more that they are not very productive per hour, and often don't have very high standards (which I write on this US-run forum which is now giving errors all the time).

But Toyota famously took perhaps the worst GM car factory and turned it around completely:

 

[jpiniero]

https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2024-01/3e43ab2cb1ddad664ef9fe09c6ae80fa69167eea/4Q23Presentation%28E%29.pdf

From TSMC's earnings report, The "3 nm" nodes are 15% of their revenue... "7 nm" ones are only 17%.

Says N2 "volume production" next year and N2 with backside power delivery "available to customers in the second half of 2025" with production in 2026. So Apple should be able to use it for the 2026 iPhone.

 

[Aapje]

Interesting. TSMC is only reporting revenue from N3 starting in Q3 2023, even though they announced volume production in December 2022. So apparently it either took 3/4 of a year to actually get volume production going for realsies, or the rumor about having Apple as a customer for N3B is false and they had zero customers for that node. After all, they promised N3E production for 2H 2023. Assuming that is true, of course.

Also, I see that even in Q4 2023 they were still scaling up N5.

 

[Hitman928]

Interesting. TSMC is only reporting revenue from N3 starting in Q3 2023, even though they announced volume production in December 2022. So apparently it either took 3/4 of a year to actually get volume production going for realsies, or the rumor about having Apple as a customer for N3B is false and they had zero customers for that node. After all, they promised N3E production for 2H 2023. Assuming that is true, of course.

Also, I see that even in Q4 2023 they were still scaling up N5.

Apple and Intel are the only N3B customers and neither was ready for chips when N3B was ready for volume production. I believe TSMC only counts revenue once the customer actually receives the chips so it seems TSMC started sending volume chips to Apple right at the beginning of Q3. Intel isn't needing chips until later 2024.

 

[igor_kavinski]

Intel's German fab will be most advanced in the world and make 1.5nm chips, CEO says

Europe is set to get the world's most advanced fab for the first time in years.

www.tomshardware.com

Any predictions on what these 15A chips would be like?

Performance, power, area, heat generation?

How many cores we can look forward to? How much cache?

Is this going to be the process where almost 50% of the chip is cache and the rest is shrunken silicon?

Will just a mere volt be enough to kill these chips?

 

[Aapje]

Intel's German fab will be most advanced in the world and make 1.5nm chips, CEO says

Europe is set to get the world's most advanced fab for the first time in years.

www.tomshardware.com

The original plan was to be done building in 2027, but the first delay has already happened before they started, so best case is probably 2028.

TSMC plan N2 volume production in 2025, so TMSC's 15A should be done before 2028 already, right?

I smell the brown stuff from male cows.

 

[DrMrLordX]

A big problem is the difference in culture. Americans don't want to slave away like they do back at TSMC.

Americans can get the job done. They just want more pay to do so.

 

[Doug S]

https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2024-01/3e43ab2cb1ddad664ef9fe09c6ae80fa69167eea/4Q23Presentation%28E%29.pdf

From TSMC's earnings report, The "3 nm" nodes are 15% of their revenue... "7 nm" ones are only 17%.

Says N2 "volume production" next year and N2 with backside power delivery "available to customers in the second half of 2025" with production in 2026. So Apple should be able to use it for the 2026 iPhone.

Once I heard Ian's report that N2 with BSPDN comes six months after "base" N2 does, and knowing that N2 is H2 2025 (like N3 too late for Apple) it was clear that Apple's first use of N2 would be in 2026. They would want BSPDN for the design freedom and increased performance / decreased cost.

Anandtech's article seems to indicate that BSPDN comes with N2P but Ian was pretty clear that was separate from N2P (and maybe you can get N2P with and without BSPDN as well, we'll see) so I'm going with him.

So I'd assume Apple is doing N3E this year (and their first use of FinFlex which could benefit them beyond the small % improvements in performance/power for N3E vs N3B, another reason not to use the initial N2 which does not include it) and N3P next.

 

[Hitman928]

Intel and UMC just announced a deal for Intel to manufacture a 12 nm node for UMC at their Arizona fab site. UMC will handle all customer facing things (PDK, sales, support, etc.) and Intel will do the actual manufacturing.

Intel and UMC Announce New Foundry Collaboration

www.intc.com

 

[Ajay]

12nm?? I thought Intel had 10 and 14?

 

[Hitman928]

12nm?? I thought Intel had 10 and 14?

They are co-developing a 12 nm process. While not stated, I'm pretty sure it is just a tweaked UMC process based on their (UMC's) 14 nm node.

 

[DrMrLordX]

Intel sometimes has nodes with funky names based on ones we all know and love/hate. Like 22FFL which was based on 14nm, and then there's a 16nm node which I guess was also based on 14nm? My guess is that if Intel is collaborating on a 12nm, node, it'll be based on 10nm/Intel 7 somehow.

 

[moinmoin]

While not stated, I'm pretty sure it is just a tweaked UMC process based on their (UMC's) 14 nm node.

Why would Intel manufacture a UMC node? The PR mentions:

"The 12 nm node will utilize Intel’s U.S. based high-volume manufacturing capacity and experience in FinFET transistor design (...) The new process node will be developed and manufactured in Fabs 12, 22 and 32 at Intel’s Ocotillo Technology Fabrication site in Arizona. Leveraging existing equipment in these fabs will significantly reduce upfront investment requirements and optimize utilization."

Which sounds like an Intel node to me. For UMC this seem to be an easy way to extend its service portfolio while IFS gets to tap into UMC's existing customer base.

 

[clemsyn]

I hope they don't buy UMC like they did with Altera.

 

[moinmoin]

I hope they don't buy UMC like they did with Altera.

Intel seems to be intent on splitting off Altera again. Would it even have the money to buy UMC?

 

[gdansk]

Americans are actually known for working pretty long hours for a Western nation. It's more that they are not very productive per hour, and often don't have very high standards (which I write on this US-run forum which is now giving errors all the time).

In fact there are only 5 OECD countries with a higher productivity per hour worked than the United States and it's Ireland, Luxembourg, Norway, Switzerland, and Sweden. Japan and Korea are nowhere close. They work many hours. I have no idea about Taiwan, since they're not in the OECD, but it's probably closer to Japan than the US even with their recent success.

Source: OECD

 

[DavidC1]

They are co-developing a 12 nm process. While not stated, I'm pretty sure it is just a tweaked UMC process based on their (UMC's) 14 nm node.

Which sounds like an Intel node to me. For UMC this seem to be an easy way to extend its service portfolio while IFS gets to tap into UMC's existing customer base.

Why does it have to be one OR the other? It says it's a collaboration, a mix of both.

The 12 nm node will utilize Intel’s U.S. based high-volume manufacturing capacity and experience in FinFET transistor design, offering a strong combination of maturity, performance and power efficiency. The production will markedly benefit from UMC’s decades of process leadership and history of providing customers with Process Design Kit (PDK) and design assistance for effectively providing foundry services.

Americans can get the job done. They just want more pay to do so.

Ever heard of the phrase "there are no bad products, only bad prices?"

If someone will do the same job as you do but for less, who in the right mind hire the more expensive one?

Japan and Korea are nowhere close. They work many hours. I have no idea about Taiwan, since they're not in the OECD, but it's probably closer to Japan than the US even with their recent success.

Japan has been in decline for a while, which will hit most of the places.

The video is about Toyota though. They are excellent when it comes to reliability, no coincidence due to treating them as human rather than biological machines.

 

[DrMrLordX]

If someone will do the same job as you do but for less, who in the right mind hire the more expensive one?

Someone who needs redundancy in other countries.

 

[Hitman928]

Why would Intel manufacture a UMC node? The PR mentions:

"The 12 nm node will utilize Intel’s U.S. based high-volume manufacturing capacity and experience in FinFET transistor design (...) The new process node will be developed and manufactured in Fabs 12, 22 and 32 at Intel’s Ocotillo Technology Fabrication site in Arizona. Leveraging existing equipment in these fabs will significantly reduce upfront investment requirements and optimize utilization."

Which sounds like an Intel node to me. For UMC this seem to be an easy way to extend its service portfolio while IFS gets to tap into UMC's existing customer base.

It's possible that the base is an Intel node, I just think it's far more probable to be based on UMC's 14 nm node than Intel's. UMC's node will already have proven out 3rd party tools, quality external customer documentation, and support staff. As far as I know, this doesn't exist for Intel's 14 nm node. I think it makes far more sense for UMC to share their 14 nm recipe with Intel (this isn't something valuable to Intel) and tweak it a little with them which will require only minor updates to their flow, than it is for Intel to develop all of the design flow for 3rd party customers just for UMC to get a piece of the pie. This allows Intel to get use out of their aging foundries with minimal effort and UMC gets additional capacity, win-win.

Why does it have to be one OR the other? It says it's a collaboration, a mix of both.

Developing a totally new node takes significant time and money. You wouldn't do that when you have mature nodes already available to base it on. It's far cheaper and easier to develop an enhanced node based upon an existing mature node than to try and create a new node altogether.

 

[Thunder 57]

Why does it have to be one OR the other? It says it's a collaboration, a mix of both.

The 12 nm node will utilize Intel’s U.S. based high-volume manufacturing capacity and experience in FinFET transistor design, offering a strong combination of maturity, performance and power efficiency. The production will markedly benefit from UMC’s decades of process leadership and history of providing customers with Process Design Kit (PDK) and design assistance for effectively providing foundry services.

Ever heard of the phrase "there are no bad products, only bad prices?"

If someone will do the same job as you do but for less, who in the right mind hire the more expensive one?

Japan has been in decline for a while, which will hit most of the places.

The video is about Toyota though. They are excellent when it comes to reliability, no coincidence due to treating them as human rather than biological machines.

There are certainly bad prodcuts. Like power supplies that catch on fire.