TSMC's N3E Node Said to Have Good Yields, Volume Production Expected Q2 2023

[TheLostSwede]

Back in March there were reports of TSMC's N3E node having been moved from 2024 to the end of 2023. However, it seems like the node is already seeing better than expected yields and is now being pulled in further and TSMC is expecting to start volume production as early as Q2 in 2023. The node does appear to have been frozen when it comes to further development as of the end of March. Yields are said to be much higher than the N3B node, which is also under development, but with limited information available about it.

The first customer for the new node is expected to be Apple, as the company is largely paying for much of the cutting edge node development at TSMC. However, both Intel and Qualcomm are said to be some of the first customers for the new node. More details should hopefully be announced tomorrow during TSMC's first quarter earnings call. The N3E node is a reduced layer EUV process, but before it goes into mass production, it's likely we'll be seeing the N3 node first. Early production of 3 nm parts later this year is expected to be at around 10 to 20k wafers per month initially, rising to about 25 to 35k a month once TSMC's new fab is ready. Once the N3E node is in full swing, the monthly capacity of 3 nm parts should be around 50k wafers a month, but depending on customer demand, it might end up being even higher.

View at TechPowerUp Main Site | Source

 

[Deleted member 24505]

How did Intel get ahead of AMD in the queue?

 

[TheLostSwede]

How did Intel get ahead of AMD in the queue?

They paid more I guess?
The customer list of obviously not complete, nor from TSMC, it's just companies that are expected to want access to the N3E node early on.

 

[Deleted member 24505]

They paid more I guess?
The customer list of obviously not complete, nor from TSMC, it's just companies that are expected to want access to the N3E node early on.

I though AMD would have been ahead, that's always the impression i got

 

[ghazi]

As expected, TSMC is way ahead by sticking with FinFET for the 3nm generation. Let's see if Samsung's node manages to perform any better.

 

[windwhirl]

I though AMD would have been ahead, that's always the impression i got

Nah, that would be Apple. AMD does use and pay for advanced nodes, but the only one consistently using the most "bleeding-edge" node is Apple. AMD tends to follow right behind though, they don't really delay much before jumping into whatever is the newest node.

 

[Steevo]

Intel needs the better process to be competitive, their GPU is already being built on 6nm and I bet they are hungry for an advantage over AMD. I just hope that AMD and Nvdia get some line capacity too.

 

[mechtech]

Good news.................hopefully it won't end up like Samsung poor yield news later on.............

 

[ModEl4]

Intel probably is going to be competitive with 18A vs TSMC's 2nm (a little bit better in performance/watt, a little bit worst in performance and worst in density).
Before a few days they announced that they moved 18A risk production from H1 2025 to H2 2024, still i don't think less than 9 months between generations in the client space make sense, so probably we will have the following regarding launches:
Q4 2021 Alder Lake 7nm
Q3 2022 Raptor Lake 7nm
Q2 2023 Meteor Lake 4nm
Q1 2024 Arrow Lake 3nm
Q4 2024 Lunar Lake 20A
Q3 2025 Nova Lake 18A
The rumour goes that Arrow i9 will be 8P+32E cores and that Nova will be their most extensive redesign since Core era, with +50% performance vs Lunar, or something like that, we will see.

 

[simlife]

Intel probably is going to be competitive with 18A vs TSMC's 2nm (a little bit better in performance/watt, a little bit worst in performance and worst in density).
Before a few days they announced that they moved 18A risk production from H1 2025 to H2 2024, still i don't think less than 9 months between generations in the client space make sense, so probably we will have the following regarding launches:
Q4 2021 Alder Lake 7nm
Q3 2022 Raptor Lake 7nm
Q2 2023 Meteor Lake 4nm
Q1 2024 Arrow Lake 3nm
Q4 2024 Lunar Lake 20A
Q3 2025 Nova Lake 18A
The rumour goes that Arrow i9 will be 8P+32E cores and that Nova will be their most extensive redesign since Core era, with +50% performance vs Lunar, or something like that, we will see.

expect all of that wont matter tech is dying only reason ppl still buy phone the battery isnt replaceable 12 gb phones were out like 2 years ago..years ago yet a 4 gb ram switch is the most popular tech item in forever the switch even though its weaker then the 2013 consoles that came out 4 years earlier im not wrong look at the the steam hardware ppl are using 2016 low end hardware or weaker then that as a majoritey if 2k is all you need(no way they are at that) even without dlss every game can run it... most phones are more powerful then the swtich... ringt now and 8k mathamticly is bad no matter the viewing distance

 

[watzupken]

I though AMD would have been ahead, that's always the impression i got

AMD’s strategy when they started using TSMC is likely to stick with the second most cutting edge node. Thus, they have been sticking around with 6/7nm while 5nm was the best that was offered by TSMC. So no surprises here either. There is little reason to try and compete with Apple for cutting edge node allocation unless you have deep pockets. Not sure what will Intel be using TSMC 3nm for though because whatever they produce on it is unlikely to be cheap, and GPU is quite unlikely since the current ARC is expected to be using 6nm(7nm) TSMC node.

 

[ModEl4]

AMD’s strategy when they started using TSMC is likely to stick with the second most cutting edge node. Thus, they have been sticking around with 6/7nm while 5nm was the best that was offered by TSMC. So no surprises here either. There is little reason to try and compete with Apple for cutting edge node allocation unless you have deep pockets. Not sure what will Intel be using TSMC 3nm for though because whatever they produce on it is unlikely to be cheap, and GPU is quite unlikely since the current ARC is expected to be using 6nm(7nm) TSMC node.

3nm TSMC GPU is confirmed (at least IGP, if not all arc series)

https://www.screenhacker.com/screensite/wp-content/uploads/2022/02/dutsR8qBw5Co4HAs.jpg

The peculiar thing is that the official slide suggests 2024 Arrow Lake comes with TSMC N3 graphics which is extremely early!

expect all of that wont matter tech is dying only reason ppl still buy phone the battery isnt replaceable 12 gb phones were out like 2 years ago..years ago yet a 4 gb ram switch is the most popular tech item in forever the switch even though its weaker then the 2013 consoles that came out 4 years earlier im not wrong look at the the steam hardware ppl are using 2016 low end hardware or weaker then that as a majoritey if 2k is all you need(no way they are at that) even without dlss every game can run it... most phones are more powerful then the swtich... ringt now and 8k mathamticly is bad no matter the viewing distance

I don't know if tech is dying because sales doesn't suggest that, but we are starting to see an acceleration in performance/price point in relation with what was thought probable before let's say 3 years. If you are trying to suggest that advancement in tech keeps getting harder to convince people to upgrade because for some people the level we are now is good enough for many people, yes OK maybe (the economy also plays a major role for many people not deeming necessary to upgrade), certainly many companies trying to artificially limit their designs somehow in order to force their customer base to upgrade sooner than later, whether this is phone batteries, or windows compatibility-support or GPU ram (3070-3080) etc, but the switch success is a special case for many reasons, if Microsoft had tried to launch a console with switch hardware level capabilities probably would have died a miserable death, the same would be true if Nintendo was trying to compete with switch level hardware in a more traditional style console design like PS5/XBOX SX. But anyway i see your point! (btw your writing style reminds me someone in Anandtech, i used to post there 13 years ago, I don't remember the name, but his avatar's picture was a hawk if i remember correctly)

 

[Deleted member 24505]

With Intel spending so much on new FABS and AMD seemingly spending nothing and relying on TSMC imo Intel has the potential to be in the lead within the next 5 years. AMD needs to start thinking about spending some of the RYZEN profits on their own FAB, it does not pay to rely on someone else. Intel is only using TSMC till they new FABS come to fruition then imo they will pull out and sort their own stuff out.

 

[windwhirl]

With Intel spending so much on new FABS and AMD seemingly spending nothing and relying on TSMC imo Intel has the potential to be in the lead within the next 5 years. AMD needs to start thinking about spending some of the RYZEN profits on their own FAB, it does not pay to rely on someone else. Intel is only using TSMC till they new FABS come to fruition then imo they will pull out and sort their own stuff out.

I disagree. The cost for AMD of making fabs would be in the dozens if not hundreds of billions. Financially risky.

TSMC to Spend $100B on Fabs and R&D Over Next Three Years: 2nm, Arizona Fab & More

www.anandtech.com

TLDR, TSMC is investing 100 billion dollars (100.000.000.000) between R&D and the fabs themselves, between 2021 and 2023. I'm not sure AMD can invest that much money.

 

[Deleted member 24505]

I disagree. The cost for AMD of making fabs would be in the dozens if not hundreds of billions. Financially risky.

TSMC to Spend $100B on Fabs and R&D Over Next Three Years: 2nm, Arizona Fab & More

www.anandtech.com

TLDR, TSMC is investing 100 billion dollars (100.000.000.000) between R&D and the fabs themselves, between 2021 and 2023. I'm not sure AMD can invest that much money.

So they will always be reliant on TSMC whereas Intel will not. It will pay off for Intel at some point, there is no doubt imo.