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

 

[igor_kavinski]

Oh wow that takes me way back not personally. I'm guessing this is pre pentium 1 era?

Radiation-hardened Pentium&#174 processor

Finally, a good use for Intel’s lowly Atom CPU: Inside NASA’s $2.7B Perseverance rover

Low bid contract: Nasa's Mars probe packs Intel Atom and a 1990s iMac CPU.

www.pcworld.com

Why NASA’s newest space shuttle uses a computer chip from 2002

The Orion, which took its first unmanned test flight Dec. 5, launching from Cape Canaveral, Florida and orbiting for four and a half hours before splashing into the Pacific Ocean, is NASA’s latest human-rated spacecraft project, and probably the most ambitious ever. Its ultimate goal, like the...

qz.com

Why spacecraft use processors from the 20th century

Spacecraft use outdated microprocessors that were developed back in the 20th century. We'll you why this is happening.

root-nation.com

Bloomberg - Are you a robot?

How NASA gave birth to modern computing—and gets no credit for it

The space agency and MIT’s bet on the integrated circuit kicked off the digital age.

www.fastcompany.com

Enjoy!

 

[A///]

Radiation-hardened Pentium&#174 processor

Finally, a good use for Intel’s lowly Atom CPU: Inside NASA’s $2.7B Perseverance rover

Low bid contract: Nasa's Mars probe packs Intel Atom and a 1990s iMac CPU.

www.pcworld.com

Why NASA’s newest space shuttle uses a computer chip from 2002

The Orion, which took its first unmanned test flight Dec. 5, launching from Cape Canaveral, Florida and orbiting for four and a half hours before splashing into the Pacific Ocean, is NASA’s latest human-rated spacecraft project, and probably the most ambitious ever. Its ultimate goal, like the...

qz.com

Why spacecraft use processors from the 20th century

Spacecraft use outdated microprocessors that were developed back in the 20th century. We'll you why this is happening.

root-nation.com

Bloomberg - Are you a robot?

How NASA gave birth to modern computing—and gets no credit for it

The space agency and MIT’s bet on the integrated circuit kicked off the digital age.

www.fastcompany.com

Enjoy!

Already found much older docs on soi/sos development that predates all that.

 

[Roland00Address]

I know nothing, thus me asking this question. Any good information about the recent 7nm China SMIC process? Bloomberg has a temporary exclusive and it is paywalled.

The company's Mate 60 Pro is powered by SMIC's 7nm chips, according to analysis that TechInsights conducted for Bloomberg News.

 

[A///]

I know nothing, thus me asking this question. Any good information about the recent 7nm China SMIC process? Bloomberg has a temporary exclusive and it is paywalled.

The company's Mate 60 Pro is powered by SMIC's 7nm chips, according to analysis that TechInsights conducted for Bloomberg News.

Techninsights is the company that's disassembling and verifying here whereas bloomberg is reporting the findings. it stands to reason that smic has successfully made a 7nm process using asml duv machines which they can still purchase and are purchasing en masse from asml until Jan 1 when their cutoff becomes official. Ignoring the political nature and common claims against china it stands to reason that they may very well be able to side step obstacles at the moment. the kirin 7nm procesor yields are unknown but if they're above 75pc I'd consider that very good for how restricted tech has become for them.

I have my own opinions on the nature of bans or export restrictions but because it's such a touchy subject and can easily go off the wall I'll avoid saying anything more other than my hats off to smic for getting to this point on their lonesome.

It was well known that SMIC had been offering up 3-5x the salary amount to TSMC engineers to come over and they did well on that promise. It's not illegal to poach engineers.

 

[Ajay]

I know nothing, thus me asking this question. Any good information about the recent 7nm China SMIC process? Bloomberg has a temporary exclusive and it is paywalled.

The company's Mate 60 Pro is powered by SMIC's 7nm chips, according to analysis that TechInsights conducted for Bloomberg News.

Link to full archived article - https://archive.ph/HMuf1

have my own opinions on the nature of bans or export restrictions but because it's such a touchy subject and can easily go off the wall I'll avoid saying anything more other than my hats off to smic for getting to this point on their lonesome.

They had some help (like from ASML and others in the supply chain), but, no useful semi manufacturing partners that I'm aware of.

 

[A///]

Link to full archived article - https://archive.ph/HMuf1


They had some help (like from ASML and others in the supply chain), but, no useful semi manufacturing partners that I'm aware of.

As does TSMC, Intel, GloFlo, etc. Who is it you think that calibrates those machines for these companies?

 

[Roland00Address]

Techninsights is the company that's disassembling and verifying here whereas bloomberg is reporting the findings.

I know that 👍, but that was the google preview so people can get on the same page as me 🙂 (for I assume people do not stay up to date with the news, for I myself can not)

it stands to reason that smic has successfully made a 7nm process using asml duv machines which they can still purchase and are purchasing en masse from asml until Jan 1 when their cutoff becomes official. Ignoring the political nature and common claims against china it stands to reason that they may very well be able to side step obstacles at the moment. the kirin 7nm procesor yields are unknown but if they're above 75pc I'd consider that very good for how restricted tech has become for them.

I have my own opinions on the nature of bans or export restrictions but because it's such a touchy subject and can easily go off the wall I'll avoid saying anything more other than my hats off to smic for getting to this point on their lonesome.

It was well known that SMIC had been offering up 3-5x the salary amount to TSMC engineers to come over and they did well on that promise. It's not illegal to poach engineers.

 

[Ajay]

As does TSMC, Intel, GloFlo, etc. Who is it you think that calibrates those machines for these companies?

I know who does. The later part of the comment was to your point.

 

[A///]

I know who does. The later part of the comment was to your point.

Which was what? You're implying that silicon manufacturers help each other out.

 

[A///]

I know that 👍, but that was the google preview so people can get on the same page as me 🙂 (for I assume people do not stay up to date with the news, for I myself can not)

You'll have to wait until the report is finalised.

 

[Hitman928]

I'm still very skeptical this was actually made on a SMIC 7 nm process. It is trivial to mark the package with whatever label you want no matter where it came from. It will be interesting to see if they can actually put these out in any kind of real volume. If not, I think that would point to it not being made by a Chinese foundry. If someone with the means and motivation to actually xray and analyze chips for foundry provenance gets a hold of one, we'd know for sure.

 

[H433x0n]

I'm still very skeptical this was actually made on a SMIC 7 nm process. It is trivial to mark the package with whatever label you want no matter where it came from. It will be interesting to see if they can actually put these out in any kind of real volume. If not, I think that would point to it not being made by a Chinese foundry. If someone with the means and motivation to actually xray and analyze chips for foundry provenance gets a hold of one, we'd know for sure.

If you have an unlimited budget, yields no longer matter. They can have a 10% yield and it’d be irrelevant if each wafer is 90% subsidized. If you can afford to pay people 3-4x their current salary suddenly you can recruit decent talent from Taiwan and South Korea. SMIC is filled with TSMC alumni.

For Beijing the economics don’t matter, it’s about the propaganda opportunity and national agenda.

 

[Hitman928]

If you have an unlimited budget, yields no longer matter. They can have a 10% yield and it’d be irrelevant if each wafer is 90% subsidized. If you can afford to pay people 3-4x their current salary suddenly you can recruit decent talent from Taiwan and South Korea. SMIC is filled with TSMC alumni.

For Beijing the economics don’t matter, it’s about the propaganda opportunity and national agenda.

We'll see.

 

[Saylick]

If you have an unlimited budget, yields no longer matter. They can have a 10% yield and it’d be irrelevant if each wafer is 90% subsidized. If you can afford to pay people 3-4x their current salary suddenly you can recruit decent talent from Taiwan and South Korea. SMIC is filled with TSMC alumni.

For Beijing the economics don’t matter, it’s about the propaganda opportunity and national agenda.

This. The CCP's main obstacle to achieving parity with the West with regards to semiconductor technology has never been the money. It's always been about access to talent and equipment. If China has the ASML DUV machines and they've been poaching TSMC engineers like crazy, then there's a nonzero possibility that SMIC can produce a 7nm-class chip. At what volumes and at what yield? We'll never know. If it is being reported by the CCP, it will likely be inflated for national optics reasons, just like their GDP growth. The CCP is very image conscious, especially since it is what gives them legitimacy in the eyes of their citizens. "Look at what we accomplished, even with the Arrogant West looking down on us and trying to cut us off. Nothing can stop China!", will be what they say.

 

[itsmydamnation]

This. The CCP's main obstacle to achieving parity with the West with regards to semiconductor technology has never been the money. It's always been about access to talent and equipment. If China has the ASML DUV machines and they've been poaching TSMC engineers like crazy, then there's a nonzero possibility that SMIC can produce a 7nm-class chip. At what volumes and at what yield? We'll never know. If it is being reported by the CCP, it will likely be inflated for national optics reasons, just like their GDP growth. The CCP is very image conscious, especially since it is what gives them legitimacy in the eyes of their citizens. "Look at what we accomplished, even with the Arrogant West looking down on us and trying to cut us off. Nothing can stop China!", will be what they say.

4:2:1 problem , it is very much now about the money as well........

 

[Saylick]

4:2:1 problem , it is very much now about the money as well........

Well, the government can keep printing money, but yeah, the population decline will definitely put a cap to their economy, just that the effects are not heavily felt today. Maybe in like 20 years.

 

[itsmydamnation]

Well, the government can keep printing money, but yeah, the population decline will definitely put a cap to their economy, just that the effects are not heavily felt today. Maybe in like 20 years.

your out by exactly 20 years.
that's by CCP own data , its probably already worse.

Western countries can get away with it because people actually want to immigrate to them.......

but i've taken this off topic enough

 

[A///]

there was an article on their population decline a few days ago. the jist of it was life is expensive and younger people want to live their lives rather than become parents. Not sure how it here in the west.I can't really imagine it being much different. young people don't want to finish up university , get married and have children within 5 years of finishing university and beginning their careers.

 

[H433x0n]

An important note on this - Tech Insights hasn’t done electronmicroscope inspection, so I’ve got no idea what they’re basing all of this on. They’re just giving the CCP free publicity at this point to spike the football and do victory laps. I don’t get what they’re doing, they couldn’t just wait until they did a proper analysis?

The power/performance ratio is poor, with the largest heatspreader ever seen on a smart phone.

 

[Joe NYC]

Well, the government can keep printing money, but yeah, the population decline will definitely put a cap to their economy, just that the effects are not heavily felt today. Maybe in like 20 years.

People tend to cling to all sorts of reasons why China will fail, and the West will succeed, while the evidence to the contrary keeps piling on and on. Industry after industry, technology after technology, pace of infrastructure development in China that people in the West have not witnessed, unless they are like 80 years old.

The worst possible way to solve the population decline is by importing the most desperate people of the globe, with 8 years of education and IQ of 80.

It's a brainpower race, and China has the West outmatched. China has a lobsided advantage in brainpower, and it is only a matter of time their semiconductor industries become the execution engines like the rest of their industry, which is surpassing or has already surpassed the West. One by one.

For the time being, China will probably take DUV further than TSMC.

The way sanction will affect China getting their home grown EUV is that China will develop the full home grown supply chain for EUV. So, it will not be only ASML, but all of the suppliers who will one day wake up being technologically behind a Chinese competitor.

 

[H433x0n]

People tend to cling to all sorts of reasons why China will fail, and the West will succeed, while the evidence to the contrary keeps piling on and on. Industry after industry, technology after technology, pace of infrastructure development in China that people in the West have not witnessed, unless they are like 80 years old.

The worst possible way to solve the population decline is by importing the most desperate people of the globe, with 8 years of education and IQ of 80.

It's a brainpower race, and China has the West outmatched. China has a lobsided advantage in brainpower, and it is only a matter of time their semiconductor industries become the execution engines like the rest of their industry, which is surpassing or has already surpassed the West. One by one.

For the time being, China will probably take DUV further than TSMC.

The way sanction will affect China getting their home grown EUV is that China will develop the full home grown supply chain for EUV. So, it will not be only ASML, but all of the suppliers who will one day wake up being technologically behind a Chinese competitor.

You should invest in China! I’m sure you’d have no problems allocating a good chunk of your 401K into Baidu, BYD & SMIC. Good luck!

 

[Joe NYC]

You should invest in China! I’m sure you’d have no problems allocating a good chunk of your 401K into Baidu, BYD & SMIC. Good luck!

Chinese technology, and overall development is not about chasing paper profits on stock prices.

It is more about manufacturing products of ever increasing quality, at competitive costs (prices).

 

[Doug S]

I'm still very skeptical this was actually made on a SMIC 7 nm process. It is trivial to mark the package with whatever label you want no matter where it came from. It will be interesting to see if they can actually put these out in any kind of real volume. If not, I think that would point to it not being made by a Chinese foundry. If someone with the means and motivation to actually xray and analyze chips for foundry provenance gets a hold of one, we'd know for sure.

TSMC's original N7 was DUV only, so it is certainly doable for SMIC. As for what yields they get, who knows, but we have hints from the fact that the Mate 50 was available outside of China and supposedly the Mate 60 will be a China only product.

So yields are probably low which makes the chips more expensive, but in a phone selling for $1000 that's not a huge obstacle unless they were expecting Apple-like margins. If SMIC can't get the yields up the process will only be be useful for chips going into high dollar items, but that's fine for high end smartphones, electric cars, and of course China's military needs.

A process where cost is less important might even be able to extend DUV to 5nm or lower. LELELELELELELELE anyone?

 

[qmech]

I know nothing, thus me asking this question. Any good information about the recent 7nm China SMIC process? Bloomberg has a temporary exclusive and it is paywalled.

The company's Mate 60 Pro is powered by SMIC's 7nm chips, according to analysis that TechInsights conducted for Bloomberg News.

SMIC produced a "7nm class" chip last year, the MinerVA (a BitCoint mining chip). TechInsights had a deep analysis of it and concluded that had what could be considered "7nm features". This SMIC node was termed N+1 7nm.

The new chip, the Kirin 9000S, is much larger (the MinerVA was only 20mm²) and is supposedly on a new iteration of the SMIC 7nm process called N+2 7nm (although some Chinese media are calling it "5nm").

It is worth remembering at this point that TSMC's N7 process (and, indeed, Intel's many 10nm+++ variations, latest renamed Intel7) does not use EUV and instead relies on what is essentially the same DUV equipment that SMIC has access to (critically the Twinscan NXT:2000i, which doesn't come under embargo until the end of 2023, and is a newer and better scanner than the ones TSMC used for its N7 process).

TSMC's N7 process was quite good (arguably better than Samsung's 5LPE, although less dense), but although the differing design goals muddy the waters, the Intel7 process can certainly be argued to be the best pure DUV 7nm process. There is no reason that SMIC could not, eventually, achieve the same level of performance as TSMC and Intel, given that they have access to mostly the same equipment (and can analyze N7/Intel7 chips).

SMIC's N+1 7nm process wasn't quite at the level of the best DUV nodes according to TechInsights and we don't yet have the full report on the N+2 7nm process so comparisons are going to be somewhat lacking at this stage.

What we can do at this stage is compare the Kirin 9000S to various ARM chips that it is closely related to. The SoC features custom P cores that are very similar to ARM A7x cores. Like ARM with the X cores, the latest Kirin features a single "beefed-up" core (the "large"). As ARM X1 is to A78/A710, so the Kirin large core is to the medium P cores (although we don't actually know the specifics at this time).

In terms of single-threaded CPU performance, the Kirin 9000S is fairly similar to an ARM X1. It performs slightly worse and uses slightly more power, but it's close. Th e

In terms of multi-threaded CPU performance, the Kirin looks significantly better with the medium cores outperforming the A78 and closing in on the A710. The Kirin 9000S custom cores support HyperThreading, however, which boosts multithreaded performance significantly when enabled, but does hurt efficiency.

Overall, it's tough to say where the SMIC process lands, but it's probably reasonable to assume that it is very close to TSMC's N7 process. At this point, there is nothing to suggest anything beyond that and we obviously don't know anything about the yields (reports are that they are very low, but I'd categorize those "reports" as nothing more than rumors).

The die size of the Kirin 9000S is pretty big, but that might not indicate that the density is bad. Rather, the use of larger caches could explain that. Larger caches would also help improve power consumption, so if SMIC's process is struggling with power usage, spending some extra area on caches is one way to help.

In theory, SMIC's DUV equipment can handle "5nm class" feature sizes, so even if SMIC doesn't get access to better equipment over the next few years, there should still be some improvements. The biggest issue for them is that DUV is a dead-end and while they can continue to refine it, that will not really get them closer to a 3nm or better process where EUV is a must.

 

[Hitman928]

With Intel’s buyout of Tower dead, they’ve announced a new partnership in its place. If this goes well, I expect it may expand in the future.

Intel Foundry Services and Tower Semiconductor Announce New US Foundry Agreement

www.intc.com