Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel)

[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.

 

[DrMrLordX]

LOL at Intel courting Congress for money to increase domestic silicon foundry production while opening a fab in Germany and buying billions of dollars worth of wafers from TSMC to cover their own shortcomings.

 

[Ajay]

LOL at Intel courting Congress for money to increase domestic silicon foundry production while opening a fab in Germany and buying billions of dollars worth of wafers from TSMC to cover their own shortcomings.

Intel is begging at every trough. Non unusual from Multi-National companies in modern capitalist democracies. This is especially true when there is an enormous amount of industrial production dependent on the performance of one sector (as with oil companies - we tax them and then hand out huge subsidies to them).

 

[DrMrLordX]

Intel is begging at every trough. Non unusual from Multi-National companies in modern capitalist democracies. This is especially true when there is an enormous amount of industrial production dependent on the performance of one sector (as with oil companies - we tax them and then hand out huge subsidies to them).

How much money did Intel get from the EU to open a fab in Germany?

 

[Saylick]

Intel is begging at every trough. Non unusual from Multi-National companies in modern capitalist democracies. This is especially true when there is an enormous amount of industrial production dependent on the performance of one sector (as with oil companies - we tax them and then hand out huge subsidies to them).

Right... on one hand, they are begging for money and claiming they need the help, and on the other hand they say AMD is in the "rearview mirror" and will be forever.

Which one is it, Pat? Does Intel need help or is Intel dominating the competition?

 

[Ajay]

How much money did Intel get from the EU to open a fab in Germany?

IDK how much money was allocated to Intel from the European Chips Act. Not sure if that’s public yet.

 

[IntelUser2000]

You know in many other countries the government simply hands them massive amounts of money right? China does this, South Korea does this. Why do you think the latter has superfast internet(other than the tiny land mass)?

They don't care about monopolies there. Eventually every country ends up with monopolies, because lots of things favor them, such as economics of scale. And if you ask cheaper products, then only mega corporations can do so. Look at Costco!

It's true government subsidies help a lot, and for companies that are leaders in that field, they pretty much represent the nation. So of course they will be funded.

The US and Canadian postal services subsidize Chinese companies hence they are able to send them at dirt cheap shipping prices. Whether it's Chinese government incentivizing them or not it allows China as a country to achieve massive economic gains over time.

You'd pretty much lose out to companies that are based in countries like China if you don't start helping your own as shady as it might sound.

LOL at Intel courting Congress for money to increase domestic silicon foundry production while opening a fab in Germany and buying billions of dollars worth of wafers from TSMC to cover their own shortcomings.

And how's that a bad thing? If you are able to ramp up your production that way, then it's a win-win for them. Because it takes years to get there, so might as well be helped right?

If Intel gets a process lead, they'll rely less and less on TSMC. In fact it's starting to happen now, with Meteor and Arrow and even Granite originally said to be on TSMC but now is all Intel process.

 

[DrMrLordX]

And how's that a bad thing?

They're talklng out both sides of their mouth. After effectively destroying (or neutering) fab competition in the United States, now they want public money? No thanks. The money would be better spent courting Samsung or TSMC tbh.

 

[MadRat]

I'd rather see an American non-Intel foundry business open up to compete against TSMC and Samsung. The problem is there is no way to do it and be price competitive. When Intel expands it is about Intel products, not really to help other American designs.

 

[Thibsie]

I'd rather see an American non-Intel foundry business open up to compete against TSMC and Samsung. The problem is there is no way to do it and be price competitive. When Intel expands it is about Intel products, not really to help other American designs.

Proper advanced foundry in EU is also required IMO. From another industrial actor if possible.

 

[Doug S]

I'd rather see an American non-Intel foundry business open up to compete against TSMC and Samsung. The problem is there is no way to do it and be price competitive. When Intel expands it is about Intel products, not really to help other American designs.

It is about Intel products today, but not necessarily tomorrow.

Their previous attempt at becoming a foundry was obviously half hearted. They forced customers to use Intel tools and design flows instead of industry standards, and were only able to offer a small number of fairly recent processes versus everything under the sun like real foundries do. The bulk of non x86 stuff they made was almost all via acquisitions like Altera and Infineon (Apple modems) ported from other foundries with only a handful of true foundry clients, some of which Intel had to buy a stake in to make the deal happen!

This time around they're going industry standard, and making acquisitions to fill holes in their portfolio so they can become a true foundry. More out of necessity since even Intel's volume isn't enough to support them continuing to operate their own fabs in the long term, but still.

If (and that's a big if) they can deliver on their roadmap, and/or TSMC runs into roadblocks so that Intel catches up or regains the lead, they can compete for wafer starts for big players like Qualcomm (yeah I know they already announced something when Intel announced their IDM 2.0 strategy, but there were no commitments on volume so that was obviously for show rather than Qualcomm guaranteeing that Intel would get a significant percentage of overall wafer starts) Eventually they might even be able to land their long rumored white whale, Apple. Having leading edge capacity in the US would be attractive to Apple to increase their "made in the USA" percentage of BOM, though if TSMC counters with a leading edge fab in the US that could change (the TSMC announcement for AZ is for N5 generation and operational in 2024, so it will already be 1 1/2 process generations behind on opening day)

I think everyone saw how Intel claimed they were going to be a foundry last time but weren't very welcoming, so it may take some time for them to get any big wins without the usual Intel backdoor deals (or buying stake in the company again) but once they prove themselves being able to offer leading edge capacity made in the US will be a big advantage. That would justify the idea of making that Ohio complex much larger as was hinted at during last night SOTU address.

 

[jpiniero]

Their previous attempt at becoming a foundry was obviously half hearted.

The real problem was 10 nm, not any of their processes. It'll make no difference if it takes Intel 3 4-5 years to get anywhere near decent yield like it did 10 nm.

 

[Doug S]

The real problem was 10 nm, not any of their processes. It'll make no difference if it takes Intel 3 4-5 years to get anywhere near decent yield like it did 10 nm.

Even if they had delivered 10nm on time they would have had a lot of trouble getting their initial foundry attempt going. They were not getting anyone to sign up even before the first 10nm delays were announced, when everyone assumed they would maintain a process lead over TSMC. Customers don't want to have to essentially redo a design from the ground up to switch foundries or dual source, which would have been necessary with Intel's proprietary tooling before. There's a reason they can dual source their x86 chips now (or in the near future) but couldn't in the past.

 

[jpiniero]

Even if they had delivered 10nm on time they would have had a lot of trouble getting their initial foundry attempt going. They were not getting anyone to sign up even before the first 10nm delays were announced

They did, remember the customer that got screwed over that SemiAccurate reported on?

 

[MasterofZen]

Culture Clash at TSMC's Arizona Fab - EE Times Asia

Will management and workers find a middle ground as the chip giant's newest U.S. fab ramps production?

www.eetasia.com

So their secret to success is more PhDs and longer working hours?

I don't think that's true. According to official data from TSMC. Only 4.4% of their employees are PhDs. PhDs are definitely are doing R&D. No freaking way they are "using Ph.Ds to monitor a single piece of equipment on a production line."

But they are probably using master degrees and bachelors to monitor equipment on production lines.

 

[MasterofZen]

That makes it sound pretty terrible working for them. That's a BAD thing because they are saying we need to trade people's lives for corporate profit. Great, $100K+ spent on education but you end up being a lowly manager. Sounds great for humanity.

Also what's that about not needing PHDs but making it sound it was the PHDs that made the company successful.

I don't think that's true. According to official data from TSMC. Only 4.4% of their employees are PhDs. PhDs are definitely doing R&D. No freaking way they are "using Ph.Ds to monitor a single piece of equipment on a production line."

 

[MasterofZen]

It's not entirely clear what value the company derives from overqualified employees minding the daily operation of machines. Might be occasional boosts in productivity due to occasional prevention of edge-case failures. Hard to say.

This whole argument of PhDs monitoring a single equipment is totally false. According to official data from TSMC. Only 4.4% of their employees are PhDs. PhDs are definitely are doing R&D. No freaking way they are "using Ph.Ds to monitor a single piece of equipment on a production line."

But they are probably using master degrees and bachelors to monitor equipment on production lines. Could be overqualified, but I don't know.

 

[MadRat]

PhD vs MS is not always a good metric for research. PhD types are mainly in-the-box thinkers but are great at documentation. You need your rare creative types to push new ideas, and your PhDs to prove them valid. If you spend your career in documentation it does not bode well for their creative streaks. Ultimately what ideas get explored are up to decision makers that also have to count the pennies.

 

[IntelUser2000]

They did, remember the customer that got screwed over that SemiAccurate reported on?

Yes, but it doesn't change what DougS said and what I also agree on. Their attemps were either half-hearted or had little clue on how to be a real foundry. Their original goals were to assimilate everyone to become part of Intel.

I'd rather see an American non-Intel foundry business open up to compete against TSMC and Samsung. The problem is there is no way to do it and be price competitive. When Intel expands it is about Intel products, not really to help other American designs.

They will fab non-Intel american designs, that's the point of the new foundry model they are pursuing.

Also if we're on the line of countries funding to have this patriotic feeling and you fund it to companies like TSMC and Samsung, you aren't doing your job. Let's say Taiwan and Korea suddenly become enemies of the US. Do you think they will not be affected? Those "fabs in US" will no longer be such.

If it's to be American, then the only way is for an American company to to make it.

 

[DrMrLordX]

This whole argument of PhDs monitoring a single equipment is totally false.

Why does the degree matter? Even someone with a good undergraduate or masters degree is probably overqualified for monitoring a single piece of equipment all day long, especially if the MTBF is high.

 

[maddie]

The word used might be misleading.

Monitoring implies to most a passive role whereby the monitor mainly records.

Chip diffusion, etching, etc, is a complicated process involves the early adjustment of many parameters to keep it within the accepted range. The raw material inputs vary, as perfect uniformity is a fallacy of our thinking. In reality, industrial mechanical/chemical processing can more closely compared to cooking. The operator is constantly adjusting all of the factors to maintain the output within the desired range (+/-). I would imagine that ML is a big help today but managing several interrelated components is definitely not trivial.

Anyone who has worked in a chemical industrial processing plant will understand what I mean.

 

[MadRat]

I can imagine having multiple machines in play, running masks to etch each level. If one person is even 1% off at this scale it certainly creates misalignments. I would imagine masks are all applied on the same machine so that the whole bin is uniform. Otherwise feeding work from multiple sources would possibly create a situation requiring constant adjustments.

 

[maddie]

I can imagine having multiple machines in play, running masks to etch each level. If one person is even 1% off at this scale it certainly creates misalignments. I would imagine masks are all applied on the same machine so that the whole bin is uniform. Otherwise feeding work from multiple sources would possibly create a situation requiring constant adjustments.

Its not just masks, but the chemical assays of the dopants, etches, etc. Everything is X% purity and all batches vary slightly, not perfect, and as we get smaller nodes where a small amounts of atoms might cause a flaw, you need to adjust another input that might compensate. Earlier, larger nodes could get by with greater internal molecular flaws.

 

[igor_kavinski]

Unplanned power outage at Taiwanese plant interrupts semiconductor production across the region | TechSpot

TSMC is currently assessing the impact after observing voltage drops lasting between 400ms and 1000ms at some locations.

Despite TSMC's claims of little to no impact, the delays stemming from the 90-minute incident could cascade down and ultimately extend the already hobbled semiconductor industry's chip shortage. For decades, semiconductor manufacturers have relied on lean manufacturing, a manufacturing methodology that focuses on providing customers with what they need when they need it. That means low to no additional inventory or surplus and almost no room for error during manufacturing and delivery. Any delays in this type of environment can result in missed or delayed shipments, affecting both TSMC's upcoming deadlines as well as deadlines for customers that rely on those nodes for their own products.

400ms to 1000ms voltage drops can have consequences? If so, shouldn't they have thought of it before depending solely on the power grid without backup UPS?

 

[oak8292]

400ms to 1000ms voltage drops can have consequences? If so, shouldn't they have thought of it before depending solely on the power grid without backup UPS?

Without direct knowledge I can tell you with assurance that they have thought of backup UPS. These are competent engineers and business people. I will speculate that they have UPS on control systems but it is too expensive to back up the whole manufacturing process. Risk management is both engineering and business 101. Risk mitigation costs are weighed against benefit and even within the engineering and business ranks there are conflicting opinions because everybody assesses risk in different ways, either in terms of probability or consequence.

Lean manufacturing is essentially the result of regulatory control w/o disturbances. The longer a system is in stasis the benefits of 'lean' behavior improve the economics of manufacturing. A disturbance which requires flexibility has a cost, e.g. aging inventory, excess capital for production. Aging inventory in the semi-industry was very expensive in the past with Moore's Law devaluing transistors on an annual basis. Supply chain disturbances and the end of Moore's Law will adjust behavior and increase costs as 'insurance' is added.

P.S. Pat Gelsinger desire to have empty shells for opportunity production, e.g. Fab 42, 'smart capital' has a cost of unproductive Capex.

 

[Roland00Address]

What oak8292 said though I may disagree not with the logic but instead "depends on how you look at it." Certain types of goods in a national / world economy and its "circular flows" you want to create an abundance with.

This will drive a certain style of accounting to scream blood murder for you are overproducing and you are wasting one type of money / resource which should be reallocated into another sphere of life that you control. Well that is life and it depends on how you view things with different types of horizons / limits such as what time horizonS you are looking at, or what space horizonS you are looking at, so on and so on.

As long as you are making profit one should be one form of happy while simultaneously playing in your mind is there ways to view reality in different ways of seeing. There will always be challengeS with how one is doing a thing and that is actually a source of health not weakness. Likewise when one is screwing up there will also be challengeS to your paradigm.

c'est la vie