Discussion Intel leading customer for TSMC 3nm?

[Failnaught]

There is news that Intel is the co-leading customer on the TSMC 3nm node, with Apple. Did they manage to outflank AMD on process technology? What are they going to make on this node?

https://asia.nikkei.com/Business/Te...become-first-to-adopt-TSMC-s-latest-chip-tech

 

[Hougy]

"TSMC's Fab 18 phase 5-8 South plant will be responsible for the N3 process; Fab 18 phase 4 is used for N5 process expansion. In addition, TSMC plans to build a new Fab 20 plant in Hsinchu for N2 production." - Jun 02, 2021

Fab 18 Phase 1-4 = ~50K*4; Max Capacity 200K of N5
Fab 18 Phase 5-8 = ~50k*4; Max Capacity 200K of N3
Fab 20 Phase 1-4 = ~50k*4; Max Capacity 200K wpm of N2

AMD only gobbles 20K initially and if the node is well liked they increase meal capacity to 30K. 200K-30K = 170K left. There is plenty of room going forwards for AMD.

How many wpm does Apple use?

 

[dmens]

I remember it being stated that the recent increases in Intel server volumes were related to the security problems of late. It was simpler for customers to add hardware rather than change systems with all the associated validation issues. In other words, the security issues with mainly Intel hardware counterintuitively led to more sales by Intel. That period seems to be ending, as the following quote in the article by dmens states.

On a call with analysts, Intel Chief Executive Pat Gelsinger, in his first earnings report since officially taking over, noted that Intel had just recently launched its new generation of Xeon server chips, codenamed “Ice Lake,” and that the industry is just beginning to emerge from a digestion phase from data centers and that Intel is “starting to see signs that they want to start the next build phase in their cloud.”

The new era begins.

The cloud digestion excuse was thoroughly debunked by continual record quarters by AMD. Also you still cannot buy an Icelake server from Dell or Supermicro.

 

[jpiniero]

Also you still cannot buy an Icelake server from Dell or Supermicro.

You can from Dell but it won't ship until August.

 

[Doug S]

How many wpm does Apple use?

No one knows for sure, all we can do is make some assumptions and guess...

For iPhone alone, assuming a 90 mm^2 die and 90% yield they get 600 working chips per wafer. If you assume they sell 20 million iPhones per month (which is probably about in line with now though historically it is hasn't been quite that high) with half being the latest model and half being older models (I have no clue on the product mix) they'd need 16,666 wafers per month just to supply iPhone 12. Since they sell the most iPhones right after launch and fewer as next year's launch approaches the peak may be 25-30k wpm, and down to 10K or less around this time of year.

That's just on the most leading edge process (i.e. N5) If you want their overall total you have to add in not only wafers for older iPhones but iPad / iPad Pro, Mac, Watch, ATV, etc. Their total average across all products is probably probably at least 40k, perhaps as high as 50k, but they use chips as old as A8 in some products (Apple TV HD) which was made in 20nm. Who knows if they even still make those, they might have a stockpile and will upgrade Apple TV HD to a faster SoC when the A8 stock runs out.

Apple's iPhone dies are going to get bigger in a year or two when they begin integrated their modem. I expect they will support only LTE and 5G to make the baseband software simpler and reduce patent licensing cost, but that won't save much chip area.

 

[NostaSeronx]

How many wpm does Apple use?

From what I can find, they had ~80% of risk production which was ~24,000 of 30,000 5nm wafers per month in 1H2020. With initial mass production in 2H2020 they had ~53% which was ~32,000 of 60,000 5nm wafers per month.

 

[maddie]

The cloud digestion excuse was thoroughly debunked by continual record quarters by AMD. Also you still cannot buy an Icelake server from Dell or Supermicro.

What?

 

[itsmydamnation]

wasn't there a graph from ARM showing arm x86 vendor 1 and 2 and deployment rate for Amazon and ARM and x86 vendor 2 were eating x86 vendor one's lunch quite considerably.

 

[beginner99]

Not just profit, but big fat margins since Intel is always run by beancounters first and foremost.

You know what gives fat margins? being able to sell everything that rolls out of your 7 year old 14nm fabs. the pandemic and chip shortage was intels biggest break. can't be picky with performance or performance/watt when the supply is this limited.

 

[Doug S]

You know what gives fat margins? being able to sell everything that rolls out of your 7 year old 14nm fabs. the pandemic and chip shortage was intels biggest break. can't be picky with performance or performance/watt when the supply is this limited.

Not only that but most PC buyers are not performance sensitive. PCs have been more than fast enough for years for the average consumer or corporate buyer. Intel's production cost for 14nm CPUs is tiny since they've fully depreciated those fabs by now, so they are making bank selling those in PCs whose buyers don't care that they are little improved over the CPUs installed in PCs five years ago.

The people who say "oh AMD has access to better processes and offers faster CPUs so they are going to take over the market" don't understand how the PC market works. Because they care so much about performance, and hang around on forums where everyone else cares about performance, they keep waiting for Intel's profit to tank. They are making a lot more per CPU sold than AMD does.

The performance difference really only matters for servers - but since Intel is able to sell all the Xeons it makes currently they don't have to care and won't until that's no longer true.

 

[dmens]

You know what gives fat margins? being able to sell everything that rolls out of your 7 year old 14nm fabs. the pandemic and chip shortage was intels biggest break. can't be picky with performance or performance/watt when the supply is this limited.

That will provide a financial buffer until the ASP is too low to matter. Intel got lucky with COVID, they would be in far deeper trouble otherwise.

 

[coercitiv]

Get ready for a delicious short read that matches this thread like a glove. I'm amazed we missed it so far, or at least I did.

More than manufacturing: Investments in chip production must support U.S. priorities
By PAT GELSINGER, CEO, INTEL | Sponsored content

The federal government should invest in American intellectual property and capabilities. It should invest U.S. tax dollars in companies that are based here and house their most critical assets ― including patents and people ― domestically;

By contrast, foreign chipmakers vying for U.S. subsidies will keep their valuable intellectual property on their own shores, ensuring that the most lucrative and cutting-edge manufacturing stays there and requiring the U.S. to make the difficult choice between forgoing the advanced chips necessary for critical national security applications or relying on insecure, foreign supply chains for them. According to an analysis in The Economist, even after Taiwanese chipmaker TSMC opens its first plant (called a “fab”) in the U.S. in 2024, the company will still be making its most advanced products in Taiwan.

To maximize the return on the U.S. government’s investments in the semiconductor industry, policymakers need to stimulate the development of advanced semiconductor process technology and manufacturing under a set of common objectives.

The federal government is poised to invest $52 billion by 2026 to support the semiconductor industry. We have two choices in investing this money: simply providing incentives that subsidize production of chips in the U.S. or fostering an ecosystem that makes the U.S. the world’s best place to manufacture technology for years to come.

TL;DR: Dear government, we lost manufacturing leadership while maximizing profits for our shareholders. If you give us the lion's share of those $52 billion we promise not to spend them on... uhm.. profits. Pinky swear. TSMC bad, they won't give us 3nm on a silver platter. Let's choose wisely, wink wink.

Can't wait for the sponsored article from TSMC.

 

[Vattila]

Pat Gelsinger: "The federal government should invest in American intellectual property and capabilities. It should invest U.S. tax dollars in companies that are based here and house their most critical assets ― including patents and people ― domestically;"

If his pleas are heard, support will (or should in all fairness) come with conditions of full independence between Intel product design and Intel Foundry Services. This seems to be the direction of travel. With the clear path that he has set out for IFS to become a successful foundry, independence is a necessity. Also, I doubt he wants to have that low-margin business weigh down their product business, so a full spin-off of IFS and fabs will eventually happen, I think. They will accelerate towards that outcome, if their product division buys large leading-edge volumes from TSMC.

IFS Fact Sheet: (intel.com)
Speculation: Intel will become fabless | AnandTech Forums: Technology, Hardware, Software, and Deals

 

[DrMrLordX]

Get ready for a delicious short read that matches this thread like a glove. I'm amazed we missed it so far, or at least I did.

A lot of it is blather so honestly I would have ignored it had I seen it. Regardless, Intel would be double-dealing out the yin-yang if they took $52 billion from the Feds and spent it on 3nm wafer access from TSMC. As it stands, throwing that kind of money at Intel (or GF or any US-based silicon foundry) would be wasted since they have all demonstrated an inability to stay on leading edge processes. Maybe IBM will save them, or maybe not.

 

[scannall]

I am somewhat skeptical about this rumor. We'll see.

But giving Intel 52 billion dollars to bail them out of at least 15 years of bad decisions is the wrong thing to do. Heck, they've spent more than that pumping their stock price, and now give them a handout? No thanks. I'd rather see them build up Global Foundries.

 

[Gideon]

Get ready for a delicious short read that matches this thread like a glove. I'm amazed we missed it so far, or at least I did.

The way Intel managed to lobby all of the government support behind themselves, killing off all viable competitors like TI, no wonder they are now the "only hope". The reality is it would be very wise to get TSMC statesides aswell ASAP and with more than one fab.

 

[DisEnchantment]

The way Intel managed to lobby all of the government support behind themselves, killing off all viable competitors like TI, no wonder they are now the "only hope". The reality is it would be very wise to get TSMC statesides aswell ASAP and with more than one fab.

Those monopolistic and anti-competitive traits still run deep within their DNA. They have to move on from this, no place for this in this day and age.
If there is one thing the US political leadership should not do is to stifle competition by only helping Intel at the cost of others even for the right reasons.
No competing domestic fab has led to US losing the technological leadership in this important aspect of Semiconductor because the sole foundry is only interested in earning their shareholders and executives more money and suddenly waking up now at the back of the race.

GF could be a candidate too to receive the Govt funding, but not before they lose their incompetent masters.

On top of this, I hope DARPA/DoE gets lots and lots of slush money to keep the innovation going forward.
They surely do not operate for profit, but I bet they would be more than happy to see their seeds bearing fruit with some partners who have been able to deliver on their investment.
DARPA and DoE have to start thinking far beyond the Exascale. Quantum, Photonics, new semiconductor materials, etc.

 

[A///]

A lot of it is blather so honestly I would have ignored it had I seen it. Regardless, Intel would be double-dealing out the yin-yang if they took $52 billion from the Feds and spent it on 3nm wafer access from TSMC. As it stands, throwing that kind of money at Intel (or GF or any US-based silicon foundry) would be wasted since they have all demonstrated an inability to stay on leading edge processes. Maybe IBM will save them, or maybe not.

While you say that I don't have any faith in Intel delivering 10nm SuperFIN or whatever BS marketing they came up with at Intel day last year, let alone their 7nm process which has gone from bad to good to bad to good in a matter of a year.

 

[Doug S]

I'm still baffled as to how they were unable to get any 3nm wafers and instead allowed Intel to take all of the limelight. I'm confident that if given the choice between giving the same quantity of Wafers to AMD or Intel at a comparable pricing, TSMC would choose AMD because they aren't competing with them anywhere.

It wasn't a matter of "choosing". If you come to TSMC with cash in hand they'll sell you future wafers. Intel bought up future capacity before AMD did, that's why they have it.

If you want to blame someone, blame AMD's management for deciding to spend precious cash on stock buybacks instead of future wafer allocation.

 

[jpiniero]

If you want to blame someone, blame AMD's management for deciding to spend precious cash on stock buybacks instead of future wafer allocation.

Timing may be an issue for AMD. You have to have the designs ready and it has to make sense from a sales and marketing perspective. Intel is so big that they can do whatever, but AMD can't.

 

[beginner99]

#4 If accurate, that's not good news for AMD (the fact that Intel will be producing 3nm chips in 2022 while AMD is only doing 5nm)

I suspect TSMC will mostly do GPU/compute parts for intel and not x86.

 

[jpiniero]

I suspect TSMC will mostly do GPU/compute parts for intel and not x86.

It has to be Core products. Intel's not going to pay this kind of cash for GPUs.

 

[A///]

We know Ponte Vechio is on TSMC, but not anything else. I doubt Intel would waste their own wafers on their DG line. Especially if it's become a money sink. I can see small/cheap lines of x86 being done for them, but nothing in the performance sector. That would be interesting to witness if they did. The company who's been adamant about running fabs for decades? The fall out will be spectacular beyond anyone's wildest imaginations. Intel has yet to make a solid recovery from their July report last year when their value nosedived.

 

[LikeLinus]

I'm curious, a lot of people have been posting all from this mindwhatever (some German retail site) and sites talking about how the 5000x is so hard to get because they are selling so well and no one could keep them in stock. How is it that Intel gained marketshare in 1Q 2021 over AMD? You know, cause people for the past two years have been saying Intel was dead and they can't survive. Makes me curious?

I don't think Intel is using TSMC for GPUs. If anything, they are using it for 3nm and using it to their advantage. Even if they slipped on 10nm? You'd be dumb not to use every resource possible to gain back your advantage. You can't hate Intel for using TSMC when AMD uses the same company. That's just stupid.

But I will say I was surprised based on all the comments/threads here on how AMD was selling out everything. Intel still gained marketshare in the Laptop/desktop segment. Congrats to Intel and AMD!

 

[Ajay]

Timing may be an issue for AMD. You have to have the designs ready and it has to make sense from a sales and marketing perspective. Intel is so big that they can do whatever, but AMD can't.

I think @Doug S is right. AMD is tossing $4B to investors to pump up their stock prices (and make AMD executive rich). It's BS. What if they waved that cash in front of TSMC? That's allot to pass up. AMD is going from being the smaller scrappier company to another typical Wall Street playa. It's a sad turn that could jeopardize their medium term success.

 

[LikeLinus]

I think @Doug S is right. AMD is tossing $4B to investors to pump up their stock prices (and make AMD executive rich). It's BS. What if they waved that cash in front of TSMC? That's allot to pass up. AMD is going from being the smaller scrappier company to another typical Wall Street playa. It's a sad turn that could jeopardize their medium term success.

Truth be told, all companies want to be a wall street "playa". For some people, they are too ignorant to realize that every company that sells products is out for one thing, money. For the most part, any larger company, there really isn't one better than the other. It's just picking a side and gambling on who big fish take little fish. This is exactly why I told a few people in this forum that, despite them thinking Intel was going under, it wasn't going to happen because deep pockets can fish deeper. Not that it cant happen, but the likelihood of someone like Intel falling, well, it's unrealistic. It's been two years though and I doubt they will resurface to tell us all they were right?

AMD is only doing what every major company does and is in their best interest. Pay the shareholder. Facts of life.