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

 

[Racan]

This doesn't bode well for AMD at all, what 3nm capacity will they be left with?

 

[NostaSeronx]

This doesn't bode well for AMD at all, what 3nm capacity will they be left with?

"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.

 

[Gideon]

Yeah that's not great at all for AMD if true (the fact that Intel will be producing 3nm chips in 2022 while AMD is only doing 5nm)

I still can't imagine how they were unable to snag any 3nm wafer and just let Intel in to steal all of the thunder. I'm pretty sure that when given a choice to give the same amount of Wafers at similar price to AMD vs Intel, TSMC would prefer AMD, as they aren't competing with them elsewhere.

EDIT:
But I still can't figure out what products exactly would Intel produce there in that timeframe, Meteor Lake?

Normally TSMC has rules that engineers working on a node can't at the same time work with a competitors node, does that not apply for Intel?

 

[lobz]

Yeah that's not great at all for AMD if true (the fact that Intel will be producing 3nm chips in 2022 while AMD is only doing 5nm)

I still can't imagine how they were unable to snag any 3nm wafer and just let Intel in to steal all of the thunder. I'm pretty sure that when given a choice to give the same amount of Wafers at similar price to AMD vs Intel, TSMC would prefer AMD, as they aren't competing with them elsewhere.

EDIT:
But I still can't figure out what products exactly would Intel produce there in that timeframe, Meteor Lake?

Normally TSMC has rules that engineers working on a node can't at the same time work with a competitors node, does that not apply for Intel?

It does, that's for sure, so we'll have to wait and see what they wanna do with that.

Still, if AMD was adamant in not pursuing bleeding edge processes and let Intel be co-lead in 3nm out of complacency instead, we can all just congratulate them. We don't know what and how they're planning, but judging from the information we do have, it doesn't look good.

 

[Zucker2k]

Ballsy. Should take some pressure off everyone at Intel, especially the process engineers. This move was always an option, pride aside, and I'm happy to see Intel go this route by eliminating the process advantage AMD has enjoyed in the last few years. Let's see what they do with the power and density advantages this will bring them; that's if this news is credible, and the source seems so.

 

[DrMrLordX]

Why would TSMC agree to that? Intel had to make some serious concessions if TSMC is going to put them on a leading-edge node. It was something TSMC publicly stated awhile back that they would NOT do.

 

[lobz]

^ ^ ^ ^ ^

Other than that, the wccf article states "Nikkei's sources also believe that the overall chip volume TSMC has dedicated to Intel is higher than that dedicated to Apple, resulting in the smartphone maker delaying adopting the 3nm process for its next year's smartphone upgrade." - which sounds really counter-intuitive for me.

 

[majord]

Ballsy. Should take some pressure off everyone at Intel, especially the process engineers. This move was always an option, pride aside, and I'm happy to see Intel go this route by eliminating the process advantage AMD has enjoyed in the last few years. Let's see what they do with the power and density advantages this will bring them; that's if this news is credible, and the source seems so.

Well, based on current competing architectures, the power and denisty advantages would only serve them to 'catch up' on both metrics.. to put it another way - to gain an advantage again they need more than just one node advantage

Lets not forget, 10nm++ isn't an uncompetitive node performance and density wise, but Zen 3 still has the upper hand on both size and efficiency

 

[DisEnchantment]

Hmmm lets see, TSMC enables it competitor to render its customers uncompetitive.
Difficult to digest.
Unless there is so much capacity to go around that it makes no difference. But if this is the case Intel Foundry business is dead on the water.

Besides that, architecture is the other big factor. We are not talking about off the shelf ARM designs here where only securing the best node will give you the edge.
Additionally, N7-> N5 Apple got only 1.5x scaling not 1.8x as advertised by TSMC. N5 --> N3 probably less than 1.35x vs advertized 1.7x. Ignoring efficiency, N5P will have comparable perf to N3

My bet is that some "Analyst" want to tank AMD shares to acquire lots of it.

 

[Gideon]

^ ^ ^ ^ ^

Other than that, the wccf article states "Nikkei's sources also believe that the overall chip volume TSMC has dedicated to Intel is higher than that dedicated to Apple, resulting in the smartphone maker delaying adopting the 3nm process for its next year's smartphone upgrade." - which sounds really counter-intuitive for me.

Yeah this seems ridiculous if all TSMC got was a slightly higher price per wafer.

Not only would it benefit Intel by allowing them to produce better chips than they can by themselves but they would also deny bleeding-edge wafers to main competitors. a win-win for Intel.

Actually a win-win-win when you consider that this would help Intel a ton with Intel's own open foundry efforts:

They'd get to see how TSMC works with customers first hand. And not just a customer, the biggest leading-edge customer on their best jewel node! They could copy all they like: All the design rules, methodology, how they communicate/troubleshoot with them, etc .... Not to mention extra insight into how TSMCs best nodes work.

And when this enables them to create the best products on the market, selling them at atrocious prices, they'd also end up in the green with this endeavour Which in turn would fuel their own future foundry efforts dumping TSMC ASAP.

When considering the opposite (not allowing Intel to be the main customer on their best node ASAP):

• TSMC would still sell every single 3nm wafer anyway (and with pretty good prices guaranteed)
• Every single Intel x86 chip loss would have an ultra high chance of ending up produced at TSMC be it:
  • Other x86 chips produced for AMD
  • M1 chips for Apple
  • Graviton chips for Amazon
  • Ampere Altra or other for ARM server vendors (e.g. oracle) - all use TSMC
  • Google's youtube silicon (replacing millions of Skylake Xeons)
  • Hell even Xilinx FPGAs
  • not to mention Future Nvidia/Qualcomm laptop chips (some might also be Samsung but still a loss to Intel)

ALL of the above would end up produced at TSMC anyway, as long as they keep their foundry advantage.

So there's got to be more TSMC gets out of it if it's true, or their strategist are full-on-braindead

 

[Gideon]

Let me condence the above post down a bit:

So if this is true Intel get's to:

• Build more competitive chips they can themselves
• Suffocate all of their competitors out of leading-edge chips (Apple, Hyper-scalers, not to mention AMD)
• Get unrivalled insight into how TSMC (and open foundries in general) work with their best customers, allowing them to:
  • Redesign from scratch how they design their chips, to be more compatible with the industry
  • Redesign all their (soon to be open) foundries to be more compatible and competitive with TSMC and others
  • get super-early mega-detailed access to how TSMCs nodes work and what parameters they have

TSMC gets .... slightly better prices for their N3 node that would 100% operate at capacity anyway?

 

[DrMrLordX]

So there's got to be more TSMC gets out of it if it's true, or their strategist are full-on-braindead

First let me state that I do not immediately accept the news to be true. I can only speculate as to what might drive TSMC into such a deal.

TSMC's ultimate goal is to eliminate competition and raise prices. Intel is/was a competitor to TSMC indirectly since Intel's Foundry 1.0 was never a serious threat to their business; nevertheless, it is in TSMC's best interest to force Intel out of the IDM and foundry game. If TSMC forced Intel to yield a LOT of cash as well as cede all future EUV machine sales to TSMC, then Intel is essentially out of the leading-edge fab business permanently. That would put near-100% of all x86 CPU production on TSMC nodes post 2023 or so, unless AMD decides to take a shot at Samsung nodes.

The obvious downside to helping Intel is that TSMC has within its capacity to save Intel from their own mis-steps. Intel can't buy themselves a functional leading-edge node, but TSMC sure can sell them access to one. Their entire enterprise division would be saved overnight, and by overnight I mean once Intel finally ports their designs over to N3. Revenue and market share restored! And that would give Intel enough future revenue to collaborate with IBM on a working 2nm process, which Intel could use to get back into the leading-edge fab game in a brazen attempt to stab TSMC in the back. Assuming TSMC hasn't forced Intel to agree not to do that anymore.

Which I'm sure Intel can totally be trusted not to do, pinkie swear.

 

[SKORPI0]

Reuters: TSMC may build its advanced 3nm foundry in the US instead of Europe

The EU is more interested in older nodes, which are used by its auto industry. Building more semiconductor foundries in the midst of a global chip...

www.gsmarena.com

 

[coercitiv]

Suffocate all of their competitors out of leading-edge chips (Apple, Hyper-scalers, not to mention AMD)

I see this mentioned a lot as some kind of super-strategic move Intel can pull to outmaneuver competition, and every time it is discussed without looking at the entire picture. I'd like to see someone explain how Intel is supposed to maintain their manufacturing plans (including R&D and CAPEX required to stay in the game) while also paying for the privilege of reserving the biggest chunk of TSMC capacity on a leading node.

What happens if TSMC does manage to (roughly) meet demand from main Intel competitors? What's Intel going to do with a huge allotment from TSMC and their own leading edge fabs asking for orders while competition eats away their market share?

 

[jpiniero]

I see this mentioned a lot as some kind of super-strategic move Intel can pull to outmaneuver competition, and every time it is discussed without looking at the entire picture. I'd like to see someone explain how Intel is supposed to maintain their manufacturing plans (including R&D and CAPEX required to stay in the game) while also paying for the privilege of reserving the biggest chunk of TSMC capacity on a leading node.

Begging for Biden Bux.

TSMC did say they wouldn't go out of their way to build capacity specifically for Intel.

 

[Zucker2k]

What happens if TSMC does manage to (roughly) meet demand from main Intel competitors? What's Intel going to do with a huge allotment from TSMC and their own leading edge fabs asking for orders while competition eats away their market share?

Besides, TSMC are in this for the money. Also, besides Intel, all their other customers have nowhere to go. Heck, even Intel may be stuck with them for a long while now. Might as well take the dough while it's being offered, for expansion and other purposes.

I don't understand how AMD could allow themselves to be shuffled back this way. They ought to have secured their position in line in what is essentially the main source of their recent successes over Intel.

 

[naukkis]

First let me state that I do not immediately accept the news to be true.

It should not be a shock news as Gelsinger revealed that deal months ago. What it basically means that Intel is given up on leading edge nodes even they still don't publicly admit such happening.

 

[lobz]

It should not be a shock news as Gelsinger revealed that deal months ago. What it basically means that Intel is given up on leading edge nodes even they still don't publicly admit such happening.

Gelsinger has revealed no such deal, where they'd get more allocation on a bleeding-edge 3nm node than Apple, resulting in Apple of all companies changing their plans.

 

[naukkis]

Gelsinger has revealed no such deal, where they'd get more allocation on a bleeding-edge 3nm node than Apple, resulting in Apple of all companies changing their plans.

Gelsinger revealed that there is leading edge products coming from TSMC for both client and server product lines. Ok, Gelsinger didn't reveal that it was from 3nm but that information was leaked even before from Taiwan.

Gelsinger openly revealed that shock news, and most tech sited didn't even notice it. Anandtech did notice, but didn't make halo headlines from it.

 

[eek2121]

Why would TSMC agree to that? Intel had to make some serious concessions if TSMC is going to put them on a leading-edge node. It was something TSMC publicly stated awhile back that they would NOT do.

MONEY. Bucket loads of it.

TSMC would be foolish to say no. Intel will exist regardless of whether TSMC chooses to make Intel chips or not. Contrary to popular belief, Intel is still a large, very profitable company and they have zero chance of disappearing any time soon.

Why miss out on the chance to make a buck from your competitor?

It should not be a shock news as Gelsinger revealed that deal months ago. What it basically means that Intel is given up on leading edge nodes even they still don't publicly admit such happening.

They have done no such thing. Intel 7nm and TSMC 3nm are within spitting distance of each other (spec wise).

It is likely that TSMC will be used for either GPUs and possibly some portion of the CPU business. It has already been discussed that Intel has a limited number of EUV machines. They will likely rely on TSMC until they can ramp up.

 

[naukkis]

It is likely that TSMC will be used for either GPUs and possibly some portion of the CPU business. It has already been discussed that Intel has a limited number of EUV machines. They will likely rely on TSMC until they can ramp up.

No need to guess, Gelsinger also revealed that most of Intel CPU still come from their own foundries 2023. So not insignificant amount of cpus will come from TSMC - and they are leading edge products from best available process. I can't believe arguing about that as Gelsinger did reveal those details months ago - are they so shocking that most of people just refuse to accept it? And nobody didn't guess that best available process for Intel($$$) would be the best TSMC could offer?

 

[lobz]

No need to guess, Gelsinger also revealed that most of Intel CPU still come from their own foundries 2023. So not insignificant amount of cpus will come from TSMC - and they are leading edge products.

That sounds as really conflicting pieces of information, considering that todays news is talking about 2 CPUs for 3nm, now add into the mix that these 2 CPUs would get a bigger allocation than Apple does, and that most of Intel CPUs will still come from Intel fabs. Something is not adding up. I'm calling at least half of today's news fake.
It wouldn't be the first time that otherwise credible media outlets publish totally baseless news about TSMC.

 

[xilli_fiberbit]

AMD hasn't already secured wafers on TSMC N3 process ?

It can help with the TSMC fab in Arizona but which start poduction in 2024. Is for which process this fab ?

 

[naukkis]

That sounds as really conflicting pieces of information, considering that todays news is talking about 2 CPUs for 3nm, now add into the mix that these 2 CPUs would get a bigger allocation than Apple does, and that most of Intel CPUs will still come from Intel fabs. Something is not adding up. I'm calling at least half of today's news fake.
It wouldn't be the first time that otherwise credible media outlets publish totally baseless news about TSMC.

Gelsinger revealed that there is coming at least two different designs from TSMC - for client and server platforms. And as leading edge process building is slow and expensive that more than Apple gets from TSMC isn't still nowhere as much as Intel can product cpu's at it's own foundries. So there's nothing conflicting with that information.