Question Intel Q2: 7 nm in bad shape

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Doug S

Senior member
Feb 8, 2020
250
301
96
If Intel's 7nm is on way to become another 10nm disaster, probably the best move Intel can make is getting into cross-licensing agreement with Samsung. Samsung's ambition is to become biggest foundary business by 2030 and they would absolutely love to get a piece of Intel's business.
I agree that IF Intel went fabless, partnering with Samsung would make the most sense. They have plenty of spare capacity, plenty of ambition, and the US government (DoD/TLAs) would be more comfortable with chips made outside the US if they're made in South Korea. Better yet from their perspective, Intel could turn over its newer US fabs to Samsung and insist a certain percentage have to be made locally so a US source for leading edge capacity remains for the TLAs.
 

maddie

Diamond Member
Jul 18, 2010
3,159
1,829
136
I agree that IF Intel went fabless, partnering with Samsung would make the most sense. They have plenty of spare capacity, plenty of ambition, and the US government (DoD/TLAs) would be more comfortable with chips made outside the US if they're made in South Korea. Better yet from their perspective, Intel could turn over its newer US fabs to Samsung and insist a certain percentage have to be made locally so a US source for leading edge capacity remains for the TLAs.
Not an American, but isn't anyone else at least a bit concerned that all of the leading edge semi production is slowly moving to what is rapidly becoming the hottest geopolitical region of the planet?
 

Doug S

Senior member
Feb 8, 2020
250
301
96
Not an American, but isn't anyone else at least a bit concerned that all of the leading edge semi production is slowly moving to what is rapidly becoming the hottest geopolitical region of the planet?
Yes this is a huge risk separate from "national security" concerns, and everyone knows it, but isn't sure what can be done about it. Heck, it is also in the 'ring of fire' which is a separate "all your eggs in one basket" risk.

Before anyone brings up TSMC's upcoming Arizona fab, it is small by comparison to the ones they have in Taiwan and will be 3-4 years out of date when it begins production, so won't count as "leading edge" production anyway.

Look at what an earthquake does to RAM prices. Imagine an earthquake that impacts production at TSMC's biggest fabs, and throws off product cycles for Apple, AMD, all the Qualcomm licensees and on and on. At least that only knocks at most ~3 months of work. If war ever broke out between China and Taiwan it would be much worse, though that seems unlikely (China wants Taiwan intact, dropping bombs on it would be counterproductive to their goal of bringing home the "rebel province" with its economic strength intact)
 

mxnerd

Diamond Member
Jul 6, 2007
5,440
631
126
Not an American, but isn't anyone else at least a bit concerned that all of the leading edge semi production is slowly moving to what is rapidly becoming the hottest geopolitical region of the planet?
All big name chip companies AMD, NVIDIA, QUALCOMM, APPLE been doing this for many years, the trend is irreversible at this point.
 

alcoholbob

Diamond Member
May 24, 2005
6,138
257
126
Fair enough.

For myself though, I never saw Keller as an Intel guy and looked at that video as an individual speaking his mind. Sure, he's a hired gun and was used by Intel to message the wider world, but he's right. Moore's Law is really an economic one and 3D tech will allow it to continue apace.
Keller himself is mostly operating on hope. He admits even with cutting edge science and materials that don't even exist yet, he can only envision 5 or 6 full node shrinks left for transistors. He's kind of misrepresenting his position, because he's trying to confuse two different concepts, the fact that some number of node shrinks are still possible in theory versus the speed and cadence of Moore's Law. To me it's him basically absorbing some of Elon's style of marketing ploys and making some fuzzy factual claims when in reality what he's actually doing is just some cheerleading, justifying their statements with the rationale of "good things will happen in the future, be positive."
 
Last edited:
  • Like
Reactions: Tlh97 and scannall

DrMrLordX

Lifer
Apr 27, 2000
16,020
4,970
136
Im surprised no one responded about this one yesterday
I don't follow Adored, though if he can add something to this mess then so be it.

  • intel is in a game of thrones power struggle for next ceo.
  • bob swan is possibly out this year.
  • murthy is sabotaging raja's group and possibly had something to do with keller's leaving.
  • 10nm is going to be bad but 7nm is worse.
  • murthy's claims about 7nm progress could come back to bite him.
Raja isn't/shouldn't even be in contention for the CEO seat. I agree that Swan will be shown the door, but taking out Raja makes no sense. Seems like Murthy is too protective of his own people in the fabs that stand to lose their jobs if/when Intel either a). spins off their fabs or b). starts licensing process tech from Samsung or TSMC (which is the logical course of action). Raja's designs may or may not be bunk, but he's smart enough to know he can't get DG2 or Ponte Vecchio working on Intel process tech. The only way Murthy makes it through this mess with a company worth leading is if he makes an alliance with Koduri, not if he bumps him off. That's a potential revenue stream that they can easily farm out to TSMC and be done with it. No harm done to Intel's internal manufacturing since they won't be taking a product away from Murthy's people.

If Murthy really is hot to wipe out Raja, he's making a huge mistake.

As for 7nm being worse than 10nm? We may never know, since it's very possible that Intel will be forced to pull a GF and either quit the fab race entirely, or just start integrating someone else's tech. Eventually Intel may release some kind of product on a 7nm EUV process, but that process won't be their work.

Taking all of Intel's fabs offline for leading-edge manufacturing is going to spread chaos in the market. Intel still produces an enormous amount of silicon that go into current-gen products on the consumer and professional level. If Samsung and TSMC aren't reaching out to Intel right now to negotiate a way to keep those fabs running, it's either going to be a huge mess or an enormous opportunity for someone. As it stands, though, I don't think anyone has the fab capacity to replace Intel 14nm. And if you think about it, if Intel continues on their current course, they won't be able to replace their own 14nm process either. The only way the current market demand can be met is if Intel integrates someone else's tech into their own fabs and starts pumping out Golden Cove (or similar) on hybridized processes.

Why did Intel lie and mislead regarding 10nm but now they are more forthcoming about 7nm problems?
What makes you think they're entirely forthcoming about 7nm's problems? An actual 6 months delay is really nothing. A robust 7nm CPU launch in Q1 2023 is not really a problem for Intel. The underlying message is that there's a bigger delay in store for a competitive product at volumes that could meet the demands of the market.

I agree that IF Intel went fabless, partnering with Samsung would make the most sense. They have plenty of spare capacity, plenty of ambition, and the US government (DoD/TLAs) would be more comfortable with chips made outside the US if they're made in South Korea. Better yet from their perspective, Intel could turn over its newer US fabs to Samsung and insist a certain percentage have to be made locally so a US source for leading edge capacity remains for the TLAs.
Can Samsung meet all of Intel's product commitments? Also, what happens to Intel's fabs and all the engineers that come with them?
 
  • Like
Reactions: Tlh97

maddie

Diamond Member
Jul 18, 2010
3,159
1,829
136
All big name chip companies AMD, NVIDIA, QUALCOMM, APPLE been doing this for many years, the trend is irreversible at this point.
Reminds me of the Jamie Dimon quote from 2007.

“When the music stops, in terms of liquidity, things will be complicated. But as long as the music is playing, you’ve got to get up and dance. We’re still dancing.”
 
  • Like
Reactions: Tlh97

JasonLD

Senior member
Aug 22, 2017
253
203
86
Can Samsung meet all of Intel's product commitments? Also, what happens to Intel's fabs and all the engineers that come with them?
That would take at least 5 years I assume.
Well, they can probably start by getting Samsung to fab their GPUs, then some mobile client GPUs. After that, start doing joint-venture with Samsung on future node would be the best case scenario, that way they can keep operating on their own fabs.
 

maddie

Diamond Member
Jul 18, 2010
3,159
1,829
136
Keller himself is mostly operating on hope. He admits even with cutting edge science and materials that don't even exist yet, he can only envision 5 or 6 full node shrinks left for transistors. He's kind of misrepresenting his position, because he's trying to confuse two different concepts, the fact that some number of node shrinks are still possible in theory versus the speed and cadence of Moore's Law. To me it's him basically absorbing some of Elon's style of marketing ploys and making some fuzzy factual claims when in reality what he's actually doing is just some some cheer leading, justifying their statements with the rationale of "good things will happen in the future, be positive."
Well, to be honest, Moore's Law is really about the cost of circuits falling over time, which traditionally happened with, but is not tied, to node shrinks. Utilizing the 3rd dimension, as the Nand people are doing is another way of maintaining progress. We're seeing a large amount of R&D happening in this area for this very reason.


edit:
Wanted to add. Why are we almost always a couple decades from depleting a resource? Because no one is going to invest in finding something with that long of a repayment period.
 
Last edited:

uzzi38

Senior member
Oct 16, 2019
894
1,104
96
I think calling 7nm another 10nm or worse is definitely over-exaggerating. It's no good at all, don't get me wrong, and there is a solid chance Intel are still misleading people on the delays... but it's not going to be 10nm bad where we're looking at ~5 years after initial launch time period for passable yields.

As for power struggle stuff... that stuff is true. I don't know if Keller was a victim, or rather I'm nearly certain that his main reason for leaving when he did was family based.

On the flip side, I'm not at all surprised to hear both Raja and Jim are being forced out... I just thought at this point Raja was probably safe. After all, iirc after Jim left Raja was publically announced to be placed in charge of several product roadmaps, and firing two major figures like Jim and Raja in the time period doesn't quite instill much confidence in your company.
 

maddie

Diamond Member
Jul 18, 2010
3,159
1,829
136
I think calling 7nm another 10nm or worse is definitely over-exaggerating. It's no good at all, don't get me wrong, and there is a solid chance Intel are still misleading people on the delays... but it's not going to be 10nm bad where we're looking at ~5 years after initial launch time period for passable yields.

As for power struggle stuff... that stuff is true. I don't know if Keller was a victim, or rather I'm nearly certain that his main reason for leaving when he did was family based.

On the flip side, I'm not at all surprised to hear both Raja and Jim are being forced out... I just thought at this point Raja was probably safe. After all, iirc after Jim left Raja was publically announced to be placed in charge of several product roadmaps, and firing two major figures like Jim and Raja in the time period doesn't quite instill much confidence in your company.
In his time at AMD, it became quite clear that Raja is very much a political creature.
 

Genx87

Lifer
Apr 8, 2002
41,079
487
126
This is remarkable news. If Intel fabs all their leading edge chips on TSMC (or Samsung) how do they get the volumes they need? TSMC is fully booked out. Fab capacity takes years to build (so if they want to pass TSMC a couple B$ to get in front of the line, they'd better get on that last month). So... how will it work? Do they have the option of selling their fab to TSMC or licensing the TSMC recipe like how GloFo did with IBM?
Perhaps a partnership where Intel invests in a TSMC fab that also grant Intel first dibs on production?
 

Zucker2k

Golden Member
Feb 15, 2006
1,158
514
136
I don't know if Keller was a victim, or rather I'm nearly certain that his main reason for leaving when he did was family based.
Jim Keller may be a family man but he's not a nurse. Imagine this 7nm bombshell news coming out at the time Keller announced his resignation - 7nm is so delayed Keller has no option but to resign, or stay and twiddle his thumbs until 7nm is ready. That family excuse was coincidental and timely.

 

uzzi38

Senior member
Oct 16, 2019
894
1,104
96
Jim Keller may be a family man but he's not a nurse. Imagine this 7nm bombshell news coming out at the time Keller announced his resignation - 7nm is so delayed Keller has no option but to resign, or stay and twiddle his thumbs until 7nm is ready. That family excuse was coincidental and timely.

Jim Keller's role at Intel was more than just to develop a uArch. He was also there to try and change mindsets and working practices at the company.

Not to mention what you wrote doesn't really make sense, if he left because he wouldn't have much to do until 7nm got sorted out... well he wouldn't have much to do even if it was sorted out. A broken node doesn't mean you completely stop future development.

Failure of 7nm is the last reason I'd think he left. First is certainly family, and second would be due to being a victim of the corporate rot over at Intel. I'm still leaning more towards the first because the second was a rumour which - so far - only appears to be half true. Time will tell if it's completely true or not.
 

NTMBK

Diamond Member
Nov 14, 2011
8,707
1,680
126
Jim Keller's role at Intel was more than just to develop a uArch. He was also there to try and change mindsets and working practices at the company.

Not to mention what you wrote doesn't really make sense, if he left because he wouldn't have much to do until 7nm got sorted out... well he wouldn't have much to do even if it was sorted out. A broken node doesn't mean you completely stop future development.

Failure of 7nm is the last reason I'd think he left. First is certainly family, and second would be due to being a victim of the corporate rot over at Intel. I'm still leaning more towards the first because the second was a rumour which - so far - only appears to be half true. Time will tell if it's completely true or not.
The two aren't mutually exclusive. If he was in love with the job at Intel and a family crisis arose, he could have taken an extended sabbatical. But if you're already sick of the political BS, it's probably an easier call to just walk away in a situation like that.
 

uzzi38

Senior member
Oct 16, 2019
894
1,104
96
The two aren't mutually exclusive. If he was in love with the job at Intel and a family crisis arose, he could have taken an extended sabbatical. But if you're already sick of the political BS, it's probably an easier call to just walk away in a situation like that.
Not what I meant but sure
 

Zucker2k

Golden Member
Feb 15, 2006
1,158
514
136
Not to mention what you wrote doesn't really make sense, if he left because he wouldn't have much to do until 7nm got sorted out... well he wouldn't have much to do even if it was sorted out. A broken node doesn't mean you completely stop future development.
Jim Keller is all about results, though. Intel has boatloads of uarchs under their belt already, what's missing is the silicon to further development on those fronts.
 

moinmoin

Golden Member
Jun 1, 2017
1,657
1,594
106
Why did Intel lie and mislead regarding 10nm but now they are more forthcoming about 7nm problems?
Aurora Supercomputer.
As long as Intel has no external obligations it can wiggle itself out without having to lie, see Cannonlake as the supposed proof 10nm is working and on time. Aurora is linked to (so far implied to be Intel's) 7nm node so now they can't just do some throw away product launch to pretend everything is fine and dandy. Intel has to be able to deliver the specs promised for Aurora, on time.

Not an American, but isn't anyone else at least a bit concerned that all of the leading edge semi production is slowly moving to what is rapidly becoming the hottest geopolitical region of the planet?
Moving to? Taiwan always has been an important center of computer manufacturing. Even for PC and before chips were done there a lot of system parts like mainboards and cards have either been manufacture there or been handled by Taiwanese companies (the majority of EMS' and ODMs are Taiwanese). Taiwan also always has been in a politically unstable place since when the US officially switched allegiances from ROC to PRC back in 1979 (so if the US is concerned now, it's a situation of its own making).
 

uzzi38

Senior member
Oct 16, 2019
894
1,104
96
Jim Keller is all about results, though. Intel has boatloads of uarchs under their belt already, what's missing is the silicon to further development on those fronts.
Golden is complete and Ocean might be now (maybe, I don't really know there, I've only ever briefly heard about Golden), but aside from those two no other main Core's are actually complete. There's still a lot of development work to be done, and if Intel's roadmap is as delayed as it seems to be, they need all the help they can get on the uArch after Ocean.
 
  • Like
Reactions: Tlh97

ASK THE COMMUNITY