9 ms·

In short, Gelsinger and his team are configuring the company so Intel's internal manufacturing operations must earn Intel's business, in competition against eve

by cs702 5y ago

In short, Gelsinger and his team are configuring the company so Intel's internal manufacturing operations must earn Intel's business, in competition against every other fab-as-a-service provider in the world, including TSMC. The risk is obvious: The company's internal manufacturing operations can survive only if they figure out how to outperform competitors. The upside is that all other parts of Intel stand a good chance of surviving and thriving -- but that remains to be seen. Only one thing is certain: Gelsinger and his team have a lot of cojones.

lmilcin 5y ago
As an ex-Intel employee it is interesting to see. This seems to be sound long-term though it may cause significant shrink in the business in next couple of years to cut lines of business that are not above water. From the point of view of infinite game it makes a lot of sense to diversify and then cut any branch of business that is not profitable, before it drags entire company. While I think it is difficult to predict results, one thing is sure -- this will force a lot of change in Intel and we would likely not recognise the company ten years from now.
SavantIdiot 5y ago
> Intel's internal manufacturing operations must earn Intel's business This is putting lots of high-level employee's future earnings on the line in a far more direct manner. It will be interesting to see if they accept this challenge, or fight it in order to accept the slow decay that will still ensure at least a longer-term personal financial gain (i.e., instead of failing in 2 years, failing slowly over 10).
ignoramous 5y ago
> Intel's internal manufacturing operations must earn Intel's business, in competition against every other fab-as-a-service provider in the world, including TSMC. I think this might be an interesting retrospective application of The Innovator's Solution's 'be willing to cannibalize revenue streams in-order to avoid disruption.'
threatripper 5y ago
This means that Intel manufacturing is used where it makes sense and vice versa isn't used where it doesn't make sense. I can see those general outcomes: * Intel manufacturing is soon again better than other manufacturers in all aspects as it has been a long time. All Intel business goes to Intel manufacturing and things return to what they were. * Intel manufacturing is better in some aspects and continues to outcompete other manufacturers in some areas. Most Intel business goes to Intel manufacturing but some special business goes to other manufacturers. Some other companies use Intel where they have an advantage. Everybody wins. * Intel manufacturing cannot keep up with the market leader. Most Intel business goes to the leading Fab. Intel manufacturing cannot catch up as they don't earn enough money to do so. They start cutting costs and compete downmarket with lower tier fabs. In the long run Intel becomes fabless.
- chasil 5y ago
  The problem that the industry faces is that the economics and reliability of these chips have been undermined on the recent nodes. Sophie Wilson has said that cost-per-transistor is rising since we shrank below 22nm, and that a 7nm die cannot run more than half of its transistors at the same time without melting. Sophie addresses the cost rise at 22:00. https://m.youtube.com/watch?v=zX4ZNfvw1cw https://m.youtube.com/watch?v=zX4ZNfvw1cw Heat is more of a problem because finfet can't dissipate as well as planar, and reliability is declining at smaller nodes. "With a planar device, you do not have to bother about self-heating. There are a lot of ways to dissipate heat with a planar device, but with finFETs that is not the case. The heat gets trapped and there are few chances for that heat to get dissipated." https://semiengineering.com/chip-aging-becomes-design-problem/ https://semiengineering.com/chip-aging-becomes-design-proble... https://news.ycombinator.com/item?id=29889951 https://news.ycombinator.com/item?id=29889951 If I needed a 20-year service life, I would not choose 7nm. Edit: I did not realize that ARM1 had fewer transistors (~25k) than the 8086 (29k), and over ten times less than the 80386 (275k). Intel should have bought ARM in the 80s; instead Olivetti got them.
  - threatripper 5y ago
    This problem is not specific to Intel. Many chips are not using the smallest node sizes and old fabs get a second life producing those chips instead of CPUs. That could soon be the future for Intel fabs.
blihp 5y ago
I don't see it as that much of a risk. One of the reasons Intel has so thoroughly dominated for so long was that they were at least at parity to slightly ahead on process nodes. If they don't get back there fast, they are in real trouble. Intel likes to command a price premium and you can't do that in second place.
- n7pdx 5y ago
  It is quite a bit more than parity to slightly ahead. From 130nm to 22nm the Intel process advantage was enormous. That the horrible P4 architecture was able to compete and be cost-effective is testament to that. Intel architects and designers never had to work very hard to beat the competition. That is one of the reasons Intel was able to get away with their garbage in-house EDA tooling and processes. Intel was also able to get away with delays to fix their bugs (due to awful validation methodology). Even if Intel process catches up TSMC, they would have fix their design tooling, hire talented engineers and get rid of the entrenched hacks who built their careers on mediocrity.
- threatripper 5y ago
  I think some of the advantage comes from synergies of having microarchitecture development and process development under one roof. You can fine tune them together to get the best performance/power ratio. Even if both alone are just on par with the competition, together they are still ahead by a few iterations. Also they get out new products a bit faster. The problem is that by switching to another fab they lose these advantages.
strikelaserclaw 5y ago
decoupling
skynetv2 5y ago
This was always the case. If anyone tells you otherwise, they do not know. In fact, every team at Intel operates as a P&L ruthlessly, not just manufacturing.
ksec 5y ago
>Only one thing is certain: Gelsinger and his team have a lot of cojones. Trained and mentored by Andy Grove. Disciple of the old school "Only the paranoid survive". I expect nothing less. I hope someday he could do the next unthinkable, Once their foundry could at least compete against Samsung, open up the x86 ISA. Or clean up the ISA and call it something like AE86. In the very long term, x86 is pretty much dead. By 2030 you would expect all hyperscaler to be on ARM with x86 serving some minor legacy clients. You have PC manufacture eager to compete with Apple and ChromeBook. So x86 could loss up to 50% of its current volume by 2030. Then a long tail of 10 - 15 years decline.
- lmm 5y ago
  AE86 would be a good architecture for running in the '90s.
  - ksec 5y ago
    >running in the '90s. Is a new way I like to be Is a new way to set me free The naming is intended to show old things can be made good, and fast :) ( Nice to see somebody gets it )
DCKing 5y ago
> The company's internal manufacturing operations can survive only if they figure out how to outperform competitors. Is outperforming competitors necessary for Intel's survival? There's plenty of fabs in the world doing quite alright well behind TSMC, the vast majority of which can't even come close to Intel's current capabilities [1]. Even if Intel never succeeds with their existing process roadmap - which is not on pace to beat TSMC - they still possess some of the most advanced chip manufacturing processes in a world that's all the more dependant on chip manufacturing. GlobalFoundries got out of the race on 14nm - a 14nm not as good as Intel's 14nm and far behind Intel's 10nm - and is still posting YoY revenue growth, despite losing much of the volume they were producing for AMD over the last few years. In addition to that, I suppose that even if Intel merely succeeds in roughly keeping up with TSMC and Samsung (their current 10nm being on par with Samsung's and TSMC's 7nm would classify as "roughly keeping up", I think) there's American national security interests at play. Especially so if Intel's manufacturing capabilities are accessible to (American) third parties. No way the powers that be would let Intel's manufacturing plants go under. It's a pretty bold strategy at face value, but I think it's actually a pretty straightforward choice and the risks aren't all that existential. [1]: https://en.wikipedia.org/wiki/List_of_semiconductor_fabrication_plants https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat...
- seanp2k2 5y ago
  >No way the powers that be would let Intel's manufacturing plants go under. Perhaps that's part of the calculus too; if they know that Uncle Sam will backstop any failures, are they not also making the same play as their competitors were in the 80s? The problem is that by the mid-1980s Japanese competitors were producing more reliable memory at lower costs (allegedly) backed by unlimited funding from the Japanese government, and Intel was struggling to compete…
- cs702 5y ago
  In the short run (say, a time horizon under 10 years), I think you're right. In the long run, I think failing to outperform the competition persistently would lead to gradual irrelevance, unsustainable margins, and eventually a slow death of the manufacturing side. I mean, that's the path Intel's manufactoring operations were on before this recent change in strategy: the path to a slow death.
kelp 5y ago
It seems like echos of, and a less drastic version of the AMD/Global Foundries split. And then Global Foundries couldn't keep up, cancelled their 7nm work, and AMD has been sending more and more business to TSMC.
bsder 5y ago
> In short, Gelsinger and his team are configuring the company so Intel's internal manufacturing operations must earn Intel's business, in competition against every other fab-as-a-service provider in the world, including TSMC. This is a monumentally stupid idea and, even worse, we have seen it before. Every company that has done this in the past is gone. I would say that every VLSI designer in Intel should be printing their resumes, but I'm pretty sure that the decent ones left long ago. In addition, this completely throws away the advantage that having a fab brings at a time when it's finally important. Fabless works great when you can push a synthesis button every 18 months and your chip doubles in speed. Smart design can't keep up with that. However, once Moore's Law stops and you can't run every transistor without melting your chip, suddenly design matters again. You want to be able to do creative things with a bit of tweak from your fab.
bigbillheck 5y ago
Why stop there, make various parts of manufacturing compete against each other. Get this thing going Sears-style, it'll be lots of fun for those of us on the outside.

by cs702 5y ago

lmilcin 5y ago

As an ex-Intel employee it is interesting to see. This seems to be sound long-term though it may cause significant shrink in the business in next couple of years to cut lines of business that are not above water. From the point of view of infinite game it makes a lot of sense to diversify and then cut any branch of business that is not profitable, before it drags entire company. While I think it is difficult to predict results, one thing is sure -- this will force a lot of change in Intel and we would likely not recognise the company ten years from now.

SavantIdiot 5y ago

> Intel's internal manufacturing operations must earn Intel's business This is putting lots of high-level employee's future earnings on the line in a far more direct manner. It will be interesting to see if they accept this challenge, or fight it in order to accept the slow decay that will still ensure at least a longer-term personal financial gain (i.e., instead of failing in 2 years, failing slowly over 10).

ignoramous 5y ago

> Intel's internal manufacturing operations must earn Intel's business, in competition against every other fab-as-a-service provider in the world, including TSMC. I think this might be an interesting retrospective application of The Innovator's Solution's 'be willing to cannibalize revenue streams in-order to avoid disruption.'

threatripper 5y ago

This means that Intel manufacturing is used where it makes sense and vice versa isn't used where it doesn't make sense. I can see those general outcomes: * Intel manufacturing is soon again better than other manufacturers in all aspects as it has been a long time. All Intel business goes to Intel manufacturing and things return to what they were. * Intel manufacturing is better in some aspects and continues to outcompete other manufacturers in some areas. Most Intel business goes to Intel manufacturing but some special business goes to other manufacturers. Some other companies use Intel where they have an advantage. Everybody wins. * Intel manufacturing cannot keep up with the market leader. Most Intel business goes to the leading Fab. Intel manufacturing cannot catch up as they don't earn enough money to do so. They start cutting costs and compete downmarket with lower tier fabs. In the long run Intel becomes fabless.

chasil 5y ago
The problem that the industry faces is that the economics and reliability of these chips have been undermined on the recent nodes. Sophie Wilson has said that cost-per-transistor is rising since we shrank below 22nm, and that a 7nm die cannot run more than half of its transistors at the same time without melting. Sophie addresses the cost rise at 22:00. https://m.youtube.com/watch?v=zX4ZNfvw1cw https://m.youtube.com/watch?v=zX4ZNfvw1cw Heat is more of a problem because finfet can't dissipate as well as planar, and reliability is declining at smaller nodes. "With a planar device, you do not have to bother about self-heating. There are a lot of ways to dissipate heat with a planar device, but with finFETs that is not the case. The heat gets trapped and there are few chances for that heat to get dissipated." https://semiengineering.com/chip-aging-becomes-design-problem/ https://semiengineering.com/chip-aging-becomes-design-proble... https://news.ycombinator.com/item?id=29889951 https://news.ycombinator.com/item?id=29889951 If I needed a 20-year service life, I would not choose 7nm. Edit: I did not realize that ARM1 had fewer transistors (~25k) than the 8086 (29k), and over ten times less than the 80386 (275k). Intel should have bought ARM in the 80s; instead Olivetti got them.
- threatripper 5y ago
  This problem is not specific to Intel. Many chips are not using the smallest node sizes and old fabs get a second life producing those chips instead of CPUs. That could soon be the future for Intel fabs.

blihp 5y ago

I don't see it as that much of a risk. One of the reasons Intel has so thoroughly dominated for so long was that they were at least at parity to slightly ahead on process nodes. If they don't get back there fast, they are in real trouble. Intel likes to command a price premium and you can't do that in second place.

n7pdx 5y ago
It is quite a bit more than parity to slightly ahead. From 130nm to 22nm the Intel process advantage was enormous. That the horrible P4 architecture was able to compete and be cost-effective is testament to that. Intel architects and designers never had to work very hard to beat the competition. That is one of the reasons Intel was able to get away with their garbage in-house EDA tooling and processes. Intel was also able to get away with delays to fix their bugs (due to awful validation methodology). Even if Intel process catches up TSMC, they would have fix their design tooling, hire talented engineers and get rid of the entrenched hacks who built their careers on mediocrity.
threatripper 5y ago
I think some of the advantage comes from synergies of having microarchitecture development and process development under one roof. You can fine tune them together to get the best performance/power ratio. Even if both alone are just on par with the competition, together they are still ahead by a few iterations. Also they get out new products a bit faster. The problem is that by switching to another fab they lose these advantages.

strikelaserclaw 5y ago

decoupling

skynetv2 5y ago

This was always the case. If anyone tells you otherwise, they do not know. In fact, every team at Intel operates as a P&L ruthlessly, not just manufacturing.

ksec 5y ago

>Only one thing is certain: Gelsinger and his team have a lot of cojones. Trained and mentored by Andy Grove. Disciple of the old school "Only the paranoid survive". I expect nothing less. I hope someday he could do the next unthinkable, Once their foundry could at least compete against Samsung, open up the x86 ISA. Or clean up the ISA and call it something like AE86. In the very long term, x86 is pretty much dead. By 2030 you would expect all hyperscaler to be on ARM with x86 serving some minor legacy clients. You have PC manufacture eager to compete with Apple and ChromeBook. So x86 could loss up to 50% of its current volume by 2030. Then a long tail of 10 - 15 years decline.

lmm 5y ago
AE86 would be a good architecture for running in the '90s.
- ksec 5y ago
  >running in the '90s. Is a new way I like to be Is a new way to set me free The naming is intended to show old things can be made good, and fast :) ( Nice to see somebody gets it )

DCKing 5y ago

> The company's internal manufacturing operations can survive only if they figure out how to outperform competitors. Is outperforming competitors necessary for Intel's survival? There's plenty of fabs in the world doing quite alright well behind TSMC, the vast majority of which can't even come close to Intel's current capabilities [1]. Even if Intel never succeeds with their existing process roadmap - which is not on pace to beat TSMC - they still possess some of the most advanced chip manufacturing processes in a world that's all the more dependant on chip manufacturing. GlobalFoundries got out of the race on 14nm - a 14nm not as good as Intel's 14nm and far behind Intel's 10nm - and is still posting YoY revenue growth, despite losing much of the volume they were producing for AMD over the last few years. In addition to that, I suppose that even if Intel merely succeeds in roughly keeping up with TSMC and Samsung (their current 10nm being on par with Samsung's and TSMC's 7nm would classify as "roughly keeping up", I think) there's American national security interests at play. Especially so if Intel's manufacturing capabilities are accessible to (American) third parties. No way the powers that be would let Intel's manufacturing plants go under. It's a pretty bold strategy at face value, but I think it's actually a pretty straightforward choice and the risks aren't all that existential. [1]: https://en.wikipedia.org/wiki/List_of_semiconductor_fabrication_plants https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat...

seanp2k2 5y ago
>No way the powers that be would let Intel's manufacturing plants go under. Perhaps that's part of the calculus too; if they know that Uncle Sam will backstop any failures, are they not also making the same play as their competitors were in the 80s? The problem is that by the mid-1980s Japanese competitors were producing more reliable memory at lower costs (allegedly) backed by unlimited funding from the Japanese government, and Intel was struggling to compete…
cs702 5y ago
In the short run (say, a time horizon under 10 years), I think you're right. In the long run, I think failing to outperform the competition persistently would lead to gradual irrelevance, unsustainable margins, and eventually a slow death of the manufacturing side. I mean, that's the path Intel's manufactoring operations were on before this recent change in strategy: the path to a slow death.

kelp 5y ago

It seems like echos of, and a less drastic version of the AMD/Global Foundries split. And then Global Foundries couldn't keep up, cancelled their 7nm work, and AMD has been sending more and more business to TSMC.

bsder 5y ago

> In short, Gelsinger and his team are configuring the company so Intel's internal manufacturing operations must earn Intel's business, in competition against every other fab-as-a-service provider in the world, including TSMC. This is a monumentally stupid idea and, even worse, we have seen it before. Every company that has done this in the past is gone. I would say that every VLSI designer in Intel should be printing their resumes, but I'm pretty sure that the decent ones left long ago. In addition, this completely throws away the advantage that having a fab brings at a time when it's finally important. Fabless works great when you can push a synthesis button every 18 months and your chip doubles in speed. Smart design can't keep up with that. However, once Moore's Law stops and you can't run every transistor without melting your chip, suddenly design matters again. You want to be able to do creative things with a bit of tweak from your fab.

bigbillheck 5y ago

Why stop there, make various parts of manufacturing compete against each other. Get this thing going Sears-style, it'll be lots of fun for those of us on the outside.