By: Adrian (a.delete@this.acm.org), November 3, 2020 2:32 am
Room: Moderated Discussions
Doug S (foo.delete@this.bar.bar) on November 2, 2020 1:51 pm wrote:
> Mr. Camel (no.delete@this.thanks.com) on November 2, 2020 1:14 pm wrote:
> > With Apple helping to fund TSMC process dev, it seems like TSMC
> > is likely to maintain its process lead for many years to come.
> >
> > Intel may have had a chance to win Apple's business but blew it. They probably
> > feel pretty stupid now that they won't have access to Apple's coffers.
>
>
> Perhaps, but Intel would still have had to execute. No one has ever suggested lack of
> investment was the reason for Intel's failure with 10nm. Once they got stuck Apple would
> have been looking for alternatives - designing an SoC for both Intel fabs and TSMC fabs
> would be a cheap hedge for them (and something they already did with the A9)
The incapacity of Intel to manufacture what they are designing begins to worry me more and more.
When I have seen their triumphant presentation of Tiger Lake at Hot Chips, I have believed that, after so many years, they might have finally found how to fix their 10-nm process.
Unfortunately, 3 months later, even if many devices with Tiger Lake have been announced, at least in Europe, I have not succeeded to notice anyone of them being sold anywhere.
On the other hand, I see Zen 3 CPUs listed at most shops nearby, so I assume that at least some limited quantities will be available at their official launch, in contrast with the apparent incapacity of Intel to deliver their products.
All this appears like the Intel 10-nm manufacturing process would be completely random, so that they get from time to time working samples of Tiger Lake or Ice Lake Server that they can show proudly, but most of time they just produce rejects and they do not seem to have any idea how that happens, otherwise they would have already found how to fix it.
Having worked in semiconductor manufacturing in the past, it is unbelievable for me how this can happen in 2020.
Yes, I have seen mysterious yield problems, but none where finding the cause would need years. In all cases the main limiting factor in improving the yields was only having enough money for extra equipment or materials, but that is exactly what Intel always had more than enough.
> Mr. Camel (no.delete@this.thanks.com) on November 2, 2020 1:14 pm wrote:
> > With Apple helping to fund TSMC process dev, it seems like TSMC
> > is likely to maintain its process lead for many years to come.
> >
> > Intel may have had a chance to win Apple's business but blew it. They probably
> > feel pretty stupid now that they won't have access to Apple's coffers.
>
>
> Perhaps, but Intel would still have had to execute. No one has ever suggested lack of
> investment was the reason for Intel's failure with 10nm. Once they got stuck Apple would
> have been looking for alternatives - designing an SoC for both Intel fabs and TSMC fabs
> would be a cheap hedge for them (and something they already did with the A9)
The incapacity of Intel to manufacture what they are designing begins to worry me more and more.
When I have seen their triumphant presentation of Tiger Lake at Hot Chips, I have believed that, after so many years, they might have finally found how to fix their 10-nm process.
Unfortunately, 3 months later, even if many devices with Tiger Lake have been announced, at least in Europe, I have not succeeded to notice anyone of them being sold anywhere.
On the other hand, I see Zen 3 CPUs listed at most shops nearby, so I assume that at least some limited quantities will be available at their official launch, in contrast with the apparent incapacity of Intel to deliver their products.
All this appears like the Intel 10-nm manufacturing process would be completely random, so that they get from time to time working samples of Tiger Lake or Ice Lake Server that they can show proudly, but most of time they just produce rejects and they do not seem to have any idea how that happens, otherwise they would have already found how to fix it.
Having worked in semiconductor manufacturing in the past, it is unbelievable for me how this can happen in 2020.
Yes, I have seen mysterious yield problems, but none where finding the cause would need years. In all cases the main limiting factor in improving the yields was only having enough money for extra equipment or materials, but that is exactly what Intel always had more than enough.