DigiTimes: Intel To Outsource 14nm Chips To TSMC

[PaulAlcorn]

DigiTimes is reporting that Intel has outsourced 14nm H310 chipsets and 300-series processors, meaning the Coffee Lake chips, to TSMC to aid in production.

DigiTimes: Intel To Outsource 14nm Chips To TSMC : Read more

 

[jpe1701]

Recommend Epyc? I love it. It's great to see competition again.

 

[hannibal]

The problem is that if there is production capasity problem. It means that the prices of all products comes upp... as there were production capasity problem in memory department, no it goes to cpu and gpu... and only can means higher prices to us.

 

[stdragon]

Recommend Epyc? I love it. It's great to see competition again.

Going to be real interesting how that's handled. With any virtualization technology (VMWare, HyperV, Citrix...etc), you can NOT mix AMD with Intel in a single virtualization host cluster. You can however create AMD and Intel clusters with like CPUs, but even then the CPU generation feature-set will be to the lowest common denominator. This is important when having to do things like vMotion (live migration of a running VM from one host to another).

 

[littleleo]

The Intel CPU shortage is worse then the public realizes yet. Prices have been skyrocketing on popular models and lots of shortages on Intel are popping up. Expect AMD models to start going up in price and going short as SI have to go AMD to provide systems to their customers. The Intel shortages are expected to last until at least Q1 2019.

 

[JQB45]

What is causing the shortages? Lack of capacity to meet demand? Or something more serious?

 

[Kaz_2_]

Intel has limited fabs , they need to convert some 14nm to 10nm. the problem is 10nm isnt working to bring up capacity. reducing 14nm capacity to bring more 10nm capacity. in the end its big lose to Intel. They simply dont have the capacity to fulfil everyone therefore AMD will take over the remaining business. AMD is on their way up higher

 

[jasonelmore]

Intel finally gave us 6 core consumer CPU's but they did as they were transitioning to 10nm. What we seeing is a lot of consumer pimped up demand when Intel can't handle the demand due to the 10nm transition. So far Intel hasn't shown it can make a 4+ core chip on 10nm in any meaningful volume. EUV will solve the scaling problem for the time being if they can actually get it working.

 

[Giroro]

A company like Intel can't just easily switch over to another fab. They have to wait in line behind work that has already been scheduled, and when they finally get in the silicon itself takes a long time (on the scale of a month) to produce.

I don't think this is true, but if it were, TSMC would be maybe 6 months away from delivering the first shipment (but probably more like a year). It doesn't make sense for Intel to increase 14nm production for Q2 2019 ... unless Intel has decided it's 10nm is never going to happen.

 

[InvalidError]

It doesn't make sense for Intel to increase 14nm production for Q2 2019 ... unless Intel has decided it's 10nm is never going to happen.

It could make sense if you were still manufacturing a significant volume of 22nm chipsets and other support functions that don't require cutting-edge process, want to shut down those 22nm plants for upgrade to 7/10nm but don't have enough spare 10/14nm capacity to accommodate the transition.

Chipsets and other support chips will stay on 14-22nm for many more years, especially if 10nm and beyond continue being more challenging than expected.

 

[jimmysmitty]

Recommend Epyc? I love it. It's great to see competition again.

Going to be real interesting how that's handled. With any virtualization technology (VMWare, HyperV, Citrix...etc), you can NOT mix AMD with Intel in a single virtualization host cluster. You can however create AMD and Intel clusters with like CPUs, but even then the CPU generation feature-set will be to the lowest common denominator. This is important when having to do things like vMotion (live migration of a running VM from one host to another).

Its a big reason why even with a good chip like Epyc AMD is not going to be taking a ton of the market share like some people want. It would be more gradual if at all since those bigger companies are not willing to invest that much to change over for some better performance.

 

[bit_user]

Intel intends to outsource 14nm chipsets and "several other" 300-series processors, meaning the Coffee Lake chips, to TSMC to aid in production.

Chipsets. Not CPUs.

 

[bit_user]

Chipsets and other support chips will stay on 14-22nm for many more years, especially if 10nm and beyond continue being more challenging than expected.

According to this:
https://www.tomshardware.com/news/14nm-processor-intel-shortage-9000-series,37746.html

Intel's recent 300-series chipset refresh found its new chipsets coming to market with the 14nm process, which is necessary to meet California's new power standards.

So, that increase in demand for 14 nm probably could not be accommodated and is now being shunted off to TSMC.

 

[kerberos_20]

cpu has to be put somewhere, if no chipset, no mobo, cpu cant work without mobo

 

[bit_user]

Recommend Epyc? I love it. It's great to see competition again.

I wonder how much extra demand AMD can absorb, before they must also start increasing prices.

Sharing the same dies between server, HEDT, and upper-mainstream desktop has advantages and drawbacks, for them. The benefit is they can offset weak demand in one segment with healthy demand in another. But, if demand really surges in any of those segments (particularly the 4-die EPYCs), it will probably affect pricing in all.

 

[PaulAlcorn]

Intel intends to outsource 14nm chipsets and "several other" 300-series processors, meaning the Coffee Lake chips, to TSMC to aid in production.

Chipsets. Not CPUs.

That might be a distinction that is lost in translation. As we noted, it doesn't seem logical that Intel would/could shift CPU's over, but we also aren't the only media outlet to make the (possibly erroneous) decision that DigiTimes is referring to CPUs as well here (emphasis mine):

Intel intends to give priority to its high-margin products mainly server-use processors and chipsets amid its tight 14nm process capacity, and therefore plans to outsource the production of its entry-level H310 and several other 300 series desktop processors to TSMC, the sources indicated.

 

[InvalidError]

That might be a distinction that is lost in translation. As we noted, it doesn't seem logical that Intel would/could shift CPU's over

Intel makes many CPU lines. While it is highly unlikely that Intel would outsource desktop and server CPUs to an external fab, embedded, IoT, mobile, etc. wouldn't be far-fetched - let the fabs that already do low-power, lower-margin 14nm stuff for everyone else take care of your own too.

With Intel swallowing Altera, I'll hazard a guess that a fair amount of 14nm volume is going into those $500-7000 FPGAs.

 

[PaulAlcorn]

That might be a distinction that is lost in translation. As we noted, it doesn't seem logical that Intel would/could shift CPU's over

Intel makes many CPU lines. While it is highly unlikely that Intel would outsource desktop and server CPUs to an external fab, embedded, IoT, mobile, etc. wouldn't be far-fetched - let the fabs that already do low-power, lower-margin 14nm stuff for everyone else take care of your own too.

With Intel swallowing Altera, I'll hazard a guess that a fair amount of 14nm volume is going into those $500-7000 FPGAs.

The only problem with farming out low power parts is the margin game, though. You really need scale and native manufacturing to pull some solid profits out of those, but perhaps Intel could shave a few points off those legendary 60%+ margins for the sake of actually shipping product

 

[msroadkill612]

It fits a pattern from which the only conclusion is intel is in chaos.

They have clearly over committed to 10nm, and not just they, but also their long suffering poodle partners like laptop makes, are pretty screwed. I expect a new enthusiasm for amd mobile from them.

I even suspect they may have begun transitioning fabs irreversibly.

They are failing to meet existing demand, and surreally plead the proliferation of chipsets (a result of their price gouging. for which they are widely despised).

 

[Tanyac]

I'm caught between wondering if all of these shortages that cover pretty much most major components of PCs and laptops are due to pure incompetence or a new profit strategy. Perhaps both.

 

[InvalidError]

The only problem with farming out low power parts is the margin game

Spreadtrum, Mediatek and a swathe of other budget SoC designers outsource their cheap chips to TSMC, UMC, GloFo, etc. Seems to work perfectly fine for everyone else.

When your own in-house wafer processing can't meet demand, the margin on low-margin products doesn't matter much: you'd much rather spend your finite wafers on flagship products that earn over 1M$ per wafer than support functions that earn less than 100k$ and may have an effective opportunity cost exceeding 900k$ per wafer to keep in-house. If it costs 5k$ extra to outsource low-margin wafers, that's comparably dirt-cheap.

 

[hecksagon]

Wonder if some of this has to do with all of AMD's production coming from TSMC. If anybody could get priority at TSMC over AMD it would be Intel. A couple months later AMD has a shortage.