GlobalFoundries Puts its 7 nm Program on Hold Indefinitely

[btarunr]

GLOBALFOUNDRIES today announced an important step in its transformation, continuing the trajectory launched with the appointment of Tom Caulfield as CEO earlier this year. In line with the strategic direction Caulfield has articulated, GF is reshaping its technology portfolio to intensify its focus on delivering truly differentiated offerings for clients in high-growth markets.

GF is realigning its leading-edge FinFET roadmap to serve the next wave of clients that will adopt the technology in the coming years. The company will shift development resources to make its 14/12nm FinFET platform more relevant to these clients, delivering a range of innovative IP and features including RF, embedded memory, low power and more. To support this transition, GF is putting its 7nm FinFET program on hold indefinitely and restructuring its research and development teams to support its enhanced portfolio initiatives. This will require a workforce reduction, however a significant number of top technologists will be redeployed on 14/12nm FinFET derivatives and other differentiated offerings.

"Demand for semiconductors has never been higher, and clients are asking us to play an ever-increasing role in enabling tomorrow's technology innovations," Caulfield said. "The vast majority of today's fabless customers are looking to get more value out of each technology generation to leverage the substantial investments required to design into each technology node. Essentially, these nodes are transitioning to design platforms serving multiple waves of applications, giving each node greater longevity. This industry dynamic has resulted in fewer fabless clients designing into the outer limits of Moore's Law. We are shifting our resources and focus by doubling down on our investments in differentiated technologies across our entire portfolio that are most relevant to our clients in growing market segments."

In addition, to better leverage GF's strong heritage and significant investments in ASIC design and IP, the company is establishing its ASIC business as a wholly-owned subsidiary, independent from the foundry business. A relevant ASIC business requires continued access to leading-edge technology. This independent ASIC entity will provide clients with access to alternative foundry options at 7nm and beyond, while allowing the ASIC business to engage with a broader set of clients, especially the growing number of systems companies that need ASIC capabilities and more manufacturing scale than GF can provide alone.

GF is intensifying investment in areas where it has clear differentiation and adds true value for clients, with an emphasis on delivering feature-rich offerings across its portfolio. This includes continued focus on its FDXTM platform, leading RF offerings (including RF SOI and high-performance SiGe), analog/mixed signal, and other technologies designed for a growing number of applications that require low power, real-time connectivity, and on-board intelligence. GF is uniquely positioned to serve this burgeoning market for "connected intelligence," with strong demand in new areas such as autonomous driving, IoT and the global transition to 5G.

"Lifting the burden of investing at the leading edge will allow GF to make more targeted investments in technologies that really matter to the majority of chip designers in fast-growing markets such as RF, IoT, 5G, industrial and automotive," said Samuel Wang, research vice president at Gartner. "While the leading edge gets most of the headlines, fewer customers can afford the transition to 7nm and finer geometries. 14nm and above technologies will continue to be the important demand driver for the foundry business for many years to come. There is significant room for innovation on these nodes to fuel the next wave of technology."

View at TechPowerUp Main Site

 

[oxidized]

This will make TSMC the only fab for potential new products from AMD next year, who knows if they can keep up with the production volume.

 

[hat]

Heh, putting a spin on delaying 7nm like it's a good thing...

As oxidized said, that leaves only TSMC (maybe) doing 7nm. I wonder if they'll be plagued by the same issues.

 

[RealNeil]

Heh, putting a spin on delaying 7nm like it's a good thing...

Belly Laugh!

 

[newtekie1]

Heh, putting a spin on delaying 7nm like it's a good thing...

Delaying 7mm and firing a bunch of people...

 

[cadaveca]

Delaying 7mm and firing a bunch of people...

Yes, because this lowers overall costs and enables them to make further profit from existing tech while improving market longevity. Good business that, and the stock price reflects that, of course.

 

[chaosmassive]

AMD be like,...

 

[Minus Infinity]

All these foundries talking smack as though jumping from 14/12nm to 7nm is a walk in the park. They are all talking crap, the technical challenges are enormous, Intel can't even get 10nm working yet. So much BS hype over last decade with unrealistic roadmaps. Finally an admission it's too tough. Still no word on how they'll stop quantum tunneling of electrons fro example, and what are the lasers for the lithography. EUV won't do it.

 

[jabbadap]

Heh, putting a spin on delaying 7nm like it's a good thing...

As oxidized said, that leaves only TSMC (maybe) doing 7nm. I wonder if they'll be plagued by the same issues.

Well there is samsung 7nm LPP. Though it might need that EUV update which scheduled to h1 2019 before to become option.

 

[robot zombie]

All these foundries talking smack as though jumping from 14/12nm to 7nm is a walk in the park. They are all talking crap, the technical challenges are enormous, Intel can't even get 10nm working yet. So much BS hype over last decade with unrealistic roadmaps. Finally an admission it's too tough. Still no word on how they'll stop quantum tunneling of electrons fro example, and what are the lasers for the lithography. EUV won't do it.

That's always in the back of my mind. I see all of this casual talk of 7nm like it's totally coming soon, but then... ...why can nobody see that path yet? My knowledge of the whole process is limited, but I know enough to know it's approaching the limits of what is physically possible. It would be nothing short of a massive feat of modern technology. On that day, the future begins anew, you know? Those are the kinds of advancements it will take. For anyone to pull this off as a "sure thing" would really be something. That's why I kinda don't jump on that bandwagon just yet. People seem to think AMD is somehow gonna find somebody who can print their revolutionary 7nm architecture reliably enough to sell at a fair price and it's just like... ...man, it truly would be revolutionary. I know it is absolutely possible, but when you say you're gonna be the first to do something nobody has done yet, I'm a little wary. I know that they're planning to try.

And don't get me wrong. I love AMD. I love their 12nm chips and I love the direction they're heading in and the impact they're making on the landscape. I make no qualms about being team red all the way. Though I don't partake in who's better - I go with whatever suits my needs best for the money... I just like AMD more as a company. I'll buy from them when I can get what I need from em.

But I gotta be realistic here. I don't think the battle for 7nm has even begun, yet. Gonna be a long, long time before we see this happen, if ever.

 

[hat]

One really has to wonder how "easy" 7nm is going to be, when Intel has failed at 10nm for so long, and now these guys are throwing in the towel "indefinitely".

 

[cadaveca]

One really has to wonder how "easy" 7nm is going to be, when Intel has failed at 10nm for so long, and now these guys are throwing in the towel "indefinitely".

True 10nm has long been the place where we knew things would get very difficult... like back when 130nm was king. It has been very interesting for me as an enthusiast to see how those predictions so long ago have come true, and to see how all these companies approach the physical limitation problems they face and how hardcore the engineering aspect of this truly is. If I remember things right, actual 10nm designs were not supposed to come until long after 2020, like 2025 or so. So to me, this is all of no concern. This was also the point where transistor density was to stop increasing, and we should be moving away from FinFET too. I also seem to recall that in order to keep Moore's law going at that point, 3D stacking was the answer. EUV was mentioned above, but that's also not something new either. Neither is the idea that going smaller gets more and more expensive, and as such, profits will decrease, or prices will go up. That's a big part of why it made a lot of sense for AMD to get rid of their foundry capabilities way back when. It also makes sense to me that GF would hold off on "7nm" indefinitely... it simply might not be profitable... at all.

 

[hat]

Huh? I thought things got cheaper on smaller nodes... at least, once you have one running successfully.

 

[The Von Matrices]

One really has to wonder how "easy" 7nm is going to be, when Intel has failed at 10nm for so long, and now these guys are throwing in the towel "indefinitely".

Huh? I thought things got cheaper on smaller nodes... at least, once you have one running successfully.

GF basically stated that they do not have the facilities to produce enough wafers to amortize the cost of 7nm R&D before it is obsolete. They're just too small of a company to afford that research. Companies like Intel and Samsung with more fabs and more production capacity can afford to spend more on R&D.

 

[robot zombie]

Huh? I thought things got cheaper on smaller nodes... at least, once you have one running successfully.

Ahh but that's the key... getting the yields up in time.

 

[cadaveca]

Huh? I thought things got cheaper on smaller nodes... at least, once you have one running successfully.

As @The Von Matrices posted, it's just not that simple. There's this thing about how long it takes to make a wafer, and how additional steps added to the lithography processes, since just a UV exposure is no longer enough on its own, complicate the foundry line, and make it take longer for each wafer produced. So they have to take into account how long it will take to get yields into a reasonable area, how many good dies they can produce, and what they can sell them for. Do keep in mind, this is manufacturing costs, and GF doesn't sell directly to the end user (businesses or home users), so there are other companies along the line that need to make a profit too. Since Companies like Samsung and Intel do sell directly to the end user, they cut out many of the middle-men and thereby have more direct access to funds, never mind the larger market shares that those companies have. This is why for a long time I have said that AMD and Intel aren't in direct competition; how they do what they do and what they offer is very different, and it's really only marketing that would have you think you need to choose one or the other.

 

[R0H1T]

Heh, putting a spin on delaying 7nm like it's a good thing...

As oxidized said, that leaves only TSMC (maybe) doing 7nm. I wonder if they'll be plagued by the same issues.

What about Sammy, didn't AMD have an option to use Samsung if GF was capacity constrained on 14/12nm though I'm uncertain it was CPU+GPU or just the latter?

 

[ShurikN]

Well AMD's biggest money maker, which will be Rome, is coming from TSMC, that was confirmed by AMD. And so is 7nm Vega that will come later this year. So the only product that could potentially get delayed is Ryzen 3000. At least until TSMC's capacity increases that is. And we were all wondering why the sudden shift to TSMC. I guess AMD knew about this way before anyone else.
I'm guessing if GloFo can make same clock speed improvement (10%) from refining 14nm (aka 12nm), we will see Ryzen 3k from them. If not, TSMC it is.
But then again the silicon for Rome and Ryzen 3000 is exactly the same, so is it even possible (and financially viable) to have one same chip from two different manufacturers?

 

[nemesis.ie]

AMD did apparently align the design such that it would work with either GF's or TSMC's 7nm.

A bit like Apple multi-sourcing some of their chips, albeit on a much, much tougher node in AMD's case.

 

[Zubasa]

Well AMD's biggest money maker, which will be Rome, is coming from TSMC, that was confirmed by AMD. And so is 7nm Vega that will come later this year. So the only product that could potentially get delayed is Ryzen 3000. At least until TSMC's capacity increases that is. And we were all wondering why the sudden shift to TSMC. I guess AMD knew about this way before anyone else.
I'm guessing if GloFo can make same clock speed improvement (10%) from refining 14nm (aka 12nm), we will see Ryzen 3k from them. If not, TSMC it is.
But then again the silicon for Rome and Ryzen 3000 is exactly the same, so is it even possible (and financially viable) to have one same chip from two different manufacturers?

I am pretty sure this is what they are doing.
Generally the highest volume parts are the low end stuff, for example the Pentium G line up is extremely popular for OEMs and small PC builders.
The APUs are actually still on 14nm right now and will be moving to 12nm soon.
AMD should have much more bargining power over GF these days,
so they might be able to make GF sell those 12nm wafers for basically nothing and use those for margin bin APUs to compete with Pentiums and Celerons.
By next year the 12nm yields should be close to perfect anyways.

 

[Valantar]

a) Why are people talking about this as a "delay"? Last I checked, "putting [something] on hold indefinitely" is not a delay, but a cancellation in all but name. It sounds extremely unlikely that they'll pack up for a year, then hire these people back and all of a sudden make it work, after all. This is not a delay.

b) Given that AMD is already sampling both Rome and Vega 7nm to customers, it's rather obvious that they're not based on a cancelled node. At least one of them is confirmed to be made by TSMC; it would be very odd if that didn't apply to both.


Now, for what this means for the semiconductor industry, I'm not quite sure, but it definitely has me worried. If the 2nd-largest fab company (outside of vertically integrated units like Intel and Samsung) is throwing in the towel on smaller litographies, that is a very worrying sign. At least Samsung makes their fab services available to others, but still - this is fast approaching monopoly territory. I definitely don't like the direction this is going in.

As for the people saying "Intel is stuck at 10nm, so of course 7nm is even more difficult": these "node sizes" are only marketing names. For example, GF14nm, TSMC 16nm and Intel 22nm are quite comparable in terms of actual transistor density and feature size. Intel 14nm is roughly comparable to the current Samsung/TSMC 10nm processes. TSMC 7nm is reportedly slightly denser than Intel 10nm, but not by much. This does of course not mean that TSMC 7nm is easier to make than Intel 10nm. But it's not more difficult either.

Anandtech has a great write-up on this. Apparently (and their reasoning makes sense, so I don't think the corporate BS bingo sheet applies here) this is purely a business/economic decision, and not due to technical issues with the process node. In short, GF has been losing money for a long time, and while they're close to finishing 7LP, they don't actually have any fabs to produce it in. Fab 8, where they make their 14/12nm products, is running at full capacity and not showing any signs of having to slow down - demand for that is high. Retrofitting it (which would be the cheapest solution) is thus not an option, as they'd lose the ability to sell their in-demand 14/12nm products. As such, they'd either have to build a new fab or expand Fab 8, both of which would cost $10-20 billion, which they don't have, and their owner is unwilling to invest. And even if they did, they'd still have less 7nm fab capacity than Samsung and TSMC, meaning that the amortized cost of R&D (which is comparable across the three) per chip produced would need to be higher - either making them more expensive than the competition, or less profitable. For a company that's lost several billion dollars over the last few years, that's not a good business plan.


This is a crying shame, but it makes sense. Hopefully this would also mean that they'll keep moving on to smaller nodes as time passes, just not fighting to stay at the bleeding edge. If they make money off 14/12nm for a few years, that could go a long way in paying for a 7nm-ish node down the line - and while it wouldn't be bleeding-edge, it'd still be attractive to buyers. Not CPU and GPU manufacturers, though, so us PC enthusiasts wouldn't see much gain. But it's sure better than nothing.

 

[FordGT90Concept]

So. Totally. Depressing.

What does this mean for Zen 2 and Navi?

 

[oxidized]

a) Why are people talking about this as a "delay"? Last I checked, "putting [something] on hold indefinitely" is not a delay, but a cancellation in all but name. It sounds extremely unlikely that they'll pack up for a year, then hire these people back and all of a sudden make it work, after all. This is not a delay.

b) Given that AMD is already sampling both Rome and Vega 7nm to customers, it's rather obvious that they're not based on a cancelled node. At least one of them is confirmed to be made by TSMC; it would be very odd if that didn't apply to both.


Now, for what this means for the semiconductor industry, I'm not quite sure, but it definitely has me worried. If the 2nd-largest fab company (outside of vertically integrated units like Intel and Samsung) is throwing in the towel on smaller litographies, that is a very worrying sign. At least Samsung makes their fab services available to others, but still - this is fast approaching monopoly territory. I definitely don't like the direction this is going in.

As for the people saying "Intel is stuck at 10nm, so of course 7nm is even more difficult": these "node sizes" are only marketing names. For example, GF14nm, TSMC 16nm and Intel 22nm are quite comparable in terms of actual transistor density and feature size. Intel 14nm is roughly comparable to the current Samsung/TSMC 10nm processes. TSMC 7nm is reportedly slightly denser than Intel 10nm, but not by much. This does of course not mean that TSMC 7nm is easier to make than Intel 10nm. But it's not more difficult either.

Anandtech has a great write-up on this. Apparently (and their reasoning makes sense, so I don't think the corporate BS bingo sheet applies here) this is purely a business/economic decision, and not due to technical issues with the process node. In short, GF has been losing money for a long time, and while they're close to finishing 7LP, they don't actually have any fabs to produce it in. Fab 8, where they make their 14/12nm products, is running at full capacity and not showing any signs of having to slow down - demand for that is high. Retrofitting it (which would be the cheapest solution) is thus not an option, as they'd lose the ability to sell their in-demand 14/12nm products. As such, they'd either have to build a new fab or expand Fab 8, both of which would cost $10-20 billion, which they don't have, and their owner is unwilling to invest. And even if they did, they'd still have less 7nm fab capacity than Samsung and TSMC, meaning that the amortized cost of R&D (which is comparable across the three) per chip produced would need to be higher - either making them more expensive than the competition, or less profitable. For a company that's lost several billion dollars over the last few years, that's not a good business plan.


This is a crying shame, but it makes sense. Hopefully this would also mean that they'll keep moving on to smaller nodes as time passes, just not fighting to stay at the bleeding edge. If they make money off 14/12nm for a few years, that could go a long way in paying for a 7nm-ish node down the line - and while it wouldn't be bleeding-edge, it'd still be attractive to buyers. Not CPU and GPU manufacturers, though, so us PC enthusiasts wouldn't see much gain. But it's sure better than nothing.

Pretty much agreed, but according to the info we can find on wikipedia (i take that they're pretty trustworthy) Intel's 10nm is just as much (slightly smaller) as small as TSMC's 7nm at least in 2 of the measurements, so well...

http://imgur.com/LpgsPwL

 

[FordGT90Concept]

What does this mean for Zen 2 and Navi?

[Forbes] All Of AMD's 7nm Processors To Be Manufactured At TSMC

GloFo is screwed.

 

[Valantar]

[Forbes] All Of AMD's 7nm Processors To Be Manufactured At TSMC

GloFo is screwed.

Possibly, yes. At least they're out of the high-end computing race. That doesn't mean they can't make a lot of money making other chips, though - their 14nm process ought to be plenty attractive to anything other than CPUs, GPUs and high-end SoCs. Not to mention that their R&D costs will be slashed to a fraction of what they've been when working on 7LP. The AnandTech post also described some intriguing development plans going forward. But ultimately, they'll become a much smaller company unless they secure a licence for something 7nm-like in a few years at least (and can afford to build a new fab or retrofit an old one by that point).