[Anandtech]: GlobalFoundries Stops All 7nm Development !!

[dark zero]

Holy...that's quite the news. And that cannot be good for GF, pouring billions in and then giving up on it about the time they were ready to start moving to production. Either they had serious issues that couldn't be overcome, or they're about to be run into the ground by profit above all shareholders that can't think beyond the next few quarters.

While I'm sure AMD will be glad that this would seemingly end their wafer deal, I can't imagine AMD is super pleased that they lost an option for production. This means they're going to have to compete with Nvidia, Apple, and many others for production, and AMD does not have the same resources as those others. Now, maybe TSMC won't be too constrained for it to cause problems, but that also means they'll be feeling less pressure to remain competitive and that isn't a good thing for the whole industry. Hopefully Samsung will be competitive.

That would explain why AMD was aggressive in going after TSMC 7nm. I'm sure they were aware that GF was having issues. I was pretty surprised finding out that both Zen 2 and Vega 20 were TSMC (I actually kinda figured the latter would be, with GF probably handling the smaller Polaris level GPU chips, but Zen 2 EPYC being TSMC was pretty big news I thought, and I guess a harbinger). This might also explain that there was talk that APUs might not move to 7nm quickly. Oh, and I think Sony and Microsoft have TSMC fab there, so AMD is potentially going to be competing against partners for production capacity (but it could be beneficial in that they hopefully will get info/data from them to help with their own production).

This could be a big opening for Intel if they want to open their fabs more.

Seems that AMD and Samsung are gonna strike together.

But seems that GF is pretty much on a critical condition to not to say this implies they lost their funds (arabs leaving?) and they are pretty much on a condition they will sell they have and leave the industry...

There is not known 22nm chips made by GF and their 12nm chips are only AMD ones... So, they are on a position that losing AMD means Game Over and it happened.

 

[Muhammed]

AMD has way more volume than NVIDIA. Just add CPUs, GPUs and consoles and you have more than double the wafer volume than NVIDIA.

Nope, NVIDIA sells more GPUs than AMD sells CPUs + GPUs COMBINED. That might change when Ryzen and Epyc become more popular though.

NVIDIA probable created the 12nm because they wouldnt have enough 7nm volume in 2018-2019 because AMD probable knew a lot earlier that GF 7nm will not go in to production. So it seems that AMD clearly booked a lot of 7nm capacity early for 2018-2019 leaving NVIDIA out.

No way in hell that happened. NVIDIA would be involved in 7nm just as much as AMD if not more. How can NVIDIA create a custom 12nm node and ignore 7nm completely? You think they are morons or what? If they got involved in a CUSTOM 12nm, you KNOW they are involved in a standard 7nm!

 

[kawi6rr]

Nope, NVIDIA sells more GPUs than AMD sells CPUs + GPUs COMBINED. That might change when Ryzen and Epyc become more popular though.

Got a source for that or are you just blowing... smoke! Means nothing with out a source.

 

[JDG1980]

It was reported in late April that the PS5 APU was likely to be fabbed at TSMC, not GloFo. In retrospect, this should have been a warning sign; if AMD was going to produce any 7nm products at GloFo, consoles would surely top the list, since they are high-volume (and can thus be a major contributor towards the WSA) and aren't generally as demanding as enthusiast or pro-grade chips in terms of cutting-edge performance and efficiency.

 

[JDG1980]

And if Apple has a product that is selling hot or they expect it to, they can gobble up more of the capacity for longer time. Heck, if Apple just decides that they want to "compete" by just hogging as much capacity as possible, making it so that other products can't get produced, they could. They have the money to be able to lock up TSMC production and create headaches. So far, Apple hasn't had a reason to do that, so its been mostly fine.

This wouldn't be a very effective strategy for Apple since their biggest competitor in the smartphone game is Samsung, which has its own cutting-edge fabs. And TSMC wouldn't want this to happen, since their business model is based around a pay-for-play model. If Apple hogs all of TSMC's capacity over a long period of time, competitors go to Samsung, and that's a strategic blow that TSMC doesn't want to incur.

 

[wahdangun]

Could we be have this reversed?

It appears to the public, that GloFlo was moving ahead full speed until this announcement. Is it possible that AMD decided to abandon GloFlo for 7nm as they were having delays. I think that AMD was their lead/only customer for 7nm and thus loosing them resulted in no demand for the process.

The problem with that theory is if AMD abandoning GF for tmsc then AMD must pay WSA penalties, I guess the real problem was GF main investors think than bleeding edge process not worth it, even if GF still on track for their 7nm

 

[EXCellR8]

slightly unrelated but TSMC should really redesign their website...

 

[NostaSeronx]

There is another case... AMD/Qualcomm switched GloFo FinFET orders to GloFo FDSOI orders.

With the Raven Ridge 2018, it is being sold with FT4 Stoney Ridge Refresh 1(2017) and 2(2018).
With the Snapdragons; 205(2017), Snapdragon Wear, Some of the Low-end Automotive/Embedded/Home-Hub SoCs

So, the focus is more on what GlobalFoundries can mass produce. While also allowing for big companies to have two baskets rather than one basket.

Add other big GloFo FDSOI winners:
Rockchip
MediaTeks (Standalone and Wearables)

With GlobalFoundries' Chengdu
“MediaTek established our site in Chengdu back to 2010. Chengdu is quickly becoming an international destination for cutting-edge technology companies. We are thrilled to see continued investment to establish the region as a center of excellence for both manufacturing and design of GF’s FDX technology.”
- Joe Chen, Executive Vice President and Co-COO of MediaTek

“The Chengdu Hi-tech Industrial Development Zone is quickly becoming an international center for technology innovation, and we are delighted to see Chengdu’s growing partnership with GF on advanced semiconductor design and manufacturing.”
- Spencer Pan, AMD President, Greater China

“We are pleased to see Chengdu investing in an ecosystem to support GF’s innovative FDX technology. These types of partnerships are critical to supporting China’s growing fabless semiconductor industry and helping companies like Rockchip differentiate in the mobile SoC market.”
- Min Li, CEO of Rockchip

etc.

Malta was validated for 22FDX in 2015 and 12FDX in 2017. So, Malta joining up with Dresden for more FDSOI orders means more profits for GloFo. 22FDX and 12FDX has a larger consumer base than 7LP. People rather buy "cheap", rather than buy "performance".

Of the bunch of customers stuck on 40nm/28nm.
- Go to expensive 14LPP/12LP
- Go to cheap 22FDX.

Of the some customers on 14LPP/12LP.
- Go to expensive 7LP
- Go to cheap 12FDX.

There is also some subtle implied implications of this:
https://ieeexplore.ieee.org/document/8429035/
Singapore, GloFo's Fab 7, might also be making the jump to 22FDX. If all four fabs have 22FDX consumers that is >5x the profit of one FinFET Fab.

 

[beginner99]

maybe GF realized that spending billions on a so-so process to get the money back in form of WSA payments and not actually producing stuff wasn't a sustainable business beyond 2022 and the WSA expiry date.

They were probably also behind schedule or even had severe issues like intel. But let's assume they could have delivered in mit 2019. That would only give them roughly 2.5 years to make money from this process with the WSA, probably not long enough to recuperate the costs.

I would like to see how Intel will be able to compete against AMDs 7nm (TSMC) EPYC 2 from early 2019. They will lose in every department for more than a year. EPYC 2 at 7nm vs Intels 14nm could be so disruptive that AMD could increase double digit market share in just a year.
I will also like to see how they will compete against 12nm RR in mobile space.

this assumes that TSMC has enough volume assigned for AMD to even be able to grab such a high market share.
Plus assumes Procurement actually makes sane decisions (unlikely) and doesn't just buy intel.

It was reported in late April that the PS5 APU was likely to be fabbed at TSMC, not GloFo. In retrospect, this should have been a warning sign; if AMD was going to produce any 7nm products at GloFo, consoles would surely top the list, since they are high-volume (and can thus be a major contributor towards the WSA) and aren't generally as demanding as enthusiast or pro-grade chips in terms of cutting-edge performance and efficiency.

Current consoles are also made at TSMC. So that wasn't a surprise at all and confused me as well. My guess is that GF simply does not have the capacity needed.

 

[AtenRa]

Nope, NVIDIA sells more GPUs than AMD sells CPUs + GPUs COMBINED. That might change when Ryzen and Epyc become more popular though.

Even if you see the financial statements of Q2 2018 (FY 2019 for NV) you will see that cost of goods for AMD is higher than NVIDIA.

No way in hell that happened. NVIDIA would be involved in 7nm just as much as AMD if not more. How can NVIDIA create a custom 12nm node and ignore 7nm completely? You think they are morons or what? If they got involved in a CUSTOM 12nm, you KNOW they are involved in a standard 7nm!

Nobody said NVIDIA is not involved with 7nm, but for 2018-2019 with both Apple and AMD using most of 7nm volume NVIDIA didnt have any other option but just to go for 12nm.

 

[beginner99]

Nope, NVIDIA sells more GPUs than AMD sells CPUs + GPUs COMBINED. That might change when Ryzen and Epyc become more popular though.

You forgot the consoles.

 

[AtenRa]

this assumes that TSMC has enough volume assigned for AMD to even be able to grab such a high market share.
Plus assumes Procurement actually makes sane decisions (unlikely) and doesn't just buy intel.

For the entire 2019 I believe they will have enough 7nm volume. By the end of 2019 AMD could easily have double digit server market share.

 

[NTMBK]

There is another case... AMD/Qualcomm switched GloFo FinFET orders to GloFo FDSOI orders.

I take it that you missed the multiple statements from AMD about how they are "all in" on 7nm and moving their CPUs to TSMC. They're not going SOI again.

 

[CatMerc]

I take it that you missed the multiple statements from AMD about how they are "all in" on 7nm and moving their CPUs to TSMC. They're not going SOI again.

There's a 12nm Raven Ridge coming later this year. Other than that yeah 7nm.

 

[iBoMbY]

To my surprise Mubadala aparently took a sane decision.

But only after selling 22,000,000 AMD shares on 7/30/18 ... http://ir.amd.com/ownership-summary

 

[iBoMbY]

There is no way there isn't isn't a process requirement to go along with the wafers. It's probably more transitional. 90% 14nm in 2017 85% 14nm in 2018 (covering for pipe cleaners) maybe 65% 14nm and 30% 7nm. Maybe as noted above AMD has the ability produce products on every node and only nodes where GF doesn't have an equivalent do they have to buy from them or face a penalty (charging for every order they make to other partner) until the WSA is met.

But there is one big screaming obvious statement. If AMD was required to maintain the WSA. They don't have the money to break contracts like that. GF's and AMD's announcements mean one thing and one thing only. They are not beholden to the WSA, at least as far as 7nm go and won't be held liable for those purchases. So again dead but not dead.

That exact part is censored, but there is no obvious way out for AMD, the only way is mutual termination, or GloFo literally burning to the ground. The only thing that is sure is they are allowed to procure more than the default rate from external sources in an event like this, but it doesn't say they have to pay less fines for it ...

 

[NostaSeronx]

I take it that you missed the multiple statements from AMD about how they are "all in" on 7nm and moving their CPUs to TSMC. They're not going SOI again.

The lead TD for 7LP also worked on 12FDX. So, if it was known 7LP would be cancelled by AMD. Then, the 7th amendment WSA will reflect FDSOI only for GlobalFoundries.

If you won't and can't use 7LP, you better give us your money for 12FDX.

I also have this awkward french report from 2017:

So, TSMC is N7 and GlobalFoundries is 12FDX for 2019 for AMD. TSMC will be targeting the high margin premium market with 7nm. GlobalFoundries will be focused for the fast return low margin value market with 12nm SOI.

12LP partnership was 2018 and only Pinnacle Ridge got the treatment. Raven Ridge 2018 is still 14LPP+ like Raven Ridge 2017. The only reason to do that is if Picasso with Dali is 12FDX.

 

[CatMerc]

Picasso is 12nm. Stop dreaming of your 12FDX Raven Ridge, the process isn't even available yet.

There's also no point in doing it on 14LPP.

 

[DisEnchantment]

There's a 12nm Raven Ridge coming later this year. Other than that yeah 7nm.

It actually makes more sense to get mobile SKUs on FDSOI, because ultimate performance and density are secondary to efficiency.
How big would RR be on 12FDX? Also how big is the performance tradeoff vs efficiency. I mean in a mobile SKU having ultimate performance is useless if the chip is thermally constrained 90% of time.
Cost savings, most likely, would be much welcomed too,
If this is the case I would say this is indeed and interesting move.
EDIT : 12FDX is not ready
Even Samsung is planning smaller FDSOI nodes

 

[NostaSeronx]

Picasso is 12nm. Stop dreaming of your 12FDX Raven Ridge, the process isn't even available yet.

There's also no point in doing it on 14LPP.

-> 12FDX, the next node, is a high volume, high capacity play for us. - https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries
-> I asked Gary about the status of 12FDX, he noted that customers are just now taping out on 22FDX and so they are focusing on 22FDX and 7LP and not really pushing on 12FDX but he said although early in development it is "fairly mature". - https://www.semiwiki.com/forum/content/7496-imec-technology-forum-gary-patton-globalfoundries.html
-> We have 12FDX here and we know we can do the device – we are actually pretty close to the performance targets already. We have been doing work on 12FDX here in NY for over a year. ... We expect to be taking risk production on the parts early next year (2019), so we are pretty far along with the technology. - https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries

Note that the risk production wasn't what they said last year:
September 29th, 2017 -> "with prototypes expected in the second half of 2018 and volume production beginning in 2019."

With no 7LP:
https://i.imgur.com/rYVhmw2.jpg
Only 12FDX is in that sphere.

Uh oh, the focus on 22FDX and 7LP means 12FDX gets all the 7LP customers. Especially, if *puts on sunglasses* they are bleeding edge.

 

[NTMBK]

-> 12FDX, the next node, is a high volume, high capacity play for us. - https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries
-> I asked Gary about the status of 12FDX, he noted that customers are just now taping out on 22FDX and so they are focusing on 22FDX and 7LP and not really pushing on 12FDX but he said although early in development it is "fairly mature". - https://www.semiwiki.com/forum/content/7496-imec-technology-forum-gary-patton-globalfoundries.html
-> We have 12FDX here and we know we can do the device – we are actually pretty close to the performance targets already. We have been doing work on 12FDX here in NY for over a year. ... We expect to be taking risk production on the parts early next year (2019), so we are pretty far along with the technology. - https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries

Note that the risk production wasn't what they said last year:
September 29th, 2017 -> "with prototypes expected in the second half of 2018 and volume production beginning in 2019."

With no 7LP:
https://i.imgur.com/rYVhmw2.jpg
Only 12FDX is in that sphere.

Uh oh, the focus on 22FDX and 7LP means 12FDX gets all the 7LP customers. Especially, if *puts on sunglasses* they are bleeding edge.

So... what your quotes are telling me is that 12FDX is nowhere near ready? So AMD would be better porting designs to TSMC 7nm and shipping this year, instead of shipping late next year on an inferior process?

 

[CatMerc]

And those performance figures are only valid for small devices. Garry Patton himself said that FinFETs do better than FDSOI on devices like what AMD creates.

 

[NostaSeronx]

So... what your quotes are telling me is that 12FDX is nowhere near ready? So AMD would be better porting designs to TSMC 7nm and shipping this year, instead of shipping late next year on an inferior process?

7LP/12FDX share many designers. If 12FDX isn't ready, nor would be 7LP.

-> 7LP MOL layer owner and 12FDX FEOL layer owner
-> 12FDX - Team responsible for initial delivery of OPC solutions for selected layers. • 5LP/7LP+ OPC – Team beginning work on 7LP+ EUV and 5LP OPC.
-> Endura impulse chamber, CVD-Co Cap, liner and fill, Pre-clean chamber and Producer and Precision CVD products for customer 7nm and 12FDX technology.
-> ATD 7nm, 12FDX
-> structures and driving its use across the test site design community within GLOBALFOUNDRIES for 12FDX (12nm FD-SOI) and 7LP (7nm bulk) technologies.
-> Working on 7nm,14LPP,12FDX
-> Worked on leading edge technologies, like 7nm, 12FDX, 14nm, 22FDX

Since, the rest of the R&D of 7LP/7LP+/5LP/3LP money is going to 22FDX/12FDX(/7FDX) its pretty clear what is the superior process.

12FDX @ Malta started in 2016.
12FDX @ Malta completely validated in 2017
12FDX starts process to move to Dresden in 2018 for production in 2019. (Oh, wait most of our customers went 22FDX... whelp guess we will delay that to 2020 and 2021 eqv)
7LP is canned. (Whelp, 22FDX was faster than 7LP anyway.)
12FDX @ Malta ramps up with tapeouts now rather than later.
12FDX starts at Malta then Dresden and then back to USA with East Fishkill.

 

[CatMerc]

7LP/12FDX share many designers. If 12FDX isn't ready, nor would be 7LP.

-> 7LP MOL layer owner and 12FDX FEOL layer owner
-> 12FDX - Team responsible for initial delivery of OPC solutions for selected layers. • 5LP/7LP+ OPC – Team beginning work on 7LP+ EUV and 5LP OPC.
-> Endura impulse chamber, CVD-Co Cap, liner and fill, Pre-clean chamber and Producer and Precision CVD products for customer 7nm and 12FDX technology.
-> ATD 7nm, 12FDX
-> structures and driving its use across the test site design community within GLOBALFOUNDRIES for 12FDX (12nm FD-SOI) and 7LP (7nm bulk) technologies.
-> Working on 7nm,14LPP,12FDX
-> Worked on leading edge technologies, like 7nm, 12FDX, 14nm, 22FDX

Since, the rest of the R&D of 7LP/7LP+/5LP/3LP money is going to 22FDX/12FDX(/7FDX) its pretty clear what is the superior process.

12FDX @ Malta started in 2016.
12FDX @ Malta completely validated in 2017
12FDX starts process to move to Dresden in 2018 for production in 2019. (Oh, wait most of our customers went 22FDX... whelp guess we will delay that to 2020 and 2021 eqv)
7LP is canned. (Whelp, 22FDX was faster than 7LP anyway.)
12FDX @ Malta ramps up with tapeouts now rather than later.
12FDX starts at Malta then Dresden and then back to USA with East Fishkill.

When I say that, I should qualify: you are not going to get that from a big chip. If you are making a big chip, you got tons of wire capacitance, so you are not going to beat a FinFET design. FinFET has huge drive current, so if you're dealing with smaller chips, wiring capacitance isn't as a big, but gate capacitance of the transistor is more important. FD-SOI has less gate capacitance than FinFET because you don't have the gate wrapping around the whole device. So it is ideally suited for smaller chips, I say less than 150 square millimeters.

Garry Patton of Global Foundries

7LP wasn't cancelled because 12FDX is better. It was cancelled because GlobalFoundries is quitting the leading edge race and going for more financially viable routes.
And no, 22FDX isn't anywhere near 7LP either.

 

[guskline]

Smart move by AMD and an economically necessary and realistic move by Global Foundries.