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

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

AtenRa

Lifer
Feb 2, 2009
14,000
3,357
136
You are so far off the mark man. NVIDIA created a custom 12nm node with TSMC, you think they ignore 7nm just like that? They have a deep and old partnership with TSMC, much stronger than AMD. And they ship more products, So TSMC gives them priority. Just like Apple.

AMD has partnership with TSMC as long as NVIDIA has, also AMD has way more volume than NVIDIA. Just add CPUs, GPUs and consoles and you have more than double the wafer volume than NVIDIA.
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.
NV will eventually have 7nm products but later than AMD. What is not know as of today is if AMD will have any consumer 7nm products earlier than NVIDIA. They may release VEGA 20 and EPYC 2 earlier than any 7nm NVIDIA product but they may not release any consumer based GPUs at 7nm earlier than NVIDIA.

But for 2018 and 2019 AMD will still have production in GF at 12nm for CPUs (RYZEN 2) and APUs RR 12nm. Also some of the GPUs may transfer to 12nm at GF later in the year if they see they need more volume.

As for Intel, they are still going to have to fight with 14nm Server SKUs against 7nm EPYC 2 in 2019. Also with current GF news it seems that RYZEN 3 will also come from TSMC 7nm so again in desktop Intel will also face a tough competition and still have to fight with only 14nm again.
 

french toast

Senior member
Feb 22, 2017
988
825
136
Yes, it's shocking how bad their execution has been. Especially given their starting point and funds initially:

32nm and 28nm were both delayed at least 3-6 months and used the stupid "gate-first" approach, limiting performance
20nm was canned as a failure
14nm was canned as a failure, replaced by a licenced samsung process

Just when it looked that 7nm with IBMs help was coming along, they decide to just throw in the towel. With an AMD chip about to be taped out in Q4.
Yea it has been a disaster...you forgot to mention all the money AMD has thrown down the drain with all the wafer agreement amendments and nonsense, all the money wasted on helping develop the processes, design and tape outs, we are talking hundreds of millions easily, money that would have served AMD very well indeed.
Who knows how many projects were streamlined, cut back or even cancelled due to lack of funds...Lisa streamlined and focused resources alot to push ZEN out of the door whilst cutting K12, perhaps the GPU division would not be so far behind Nvidia and we would be seeing a post GCN architecture by now.
We can forget high clocks (5ghz+) for zen desktop now, I maintain 10-15% IPC and lower my clocking predictions; 4.6-4.7ghz ST turbo...4.5ghz O/C.
This means i9 9900k is going to be a solid buy imo.

AMD has partnership with TSMC as long as NVIDIA has, also AMD has way more volume than NVIDIA. Just add CPUs, GPUs and consoles and you have more than double the wafer volume than NVIDIA.
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.
NV will eventually have 7nm products but later than AMD. What is not know as of today is if AMD will have any consumer 7nm products earlier than NVIDIA. They may release VEGA 20 and EPYC 2 earlier than any 7nm NVIDIA product but they may not release any consumer based GPUs at 7nm earlier than NVIDIA.

But for 2018 and 2019 AMD will still have production in GF at 12nm for CPUs (RYZEN 2) and APUs RR 12nm. Also some of the GPUs may transfer to 12nm at GF later in the year if they see they need more volume.

As for Intel, they are still going to have to fight with 14nm Server SKUs against 7nm EPYC 2 in 2019. Also with current GF news it seems that RYZEN 3 will also come from TSMC 7nm so again in desktop Intel will also face a tough competition and still have to fight with only 14nm again.
You raise some good points about TSMC history and AMD volume, AMD really needs to void the wafer agreement somehow and toss glofo by the wayside, they have been a lead weight round their ankles for years.

AMD needs to get a long term foundry plan in place with TSMC and SAMSUNG, sack glofo off for good.
Lisa seems to have her head screwed on so no worries IMO.
 
  • Like
Reactions: tonyfreak215

majord

Senior member
Jul 26, 2015
433
523
136
AMD has partnership with TSMC as long as NVIDIA has, also AMD has way more volume than NVIDIA. Just add CPUs, GPUs and consoles and you have more than double the wafer volume than NVIDIA.
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.
NV will eventually have 7nm products but later than AMD. What is not know as of today is if AMD will have any consumer 7nm products earlier than NVIDIA. They may release VEGA 20 and EPYC 2 earlier than any 7nm NVIDIA product but they may not release any consumer based GPUs at 7nm earlier than NVIDIA.

But for 2018 and 2019 AMD will still have production in GF at 12nm for CPUs (RYZEN 2) and APUs RR 12nm. Also some of the GPUs may transfer to 12nm at GF later in the year if they see they need more volume.

As for Intel, they are still going to have to fight with 14nm Server SKUs against 7nm EPYC 2 in 2019. Also with current GF news it seems that RYZEN 3 will also come from TSMC 7nm so again in desktop Intel will also face a tough competition and still have to fight with only 14nm again.

and back when AMD were still firing on all cyl with the GPU line , they were commonly first to new nodes.
 

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
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.
 

blublub

Member
Jul 19, 2016
135
61
101
GlobalFoundries has no deals with FinFETs only FDSOI. The only node GlobalFoundries has already projected profits(no loss) from is 22FDX.(quote)14-nm and 12-nm FinFET processes for radio-frequency, embedded memory, and low-power customers.
GlobalFoundries will also focus development efforts on its 22FDX and 12FDX processes for use with RF, analog, or mixed-signal designs that need low power, relatively low cost, and high performance.(quote)

GlobalFoundries is doing Ultra High Performance only on 22FDX/12FDX. No high performance for FinFETs. Pretty sure, 7FDX when announced will wreck every node down to 4-nm Stacked Nanosheets. GlobalFoundries has only broke up with the FinFET division. What isn't said is that the CEA-Leti/STMicro (FDSOI) division is getting the money that went for the FinFET division.

Unlike FinFETs, FDSOI needs no physical optimization for high performance or low power. Taller Fins, its planar, Shorter square fins, its planar, need doping, nope have body biasing.
This is what I was asking myself when I read the news - are they cancelling all 7nm or just FinGet 7nm, the latter would imply that 7nm could pop later in a year or to as SOI HP node ...
 

exquisitechar

Senior member
Apr 18, 2017
655
862
136
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.
Apparently, people in the know say that the process is probably broken. Wouldn't be surprised if Mubadala decided to pull the plug for other reasons, either...

I haven't been optimistic about their 7nm for a while, they were being real quiet. They've lied about things too, and with their track record, I had a reason to doubt them. Just never thought it'd be quite this bad.
It looks like Intel isn't the only one having trouble at this node.
Even TSMC missed performance targets. At least when it comes to DT, Intel is well off with their fantastic 14nm iterations until they get their 10nm in order.
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
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.

I also feel this was the only reason if one goes by public statements, they ran into the mountain of problems, realized that they won't be able to get yields anywhere near what is needed and then AMD bailed out to Taiwan.

But problem with those public statements, is that it is GF and AMD we are dealing with, kings of overpromise and underdeliver. Only insiders know the real health of process ( even if going by GF CTO it was "on track" a month or two ago).
 

exquisitechar

Senior member
Apr 18, 2017
655
862
136
But problem with those public statements, is that it is GF and AMD we are dealing with, kings of overpromise and underdeliver. Only insiders know the real health of process ( even if going by GF CTO it was "on track" a month or two ago).
Right, the CTO that also said that 12LP is an optical shrink.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Won't this be a great time for AMD to buy back glofo………………………… if they had the money? Imagine both gpu and cpu running off the same silicon. Yes, it's risky, but glofo is worth pennies right now. What are the chances of Apple jumping on an opportunity to go into the foundry business? They could look into acquiring the New York facility, for example, if not Dresden as well.
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
Right, the CTO that also said that 12LP is an optical shrink.

This has been discussed on these forums way back: GF is not publicly traded (privately owned), and they can use whatever fairy tales they fancy as their roadmaps and for other claims. In normal, non WSA bound business world, their customers would get fooled once, maybe twice max.
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
This has been discussed on these forums way back: GF is not publicly traded (privately owned), and they can use whatever fairy tales they fancy as their roadmaps and for other claims. In normal, non WSA bound business world, their customers would get fooled once, maybe twice max.

Hopefully they don't employ the same fairies that intel employs for their fairy tales. They've been leading consumers and investors on for what, 3 years now that their 10nm process is just fine? Coming just around the corner, choo choo! I wouldnt be shocked from another announcement of yet another delay.

In fact, I wouldn't be shocked if Intel has lost the process lead permanentley. In the light of fierce competition from AMD and lowered processor sales/revenue that has started with Zen and only get worse with Zen2, can they afford to remain at the bleeding edge of technology when process development is increasing exponentially and their volumes are going down? I have my doubts. By the time intel gets something going on 7, TSMC will be on 5. And they have the volume and revenue to plow into R&D. The leaders in cutting edge process technology development appears to be with Samsung and TSMC. Other than foundries developing cutting edge FDSOI that is.
 

iBoMbY

Member
Nov 23, 2016
175
103
86
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.

AMD could only move to TSMC with their CPU production, if there had been serious problems at GloFo, because the WSA is otherwise pretty strict on that. The priority for outsourcing was pretty clear before, which means CPUs are the last product for a third party manufacturer. I think most of the development is pretty recent, maybe it was around April/May when it materialized that GloFo couldn't meet AMD's demand for 7nm wafers, when they were allowed double sourcing at TSMC (AMD needs an explicit agreement from GloFo to do so), so this is why there had been some work done there already, and why this isn't such a big hit for AMD now. But the fault must be on GloFo's side, otherwise they could simply demand from AMD to place everything there, as long as they can deliver, or otherwise AMD would have to pay big fines to GloFo.
 
  • Like
Reactions: lightmanek

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
AMD could only move to TSMC with their CPU production, if there had been serious problems at GloFo, because the WSA is otherwise pretty strict on that. The priority for outsourcing was pretty clear before, which means CPUs are the last product for a third party manufacturer. I think most of the development is pretty recent, maybe it was around April/May when it materialized that GloFo couldn't meet AMD's demand for 7nm wafers, when they were allowed double sourcing at TSMC (AMD needs an explicit agreement from GloFo to do so), so this is why there had been some work done there already, and why this isn't such a big hit for AMD now. But the fault must be on GloFo's side, otherwise they could simply demand from AMD to place everything there, as long as they can deliver, or otherwise AMD would have to pay big fines to GloFo.
What I believe is that there were performance/production validation dates where the process had to achieve a certain level for further approval. It failed the last one and we have these announcements.

There is no way that the development on TSMC is recent, as AMD is already allowing testing on 7nm Zen2. AMD's had to have worked on both all along.
 

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
An opinion on AMD's use of TSMC.


This is a quote from TSMC: https://www.digitimes.com/news/a20180622PD204.html

TSMC held its technology symposium in Taiwan on June 21 during which Wei made the remark to dismiss recent speculation claiming TSMC's 7nm production yield rate had been raised at a slower-than-expected pace.

"The ramp-up of 7nm chip output will boost TSMC's overall production capacity to 12 million 12-inch equivalent wafers in 2018, up 9% from 10.5 million units in 2017, Wei indicated, without elaborating on specific orders and customers."

We can see from the above quote, 1.5 million wafers/year production capacity.

Forgive this as I posted it in another thread, but think it's relevant for us to realize that AMD can produce a lot of product at TSMC.

I'm not advocating any position here, but simply trying to understand.

Every single, admittedly "back of the envelope calculation" I do has AMD being able to produce enough CPUs for 50% marketshare. In opposition to the commonly held belief by many, I don't see production capacity as a limit to growth. We will have to look elsewhere to know why they can't achieve 50%.

The method I used was simply using Intel's server revenue and assuming a rough sales price of $80 / core. Extremely rough calcs but they can provide some guidelines and trying to make it very conservative, I assumed all revenue in the group went to CPUs and the price was very low per core. Both decisions had the effect of inflating the core production number, making it worse for AMD.

Year 2016 Intel [Server sales]
Server = $ 17.226B [29% of $59.4B]
Cores = 215 Million [$80/core]

For AMD we need 107 Million cores [50%]
Equals 13.5 million 8C dies.

Using this - http://www.isine.com/resources/die-yield-calculator

On 14nm, this is roughly 5000 wafers/month @ 216 210mm^2 die/wafer [0.1 defect rate /sq. cm] [300mm diameter]

On 7nm, I get roughly 2500 wafers/month @ 446 100mm^2 die /wafer [0.3 defect rate /sq. cm] [300mm diameter]



So even with a high defect rate, 30,000 wafers/year gives us 50% of 2016 server sales. That is 2% of TSMC 7nm production volume. Desktop and graphics, I now believe, can also be accommodated.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
I'm surprised but not exactly shocked by the news.

Obviously this is bad for the industry in general, but frankly it still amuses me quite a bit at the same time.
Given how much a certain group of people have mocked and pissed on Intel due to the issues with the 10nm process, while hyping the superiority of the 7nm LP process which now got swept under the mat...

My personal amusement aside, as long as TSMC has the manufacturing capacity to provide sufficient amount of chips to AMD I cannot really see how this could be bad.
The processes available from GlobalFoundries have frankly been garbage since the last SOI processes AMD used, so I don't know how the TSMC processes could really be any worse performance wise.

The definition of insanity is doing the same thing over and over and expecting different results.

We currently know very little about the TSMC 7nm process(es), however my gut feeling tells me that AMD might now have at least a fighting chance to reach competitive clocks compared to Intel with Zen 2.
 

Hitman928

Diamond Member
Apr 15, 2012
5,182
7,633
136
I'm surprised but not exactly shocked by the news.

Obviously this is bad for the industry in general, but frankly it still amuses me quite a bit at the same time.
Given how much a certain group of people have mocked and pissed on Intel due to the issues with the 10nm process, while hyping the superiority of the 7nm LP process which now got swept under the mat...

My personal amusement aside, as long as TSMC has the manufacturing capacity to provide sufficient amount of chips to AMD I cannot really see how this could be bad.
The processes available from GlobalFoundries have frankly been garbage since the last SOI processes AMD used, so I don't know how the TSMC processes could really be any worse performance wise.

The definition of insanity is doing the same thing over and over and expecting different results.

We currently know very little about the TSMC 7nm process(es), however my gut feeling tells me that AMD might now have at least a fighting chance to reach competitive clocks compared to Intel with Zen 2.

I think people (and to a certain degree, myself included) thought GF actually had something competitive this time. They were claiming +20% frequencies and even declaring 5 GHz+ operation and talking about how they finally got it right with 7 nm. Obviously this is a lot of marketing speak, but since we were getting so close to when actual AMD products were expected to be announced on GF 7nm without hearing any kind of back-off from GF, I started to believe them.

Unfortunately, it was all just smoke and mirrors again from GF. Their ineptitude has become almost mind boggling. I don't for a second believe that this decision is not for technical reasons. You don't invest billions into something to abandon it in the final hour except you can't make it work. If it was purely a change of business plan, they would still go ahead and release 7 nm, recoup as much of their costs as possible, and then give guidance as to their change in business model.

As others have said, I think this is only good news for AMD long term to be rid of GF and the wafer agreement though they may have some short term pains. I think AMD has probably guarded against this possibility as much as possible, but it will still cause some short term pains.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
I think people (and to a certain degree, myself included) thought GF actually had something competitive this time. They were claiming +20% frequencies and even declaring 5 GHz+ operation and talking about how they finally got it right with 7 nm. Obviously this is a lot of marketing speak, but since we were getting so close to when actual AMD products were expected to be announced on GF 7nm without hearing any kind of back-off from GF, I started to believe them.

I have been openly skeptical of 7nm hype for quite some time. I particularly had a hard time believing the 5GHz claim.

Too many have believed that 7nm was going to be a panacea. Just recently this came up again, with people pointing out how great the 28nm->16/14nm transition was. But that was a unique transition because it brought Finfet to the table. It was Finfets the were responsible for much of the gains. AFAIK, there is no comparable change this time.

But I do agree, getting rid of the GF WSA albatross can only help AMD.
 
  • Like
Reactions: Insert_Nickname

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
I'm surprised but not exactly shocked by the news.

Obviously this is bad for the industry in general, but frankly it still amuses me quite a bit at the same time.
Given how much a certain group of people have mocked and pissed on Intel due to the issues with the 10nm process, while hyping the superiority of the 7nm LP process which now got swept under the mat.

I think the only reason people were hyped about the 7nm node at GF was because it was technology from IBM. If it were purely in-house, no one would have put any faith in it at all. Perhaps everyone was still too optimistic because GF hasn’t been able to execute competently in around a decade and the expectation should have been that they’d find some way to screw things up.

I think the Intel kicking mostly comes down to people getting a certain amount of joy of spitting on the king.
 

iBoMbY

Member
Nov 23, 2016
175
103
86
Intel is struggling hard at getting 10nm to work, and it is still funny because of their previous arrogance.

GloFo failing at getting the 7nm process to work means hundreds of millions of investments lost at GloFo, AMD, and probably IBM, and others.

And we still don't know anything about what will happen to the WSA, and what the future costs for AMD will be. Of course it would be best to kill it as soon as possible now, but it still may cost them millions, or even billions.

Yes, maybe in the end this could be positive, but I don't feel like cheering just yet.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
Intel is struggling hard at getting 10nm to work, and it is still funny because of their previous arrogance.

GloFo failing at getting the 7nm process to work means hundreds of millions of investments lost at GloFo, AMD, and probably IBM, and others.

And we still don't know anything about what will happen to the WSA, and what the future costs for AMD will be. Of course it would be best to kill it as soon as possible now, but it still may cost them millions, or even billions.

Yes, maybe in the end this could be positive, but I don't feel like cheering just yet.

WSA is dead at 7nm for obvious reasons. That was probably and obvious contractual must have. You can only get the contractually obligated business if you can actually deliver.

AMD will probably have to keep producing their 12nm-14nm parts at GF, but did you notice AMD emphasizing they are "All in" on 7nm. IMO this is an opportunity for AMD to put GF in the rear view mirror.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
I think sometime this year Intel decided that there was no pressing need to get a line of 10nm chips out.

14nm has proved to be able to hold the fort quite well, I think.

I don't think the lack of a 7/10nm lineup is going to bother Intel nearly as much as people thought it would.

I think they have weathered the 10nm failure storm okay.

They just have to weather AMD's incursions now.
 

iBoMbY

Member
Nov 23, 2016
175
103
86
WSA is dead at 7nm for obvious reasons. That was probably and obvious contractual must have. You can only get the contractually obligated business if you can actually deliver.

AMD will probably have to keep producing their 12nm-14nm parts at GF, but did you notice AMD emphasizing they are "All in" on 7nm. IMO this is an opportunity for AMD to put GF in the rear view mirror.

The WSA is not dead as of now, since there hasn't been any SEC filing saying it is dead. At least I'm pretty sure that would be a material event ...
 

Hitman928

Diamond Member
Apr 15, 2012
5,182
7,633
136
I have been openly skeptical of 7nm hype for quite some time. I particularly had a hard time believing the 5GHz claim.

Too many have believed that 7nm was going to be a panacea. Just recently this came up again, with people pointing out how great the 28nm->16/14nm transition was. But that was a unique transition because it brought Finfet to the table. It was Finfets the were responsible for much of the gains. AFAIK, there is no comparable change this time.

But I do agree, getting rid of the GF WSA albatross can only help AMD.

Finfets didn't bring above average gains. Finfets allowed for gains to continue where shrinking bulk nodes essentially broke. Bulk nodes down to about 28 nm worked fine (not trying to trivialize the work to get there, just the end result) but after that, bulk stopped being viable.
 
  • Like
Reactions: lightmanek

AtenRa

Lifer
Feb 2, 2009
14,000
3,357
136
I think sometime this year Intel decided that there was no pressing need to get a line of 10nm chips out.

14nm has proved to be able to hold the fort quite well, I think.

I don't think the lack of a 7/10nm lineup is going to bother Intel nearly as much as people thought it would.

I think they have weathered the 10nm failure storm okay.

They just have to weather AMD's incursions now.

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.