AMD rumored to be using TSMC's 7nm HP node for Rome and Vega

[piesquared]

Pretty significant if true. Rome has probably been planned for TSMC all along, which is why AMD seems to be executing to their roadmap flawlessly also announcing there will be no delays with the full transition to TSMC's 7nm. It's crazy how fast they are pumping out products with this really aggressive roadmap. And they even moved 7nm Vega for Radeon Instinct ahead, all on a significantly lower budget than either of their 2 competitors.
It's shaping up pretty good for them, releasing the 2 high performance parts that will cost more to build but with higher MSRP and a higher margins right now in the current competitive landscape looks to be really good timing. The Ryzen 2000 series seems to be selling really well on 12nm and if they don't release next gen Ryzen until next year around 3Q they won't experience much of an Osbourne effect. Especially if they are socket AM4 compatible and enthusiasts can just drop it in their existing boards and get a big performance lift. There will likely be some new features on the new boards but if it is a big drop in performance upgrade that would be insane.
I'm curious to see what the clocks are on Vega 20. They brought over a bunch of CPU engineers to RTG to work on getting the clocks up. If they managed that plus the rumored >%35 performance increase from TSMC's 7nm HP over their 16nm AMD might have a killer Radeon Instinct GPU on their hands too. If the consumer Vega cards had comparable clock speeds to their competing Pascal cards, the performance difference would have been practically if not completely eliminated. This is probably why Nvidia is trying to sell ray tracing as the next coming of Christ. If ray tracing doesn't sell they would have to compete with a much bigger die. They will be a fair ways behind AMD with their 7nm product.
The GPU competitive landscape seems yet to be determined but should know in a couple months when they reveal their Instinct product. It appears likely that Zen2 will have intel on the ropes though.

 

[IEC]

Zen and Zen+ had certain compromises in order to make it to market, including being designed for low power nodes. Despite the compromises they are very good offerings, particularly in value (and also in forcing the Empire Intel to respond).

It will be interesting to see just how much 7nm and a bigger budget will get them for Zen 2.

 

[piesquared]

Zen and Zen+ had certain compromises in order to make it to market, including being designed for low power nodes. Despite the compromises they are very good offerings, particularly in value (and also in forcing the Empire Intel to respond).

It will be interesting to see just how much 7nm and a bigger budget will get them for Zen 2.

The tone from Lisa Su and Mark Papermaster seems quite positive. They were aiming at what they thought would have been intel's 10nm product and thought they would have done very well against it so I think Zen 2 is the real deal and now it'll be going against intel's 14nm. It will be interesting to see what they did with it. They likely improved the Infinity Fabric substantially for sure.

 

[beginner99]

They were aiming at what they thought would have been intel's 10nm product and thought they would have done very well against it so I think Zen 2 is the real deal and now it'll be going against intel's 14nm

It will give AMD and advantage in the server space for sure due to general low clocks and importance of power use. But on desktop, intel 14nm+++?? is actually better than intels 10nm would be because it just clocks that much better.

 

[Mopetar]

Even with a 14+++ process I think Intel only really has an edge in the gaming segment. AMD APUs in 7nm are going to be a much better choice for most consumers. People also forget that AMD was pretty absent from the notebook market as well and that a solid APU is going to allow them to expand aggressively in that space as well.

The big question though is how long it will take to get there. GF abandoning their 7nm process means AMD has a much bigger limit on their supply of wafers and the server market is much more important. Similarly, any GPU is going to the professional market. When they’re constrained like that, it makes the most sense to go after the high margin markets first.

 

[french toast]

I would be looking for AMD to announce a partnership with Samsung before Christmas, right after they chop that WSA down to size.
As Samsung specialise in low power processes, plenty of capacity and also have leading edge tech, AMD most likely can negotiate a good deal to manufacture their APUs there on 7nm EUV.
Even a cheap 8nm LPP Picasso successor would be really good and would be a big step up from '12'nm glofo.

Binning that WSA agreement is a must though, after Picasso and pinnacle ridge they are useless to AMD.
Dual sourcing between TSMC and Samsung would be awesome for AMD, leading edge nodes with lots of capacity, probably good deals also as they can play them off against each other to get prices down.

Didn't realise TSMC has a HP version, that is really good news.

 

[NostaSeronx]

TSMC N7 HPC = TSMC N7 Mobile, in regards to transistors.

The difference comes from the standard cells;
Mobile default is 6-track and two fin.
HPC default is 7.5-track and three fin.

GlobalFoundries in comparison is a lot bigger:
SoC default is 6-track and two fin.
HPC default is 9-track and four fin.

Standard cells however will be built beyond those capabilities. TSMC N7 All(full package) has cells from 5-track/one fin to 12-track/six fin.

GlobalFoundries HPC => ~1.8x Frequency for ~2.75x Power.
Peak 4 GHz @ <100W TDP 14LPP
Peak 7.2 GHz @ <275W TDP 7LP HPC
=> The high TDP is from 9-track/4 fin and higher Vdd for more current. (GF HPC would imply designs with extremely low FO4 delay.)

TSMC HPC => ~1.55x Frequency for ~0.95x Power.
Peak 4 GHz @ <100W TDP 14LPP
Peak 6.2 GHz @ <95W TDP N7 HPC
=> The low TDP is from going 9-track 14LPP to 7.5 track N7 + slightly lower nominal Vdd.
<GlobalFoundries SoC is on-par with TSMC HPC.

GlobalFoundries at the time; cheaper, faster, and more simple than TSMC. Sad to see everyone jump ship because GF only had one FinFET foundry and TSMC had three+Nanjing(4).

 

[DrMrLordX]

Even with a 14+++ process I think Intel only really has an edge in the gaming segment.

The real problem Intel will have is feature size/die size. They keep improving 14nm performance, but they aren't improving density. That's the real trouble they'll have in server rooms worldwide: while AMD keeps expanding core count per socket, Intel has no plans for that.

It may not hurt them much in the short term with gamers, but in the long run it will hurt all up and down the product stack.

AMD APUs in 7nm are going to be a much better choice for most consumers.

Maybe, maybe not. I've seen arguments to the effect that AMD APUs were already a better choice for some non-zero number of consumers in the past. It never resulted in major sales. Even Raven Ridge hasn't made much headway in the low-end desktop space, and it has been out for awhile. AMD really isn't fighting hard for the low-end desktop/notebook segment the way they used to, so maybe that is part of it .. .

People also forget that AMD was pretty absent from the notebook market as well and that a solid APU is going to allow them to expand aggressively in that space as well.

OEMs are doing a lot of things that make APU-adoption in the notebook market difficult. None of that has changed that I can tell.

The big question though is how long it will take to get there. GF abandoning their 7nm process means AMD has a much bigger limit on their supply of wafers and the server market is much more important. Similarly, any GPU is going to the professional market. When they’re constrained like that, it makes the most sense to go after the high margin markets first.

They were going after the high-end markets even with GF bringing 7nm online.

 

[jpiniero]

Even with a 14+++ process I think Intel only really has an edge in the gaming segment. AMD APUs in 7nm are going to be a much better choice for most consumers.

Going to be awhile before you see 7 nm APUs. Intel might actually fix 10 nm by then. Maybe.

Also for some reason RR laptops seem to have bad idle, which hurts the battery life.

 

[positivedoppler]

Maybe, maybe not. I've seen arguments to the effect that AMD APUs were already a better choice for some non-zero number of consumers in the past. It never resulted in major sales. Even Raven Ridge hasn't made much headway in the low-end desktop space

I have a theory on that. Even though AMD APUs are generally better for iGPU gaming vs Intel, consumers overall purchasing decisions based on the potential of a system as oppose to it's current performance.

 

[maddie]

The real problem Intel will have is feature size/die size. They keep improving 14nm performance, but they aren't improving density. That's the real trouble they'll have in server rooms worldwide: while AMD keeps expanding core count per socket, Intel has no plans for that.

It may not hurt them much in the short term with gamers, but in the long run it will hurt all up and down the product stack.



Maybe, maybe not. I've seen arguments to the effect that AMD APUs were already a better choice for some non-zero number of consumers in the past. It never resulted in major sales. Even Raven Ridge hasn't made much headway in the low-end desktop space, and it has been out for awhile. AMD really isn't fighting hard for the low-end desktop/notebook segment the way they used to, so maybe that is part of it .. .



OEMs are doing a lot of things that make APU-adoption in the notebook market difficult. None of that has changed that I can tell.



They were going after the high-end markets even with GF bringing 7nm online.

Weren't we told that they reduced density for some of the latest frequency improvements? So going backwards. Might be a contributing factor to the scarcity of 14nm Fab space. Never could have imagined this perfect storm.at present. No 10nm, chipset migration, reduced density for latest 14nm+++ node. A true nightmare.

 

[PeterScott]

I would be looking for AMD to announce a partnership with Samsung before Christmas, right after they chop that WSA down to size.
As Samsung specialise in low power processes, plenty of capacity and also have leading edge tech, AMD most likely can negotiate a good deal to manufacture their APUs there on 7nm EUV.
Even a cheap 8nm LPP Picasso successor would be really good and would be a big step up from '12'nm glofo.

Binning that WSA agreement is a must though, after Picasso and pinnacle ridge they are useless to AMD.
Dual sourcing between TSMC and Samsung would be awesome for AMD, leading edge nodes with lots of capacity, probably good deals also as they can play them off against each other to get prices down.

Didn't realise TSMC has a HP version, that is really good news.

People really need to stop worrying about WSA. WSA is non issue going forward with future designs, which are all 7nm or better. AMD already announced they are all in on 7nm. GF hasn't even whispered about AMDs obligations and they won't.

WSA is completely irrelevant to whether those 7nm designs are at TSMC or Samsung.

Though AMD also announced they are all in with TSMC for now, and I don't see any issue with that at all. TSMC is the leading third party foundry and have been for years, and don't really expect that to change anytime soon. They are also ahead of EVERYONE at 7nm, so I don't get any real push to Samsung.

 

[french toast]

Going to be awhile before you see 7 nm APUs. Intel might actually fix 10 nm by then. Maybe.

Also for some reason RR laptops seem to have bad idle, which hurts the battery life.

I think infinity fabric has a role on power consumption, not sure about idle, but i have read ancouple of articles that make mention of the uncore on ryzen consuming much more than intel, which hurts idle.
AMD has much better igpu, lower power cores, but higher power uncore and the cores themselves can't turbo nearly as high.

People really need to stop worrying about WSA. WSA is non issue going forward with future designs, which are all 7nm or better. AMD already announced they are all in on 7nm. GF hasn't even whispered about AMDs obligations and they won't.

WSA is completely irrelevant to whether those 7nm designs are at TSMC or Samsung.

Though AMD also announced they are all in with TSMC for now, and I don't see any issue with that at all. TSMC is the leading third party foundry and have been for years, and don't really expect that to change anytime soon. They are also ahead of EVERYONE at 7nm, so I don't get any real push to Samsung.

I know they can fab elsewhere, but isn't there a monetary penalty to do so? Maybe glofo has relaxed this with their side falling down?
Anyway, AMD would be better off with..and presumably prefer to have two fabs to produce wafers off, they can haggle down prices, guarentee capacity, pick the best process that suits their target market, have a contingency plan should one can have a global foundries moment or Apple/Nvidia/Qualcomm pay to take capacity of leading edge nodes that may struggle with volume or yields.

The WSA agreement has cost hundreds of millions in amendments and forfeit costs, there is no value in it anymore, AMD should scrap it once and for all after 12nm.

 

[maddie]

I think infinity fabric has a role on power consumption, not sure about idle, but i have read ancouple of articles that make mention of the uncore on ryzen consuming much more than intel, which hurts idle.
AMD has much better igpu, lower power cores, but higher power uncore and the cores themselves can't turbo nearly as high.

I know they can fab elsewhere, but isn't there a monetary penalty to do so? Maybe glofo has relaxed this with their side falling down?
Anyway, AMD would be better off with..and presumably prefer to have two fabs to produce wafers off, they can haggle down prices, guarentee capacity, pick the best process that suits their target market, have a contingency plan should one can have a global foundries moment or Apple/Nvidia/Qualcomm pay to take capacity of leading edge nodes that may struggle with volume or yields.

The WSA agreement has cost hundreds of millions in amendments and forfeit costs, there is no value in it anymore, AMD should scrap it once and for all after 12nm.

I always saw the agreement as being forced on AMD, or at the least a necessary concession made during the initial sale, years ago. I really don't think they could force Mubadala to agree to anything. Using the courts would be irrelevant time wise even if successful.

Now, with the abandonment of 7nm finfet by GloFlo, AMD might have more bargaining power, but I still don't see them having the upper hand. Too many here assume that because their opinion sounds reasonable to them, it therefore must be correct. We have a few claiming the WSA is irrelevant, and that might be true or might not be true. A failure of imagination in my view, as we just don't know what is in the WSA agreement.

A question. Would a big institutional investor know what is in the agreement?

 

[PeterScott]

I know they can fab elsewhere, but isn't there a monetary penalty to do so? Maybe glofo has relaxed this with their side falling down?
Anyway, AMD would be better off with..and presumably prefer to have two fabs to produce wafers off, they can haggle down prices, guarentee capacity, pick the best process that suits their target market, have a contingency plan should one can have a global foundries moment or Apple/Nvidia/Qualcomm pay to take capacity of leading edge nodes that may struggle with volume or yields.

The WSA agreement has cost hundreds of millions in amendments and forfeit costs, there is no value in it anymore, AMD should scrap it once and for all after 12nm.

It's completely absurd that GF could impose a penalty for fabbing 7nm elsewhere, when GF 7nm is not an option.

From everything we have seen, AMD already had the concessions to do as much 7nm outside GF, that GF couldn't supply, long before GF announced the shutdown of 7nm.

Also bear in mind that GF is likely in breach of contract for shutting down promised 7nm production, which is yet another reason this was likely all settled quietly behind doors.

As far as two fabs being preferably. Look how often NVidia has dual fabbed. A negligible amount. It makes even less sense for a company with less resources since your teams essentially have to learn a new set of design rules for the other fab, and work through a new set of hurdles. Even more so when the second fab is running late. Maybe a couple of years down the road with both Fabs are running a refined 7nm+ with all the bugs worked out, they could send a couple of different designs to a second fab, but there is no point rushing that now.

There was never any value in WSA for AMD. it was the cost of selling the fab business. The thing is who cares where AMD Fabs old parts/chipsets going forward? It's essentially irrelevant to AMD. WSA is no longer a drag on leading edge parts, and that is what matters to AMD. Why pay to get out of an agreement that really doesn't hamper their main products anymore?

 

[french toast]

Who knows what the ramifications of all this is? Alot of presumptions are being made.
What we do know is AMD paid hundreds of millions just to amend the WSA agreement to be able to have this flexibility in the first place.
It is likely then they have already paid money to Globalfoundries to fab both Vega 20, Rome and Ryzen 3xxx, it has been a disaster financially.
I don't know, (neither do you) whether mubudala can force AMD to still produce 12nm (+derivatives) or FDX chips at glofo to meet the WSA agreement quota's for ever more.

I am not saying they can, but i think they have to force Mubudala to scrap the whole thing after 12nm...for FREE.
It would take a huge dose of good will on Mubudala part to void the whole thing, when in theory they could sit back and collect fees for AMD fabbing chips elsewhere, or even force AMD to keep designing chips on their crap processes.

We don't know the constraints of that contract nor do we know the attitude/will of Mubudala, who have lost billions in this adventure already and may not want to give up ANY potential revenue without a court battle, whether we see that as logical/unfair or whatever.

As for dual sourcing, Nvidia has a less diverse portfolio than AMD, also should AMD want to increase market share significantly, they will want to increase production significantly, AMD has also been burned by process disasters more than anyone, so there is that motive.
Your forgetting AMD has already starting designing chips at two different fabs, it is their PRESENT strategy.
Qualcomm, Apple, Nvidia, even Huawei..can pay the big bucks to guarentee volume they want, (lots) AMD will not necessarily be a priority customer amongst that lot...AMD needs guarentees.
Much better to have Samsung making some stuff.
Let's see what happens.

Edit; Nvidia has also used Samsung for their 14nm process whilst using TSMC simultaneously, so has Apple and Qualcomm swapped back and forth, AMD has swapped between TSMC and Glofo, even paying for the ' privilege' to do so, even Intel has used outside fabs on occasion.

 

[PeterScott]

Who knows what the ramifications of all this is? Alot of presumptions are being made.
What we do know is AMD paid hundreds of millions just to amend the WSA agreement to be able to have this flexibility in the first place.
It is likely then they have already paid money to Globalfoundries to fab both Vega 20, Rome and Ryzen 3xxx, it has been a disaster financially.
I don't know, (neither do you) whether mubudala can force AMD to still produce 12nm (+derivatives) or FDX chips at glofo to meet the WSA agreement quota's for ever more.

I am not saying they can, but i think they have to force Mubudala to scrap the whole thing after 12nm...for FREE.
It would take a huge dose of good will on Mubudala part to void the whole thing, when in theory they could sit back and collect fees for AMD fabbing chips elsewhere, or even force AMD to keep designing chips on their crap processes.

That just strikes me as completely out there nonsense.

On one hand you are worried, that GF can force AMD to keep producing 12nm parts it doesn't need, yet on the other you think AMD can force GF to scrap the WSA completely for FREE. You are just bouncing between two unlikely extremes.

As I already said, and you acknowledged, AMD already has the concessions to do all the 7nm fabbing the want elsewhere. This was public before GF threw in the towel on 7nm.

Beyond that, the sane middle ground is that AMD just keeps fabbing the >7nm parts it actually needs, at GF. That is the biggest win for both sides.

Your forgetting AMD has already starting designing chips at two different fabs, it is their PRESENT strategy.

Not by choice. That was the WSA forcing their hand. I am sure they will happily stick with one leading edge fab for years, now that they are free to go with the best fab.

 

[french toast]

That just strikes me as completely out there nonsense.

On one hand you are worried, that GF can force AMD to keep producing 12nm parts it doesn't need, yet on the other you think AMD can force GF to scrap the WSA completely for FREE. You are just bouncing between two unlikely extremes.

As I already said, and you acknowledged, AMD already has the concessions to do all the 7nm fabbing the want elsewhere. This was public before GF threw in the towel on 7nm.

Beyond that, the sane middle ground is that AMD just keeps fabbing the >7nm parts it actually needs, at GF. That is the biggest win for both sides.




Not by choice. That was the WSA forcing their hand. I am sure they will happily stick with one leading edge fab for years, now that they are free to go with the best fab.

I am not 'worried' by anything, it doesn't affect me.
What I am saying is; AMD will also be better off scraping this tie in if they can, if glofo is going to pull a move like this..whether financially necessary or not, AMD should use this as leverage to scrap the whole thing after 12nm, glofo serve no benefit to AMD after that, whilst I agree it is not been their choice up to now, Mubudala can't expect AMD to stick with them if they can't fulfill their end of the bargain or continue to produce leading edge nodes...in theory it breaks the contract.

What your saying is also nonsense, Glofo has already said they are stopping all leading edge manufacturing and research, this includes 7nm..so why are you throwing that non starter into the mix?..if there is a non get out production quota...AMD will have to use other available processes at glofo..ie rubbish.

You are also saying AMD should just continue on with the WSA agreement, and use its flexibility to keep fabbing elsewhere, like I said, I'm pretty sure there is a penalty to do that...millions of $ likely, also the production quota, we don't know the details of it or even if there is provisions for no leading edge nodes = AMD gets a pass, there might be, or AMD will either need to re negotiate the WSA agreement AGAIN ( to great cost ) or in my suggestion, push to scrap it altogether.

Let's hope Mubudala can show some good will and settle this stupid agreement, both parties can move in a more beneficial direction.

 

[french toast]

"The warrant will give Mubadala the option to buy up to 75 million AMD shares at a currently below-market price of $5.98/share, so long as they continue to own less than 20% of AMD. AMD is valuing the warrant at $235 million, which will bring the total one-time-costs of the latest agreement to $335M."

"AMD will be paying the foundry for this new flexibility. GlobalFoundries will be receiving a $100M payment from AMD, spread out over the next 4 quarters. Meanwhile starting in 2017, AMD will also have to pay GlobalFoundries for wafers they buy from third-party foundries. This in particular is a notable change from past agreements, as AMD has never paid a penalty before in this fashion. Ultimately this means AMD could end up paying GlobalFoundries two different types of penalties: one for making a chip at another fab, and a second penalty if AMD doesn’t make their wafer target for the year with GlobalFoundries."

https://www.anandtech.com/show/10631/amd-amends-globalfoundries-wafer-supply-agreement-through-2020

It's stupid to expect AMD to just carry on paying all this money through penalties, WSA amendments and payments to fab outside glofo.
Especially as AMD only just made a small profit since 2011;
https://www.fool.com/investing/2018/02/05/amd-turns-its-first-profit-since-2011.aspx

 

[PeterScott]

I am not 'worried' by anything, it doesn't affect me..

Yet you keeping dragging it up in multiple threads.

WSA has never mattered less than it does now. AMD is free at 7nm.

The time to complain was back when they were forced to produce leading edge parts at GF, that is no longer the case.

You seem to be operating under the assumption that AMD was utterly incompetent on their side of the contract, that they have no provisions for GF failing to deliver on it's end were made.

Anyone with a brain would have tied quotas to GF's ability to actually deliver leading edge process parts.

 

[maddie]

"The warrant will give Mubadala the option to buy up to 75 million AMD shares at a currently below-market price of $5.98/share, so long as they continue to own less than 20% of AMD. AMD is valuing the warrant at $235 million, which will bring the total one-time-costs of the latest agreement to $335M."

"AMD will be paying the foundry for this new flexibility. GlobalFoundries will be receiving a $100M payment from AMD, spread out over the next 4 quarters. Meanwhile starting in 2017, AMD will also have to pay GlobalFoundries for wafers they buy from third-party foundries. This in particular is a notable change from past agreements, as AMD has never paid a penalty before in this fashion. Ultimately this means AMD could end up paying GlobalFoundries two different types of penalties: one for making a chip at another fab, and a second penalty if AMD doesn’t make their wafer target for the year with GlobalFoundries."

https://www.anandtech.com/show/10631/amd-amends-globalfoundries-wafer-supply-agreement-through-2020

It's stupid to expect AMD to just carry on paying all this money through penalties, WSA amendments and payments to fab outside glofo.
Especially as AMD only just made a small profit since 2011;
https://www.fool.com/investing/2018/02/05/amd-turns-its-first-profit-since-2011.aspx

This really indicates the tiny negotiating power that AMD has vis-a-vis Mubadala concerning the WSA. It certainly wasn't a normal contract where the 2 parties have close to equivalent benefits and get to stake out redline positions. This was almost business rape and makes us aware of the dire straights that they were in at the time. To survive long term you first have to survive today, and then you get to have the "luxury" of worrying about tomorrow's ills.

 

[french toast]

Yet you keeping dragging it up in multiple threads.

WSA has never mattered less than it does now. AMD is free at 7nm.

The time to complain was back when they were forced to produce leading edge parts at GF, that is no longer the case.

You seem to be operating under the assumption that AMD was utterly incompetent on their side of the contract, that they have no provisions for GF failing to deliver on it's end were made.

Anyone with a brain would have tied quotas to GF's ability to actually deliver leading edge process parts.

No I didn't say anything about AMD being incompetent, you are making assumptions we know nothing about and mischaracterising my posts, what happened to your 7nm comments?..your making it up as you go along.

Like I said AMD pays millions to fab elsewhere, that alone is not sustainable on the long term, the flexibility was there to account for some delays in the process, not the whole thing cancelled.
You mention anyone with a brain (me? ) would know that if glofo couldn't provide leading edge nodes = AMD would have a contractual get out....I don't know about that, did AMD get out after the cancelled 20nm and 14nm XM?..the massive delays with 32 & 28nm?..would Mubadala play 'fair' if there was no clause? we don't know anything, so stop presuming contracts are written to your logic, they are not.
AMD have not got the power in this relationship, Mubadala have all the power and this is reflected in the lopsided and expensive WSA agreements, Maddie seems to understand this.

Making assumptions about contractual clauses with no info, being adamant that AMD should just Willy nilly just keep paying glofo to fab all leading edge wafers elsewhere is stupid, that is forgetting completely any production quota's.
Anyone with a 'brain' can see AMD has no future with Globalfoundries if they are quitting leading edge processes, why should they continue to hand over money to Mubadala IF there is anyway out?.
If what you are half suggesting is true and magically Mubadala asks for no costs for dropping their end of the bargain, why keep the the WSA anyway??..that doesn't make any sense.

 

[moinmoin]

AMD's strategy to go all in with 7nm was formed years ago already. This put pressure on GloFo to (also) deliver with 7nm. This is very likely why WSA was extended with the 3rd party foundry wafer penalty, to give GloFo a guarantee that once GloFo actually offers 7nm AMD has a strong interest to make good use of GloFo's node instead one of the competition. That situation will never happen now, so there is no basis for a 3rd party foundry wafer penalty on 7nm anymore. As PeterScott rightly wrote, due to GloFo giving up on 7nm AMD is now free of the WSA shackles wrt all 7nm and beyond chips.

 

[french toast]

AMD's strategy to go all in with 7nm was formed years ago already. This put pressure on GloFo to (also) deliver with 7nm. This is very likely why WSA was extended with the 3rd party foundry wafer penalty, to give GloFo a guarantee that once GloFo actually offers 7nm AMD has a strong interest to make good use of GloFo's node instead one of the competition. That situation will never happen now, so there is no basis for a 3rd party foundry wafer penalty on 7nm anymore. As PeterScott rightly wrote, due to GloFo giving up on 7nm AMD is now free of the WSA shackles wrt all 7nm and beyond chips.

Right, so that means the WSA is useless, which goes nicely along with me saying 'scrap it'.
If the WSA is not scrapped(or made impotant until ran out) then that must mean Globalfoundries are making money out of AMD (if renewed)...contrary to peterscotts arguement.
As Mubadala own ~20% of AMD, hopefully they see sense and do one of the former.
*Edited.

 

[PeterScott]

You mention anyone with a brain (me? ) would know that if glofo couldn't provide leading edge nodes = AMD would have a contractual get out...

No one with a brain would negotiate a contract that lets their supplier fine them, for buying elsewhere, what that supplier can't deliver.

That actually gives a supplier incentive not to deliver.

They don't have teams of lawyers working on contracts to miss such basic items.

Right, so that means the WSA is useless, which goes nicely along with me saying 'scrap it'.

A WSA that maintains >10nm business at GF, has obvious value to GF. It ensures a predictable customer for the foreseeable future for all of AMDs considerable non-leading edge business.

AMD has little leverage to end WSA with regards to that business and little incentive either. This is non critical, lower cost work, and AMD would have to pay to get out of it for negligible benefit.

The sane course of action for both parties, is AMD is free to do what it want on 7nm and better, while maintaining non-leading edge business with GF. Quota will be adjusted downward to reflect GF inability to deliver on leading edge.

WSA remains, but become a non issue because it's only for non critical work going forward.

This is most likely, sane and sensible route for both parties.

Ranting about less likely what ifs, really serves no purpose.