Report: AMD cancels 28nm APUs from GlobalFoundries

Jos

Posts: 3,073   +97
Staff

It seems more troubles await AMD in the microprocessor market. A few days ago rumors started to emerge that the company had cancelled its Wichita and Krishna APUs, the 28nm the successors to its C-Series (codenamed Ontario) and E-Series (codenamed Zacate) chips for budget and mainstream notebooks. Now, a report from ExtremeTech is backing those rumors with an interesting twist on the side: AMD will move 28nm production from GlobalFroundies to TSMC and start afresh with a new design.

The decision comes in light of a slow, low-yield production ramp. A switch of foundry would be quite significant, both from technical and strategic standpoints. As ExtremeTech notes, TSMC's 28-nm high-k metal gate process has a gate-last design, but GlobalFoundries' is gate-first.

AMD may try to buy some time by releasing more efficient 28nm variants of Ontario/Zacate products, but a true replacement that brings the architectural improvements originally intended for Krishna/Wichita is likely at least 18 months away.

While AMD holds close to a 10% share of GlobalFoundries, the relationship between the two seems to have turned sour as of late. When issuing its revenue warning for the last quarter, AMD specifically blamed 32nm manufacturing issues at GlobalFoundries for the lower-than-expected revenue. They announced an agreement under which they would only pay for good 32nm dies, but with GlobalFoundries still having trouble to improve yields the firm has been stuck building Llano at a loss for most of this past year.

That deal expires on January 1, at which point AMD will go back to paying a flat fee per wafer. Apparently, GlobalFoundries is unwilling to negotiate a new wafer agreement, just as much as AMD is unwilling to pay a per-wafer fee for a slow, low-yield product ramp -- thus the breakup.

Despite setbacks in production, AMD has said it remains committed to its partnership with GlobalFoundries and that the two will continue to collaborate on AMD's main product lines. It doesn't seem like they have much options either. The company needs a competitive product on the market to fight not only Intel but also ARM, which will continue to gain territory next year as Windows 8 introduces support for the architecture.

Permalink to story.

 
I think this is going to make for a very interesting 2012/3 personally.

ARM have really catapulted themselves forwards and the issues AMD has experienced with its chips this year is probably been of huge benefit to them.

Okay, Intel will still rule the roost at the top end of the segment, but I'm looking forward to the lower end of the segment being fought out between the three companies.
 
:( :( I see a very dark future or none at all for AMD. This is very bad news indeed. I really wish they could just make a breakthrough to catapult them into atleast a subtle lead over Intels tech. But the nm race is what is going to decide chip makers of tomorrow and AMD is falling swiftly behind
 
On the other hand, this could be great news for AMD. Maybe this is their stab at ramping up CPU production and producing technically better chips?
 
On the other hand, this could be great news for AMD. Maybe this is their stab at ramping up CPU production and producing technically better chips?

Yep, Having to scrap a near-ready-for-release range of parts because the foundry you are part owner of can't provide the necessary yields at the desired specification in the appointed timeframe...resulting in lost orders, a lack of credibility in the industry, a wasted opportunity to make ground on your principle competitor, and a need to pour more expenditure into R&D to redesign from gate-first to gate-last.

If that isn't the very definition of "great news" I don't know what is.


Oh, and welcome to TechSpot.
 
Witchita and Krishna were always supposed to be manufactured at TSMC check these older links

http://www.brightsideofnews.com/news/2011/8/16/amd-to-kill-apu-in-20123b-move-to-soc-with-28nm-krishna-and-wichita.aspx

http://pcper.com/news/General-Tech/TSMC-gets-AMDs-28nm-APU-business
 
Witchita and Krishna were always supposed to be manufactured at TSMC check these older links

As far as I'm aware, Wichita/Krishna were originally set for GloFo's 28nm SHP process. It was generally reported that AMD switched from GloFo to TSMC around the same time that AMD/GloFo's contract changed from per wafer to per good die. You could probably assume that AMD were hedging their bets should GloFo's 28nm node be as problematic as 32nm.

From EE Times yesterday (quoting AMD spokesperson):

Bearing in mind that Wichita/Krishna are no more, that essentially means that Kaveri/Samara/Kabini are now the next low voltage parts at AMD...and of course, since the roadmap suddenly appeared after GloFo's manufacturing woes surfaced, there is no process/foundry attached to the parts.
Things are probably complicated by the fact that Bobcat's design team no longer work for AMD - the chief architect (Brad Burgess) quit and is now at Samsung- which probably necessitates a reworking and re-evaluation of the future parts with new personnel.
 
Things are probably complicated by the fact that Bobcat's design team no longer work for AMD - the chief architect (Brad Burgess) quit and is now at Samsung- which probably necessitates a reworking and re-evaluation of the future parts with new personnel.

You said a mouthful. I suspect that any road-map released to date is now irrelevant, and I would be very surprised if Steamroller for example, will ever come to fruition, or is even being planned for.
 
Blank slate...fresh start....who knows?
At this stage I think that either the AMD board have a grand plan which is beyond the grasp of mere mortals, or between mismanagement, overly optimistic forecasts, design miscues and some bad luck, AMD have been thoroughly derailed.

I can see how AMD are looking to a paradigm shift in design focus...unfortunately I can also see that the new direction has come piecemeal and with a lack of coherence in gathering the required resources to effect the change.
Obviously the new strategy is aimed at low voltage parts yet AMD gambled and placed all their eggs in GloFo's basket (with the exception of high-end GPU), and unlike AMD, GloFo isn't about competing with Intel, it's about competing with TSMC, so they squeeze one last shrink out of gate-first optimized for die-size/lower power -and the majority of their customers whose products use 28nm-HPP or SLP in the wireless, mobile, network businesses. I'd assume this places AMD between a rock (GF tweaked process shrink/new arch) and a hard place (TSMC's problems with 32nm, no guarantee that 28nm wont meet with problems)

I'd be happier with AMD's long term outlook if there were some substantial statements on future policy...some name-brand engineering/design hiring for example, but at this point in time all we have is Rory's word that there will be some kind of announcement in February.
Low-end and mainstream APU's probably represent AMD's best chance of turning things around, so I can understand the logic behind concentrating resources there...what I can't understand is why AMD have allowed GF's process problems to dictate AMD's fortunes to the extent they have. AMD as part owners, if not as a customer, must have been aware for GF's problems with 32/28nm for at least a year, yet leave any potential easement for 8-9 months.

Hard enough going up against Intel with a full array of SKU's in the market, but when you're hamstrung by not getting the better products into the channel (Brazos, Llano) you're basically cutting your own throat.

The personnel losses...well, where do you begin. Cutting out the deadwood is part and parcel of any company...but what precisely is AMD doing behind closed doors that causes valued personnel to jump ship. Headhunting and looking for greener pastures are a way of life, but the rate that AMD are losing key people certainly would be higher than the norm. The employee reviews at Glassdoor don't make particularly good reading from a company perspective.

Anyhow, Happy Thanksgiving.......I see the Lions are undergoing their usual Thanksgiving mauling (down by 16 to the Pack 10 minutes into the final quarter...so at least one American institution remains unaffected by change going into 2012.
 
I have a sneaking suspicion that right after the first of the year we will see the engineering hiring in droves and we will know where they are heading. i keep going back to the position Intel is in if AMD moves to a product line that hands the desktop over to them part and parcel. I wonder if things are to a point product wise that should AMD abandon the desktop market, it would not be considered monopolistic? I mean it would be terrible from a consumer standpoint...but legally?
 
Legally. I'm not so sure how Intel can be held liable for having a monopoly of the market when its main competitor has backed out voluntarily?

You certainly couldn't blame them for that decision -- it rests solely on the ****** responsible for the host of poor decisions that led to this happening in the first place.
 
Leeky said:
Legally. I'm not so sure how Intel can be held liable for having a monopoly of the market when its main competitor has backed out voluntarily?

You certainly couldn't blame them for that decision -- it rests solely on the ****** responsible for the host of poor decisions that led to this happening in the first place.

You would be surprised. while it's not their doing or fault, there are many that will put Intel under the microscope and jump on, and exploit anything that can remotely be parsed as exploiting a position of monopoly. As litigious as things are, i don't think it's a position that Intel would even favour
Think about it...microsoft got sued for including thier own browser in their OS because it was deemed non competitive.
If AMD bows out of the market..Intel might as well have Rambus move in to the corporate office so they can role play on a regular basis.
 
Yeah, I see your point Red.

But with ARM making an appearance now (well soon) I don't think it would be too much of a problem. I couldn't see AMD not existing in the CPU market, but they are going to struggle for the next couple of generations.

The next generation of processors is going to make or break AMD. I'm not personally sure how much more of a hammering it can take. It's beginning to look like the Black Pearl riddled with cannon holes and sinking into the abyss.
 
Yes, Lee. That is my understanding of the situation also re molopoly's

If Intel either forced AMD out of business using illegal business practices, or tried to acquire AMD then that becomes grounds for the FTC/SEC, but if AMD voluntarily leave the marketplace or are forced to due to their own misfortune/change of direction/consumer preference then I don't think that it becomes a monopolies commission issue.

The rise of 64-bit ARM might mitigate the monopoly argument in any case, although it does give me cause to wonder how the x86 market would be seen with regards the Intel/AMD x86 cross-licence (VIA being pretty much a non-entity) if said licence is not transferrable in any potential AMD buy out. I'm assuming that the licence is voided as an (at least 4+ years) x86-64 licence + AMD's IP would exceed AMD's market cap and company value (esp. if you add in the GPU IP).
You would be surprised. while it's not their doing or fault, there are many that will put Intel under the microscope and jump on, and exploit anything that can remotely be parsed as exploiting a position of monopoly.As litigious as things are, i don't think it's a position that Intel would even favour
Intel have been walking a pretty straight line since the EU, FTC, NY and Japanese commissions. AMD are also at a point where they can quite happily dig their own grave. Having said that, I think you're right in Intel being desirous of keeping the status quo (i.e. a small secondary competitior in x86) with the opposition limited in R&D and marketshare. Imagine what would happen if a major player like Apple decided it wanted to use in-house designed CPU/APU/GPU and could grab AMD lock stock and barrel for what amounts to small change. An unlikely scenario but not one that would benefit Intel in any way, shape or form.
Think about it...microsoft got sued for including thier own browser in their OS because it was deemed non competitive.
I think that comes down to people choosing their OS but not having the choice to use their own preferred software. IE is automatically installed with Windows in the usual install method, and if you weren't forward thinking enough to have another browser downloaded ready to go, you would have to use IE in order to obtain its replacement.
I have a sneaking suspicion that right after the first of the year we will see the engineering hiring in droves standpoint.
I'll need convincing of that. I hope it's true for every tech users sake, but I don't think that the world is awash with experienced semi-conductor engineers and chip designers- and even if it were you still have to 1. offer them better terms than where they are at present, and 2. they would be restricted from using some of their knowledge base due to non-dislosure/non-compete/proprietary information clauses in employee contracts...so basically you're starting from scratch the day they are hired. Could result in a significant amount of time between proposal and retail deployment.
 
When issuing its revenue warning for the last quarter, AMD specifically blamed 32nm manufacturing issues at GlobalFoundries for the lower-than-expected revenue. They announced an agreement under which they would only pay for good 32nm dies, but with GlobalFoundries still having trouble to improve yields the firm has been stuck building Llano at a loss for most of this past year.
Gosh, I'm so glad I'm an Intel person. With Intel you can take 32nm pretty much for granted....:rolleyes:
 
If Intel either forced AMD out of business using illegal business practices, or tried to acquire AMD then that becomes grounds for the FTC/SEC, but if AMD voluntarily leave the marketplace or are forced to due to their own misfortune/change of direction/consumer preference then I don't think that it becomes a monopolies commission issue.

I view it the same. At the end of the day AMD's misfortune has nothing to do with Intel. The blame rests squarely on them, and them alone.

It is totally unreasonable to expect Intel to be in any way to blame. It certainly couldn't be argued it was their fault for having a better product. I'd be saying the same thing if the fortunes were reversed.

AMD should consider themselves at least fortunate for having the GPU income they do as well as something thats competitive with Nvidia for now.

The rise of 64-bit ARM might mitigate the monopoly argument in any case, although it does give me cause to wonder how the x86 market would be seen with regards the Intel/AMD x86 cross-licence (VIA being pretty much a non-entity) if said licence is not transferrable in any potential AMD buy out. I'm assuming that the licence is voided as an (at least 4+ years) x86-64 licence + AMD's IP would exceed AMD's market cap and company value (esp. if you add in the GPU IP).

I have high hopes for ARM. Okay it remains to be seen if it could compete anywhere near the enthusiast price-point of Intel, but in the budget and general consumer market the higher efficiency processors are likely in my opinion to make considerable head waves.

I just hope they're overall better equipped than AMD have shown themselves to be, because they've totally screwed up a perfectly acceptable portfolio of CPU products and pretty much left themselves scraping the barrels in the hopes of a few good scraps -- cue the insanely crap ratio of "good" yields at Global Foundries. Double cue the whole BD mess -- and its a sodding landfill of one at that.

I won't comment on the whole x64 license thing between Intel and AMD, as know little of it.

I think that comes down to people choosing their OS but not having the choice to use their own preferred software. IE is automatically installed with Windows in the usual install method, and if you weren't forward thinking enough to have another browser downloaded ready to go, you would have to use IE in order to obtain its replacement.

I totally agree. The situation was totally different. In essence Microsoft "forced" you to use their browser, at the very least in order to download and then install another one to replace it. It should have been a optional extra from day one.

I didn't agree with adding the capability to select another browser after getting updates as acceptable either. The user should have the immediate choice, not have to download an update to then choose if they want to use something other than IE.

How that was even agreed by the courts is beyond me, as most uninformed users likely just cancelled the box anyway, not knowing any better.

I'll need convincing of that. I hope it's true for every tech users sake, but I don't think that the world is awash with experienced semi-conductor engineers and chip designers- and even if it were you still have to 1. offer them better terms than where they are at present, and 2. they would be restricted from using some of their knowledge base due to non-dislosure/non-compete/proprietary information clauses in employee contracts...so basically you're starting from scratch the day they are hired. Could result in a significant amount of time between proposal and retail deployment.

I personally think they're screwed. At best the influx of people (if you want to call it that) will arrive and they'll immediately be thrown to the drawing board. It is unlikely they'll be able to continue from where the designs were left by those that worked on them before. So your talking yet another generation lost while they get there head in the game, and use R&D results from research into chip design that is already several years behind Intel. Of course, this yet again all depends on whether they can get a half decent return in production. If I'm correct in thinking the following is true, then we're already half starved of the majority of the companies best processor products as it is. But it doesn't matter how good something is if you can't get enough of it to the market to actually make any money on your investment.

You lose the potential goldmine of earnings then R&D into future products gets starved. From there on its a vicious circle almost impossible to exit from without a monumental cash injection, God-like creative thinking, unimaginable good fortune and every star in the universe aligning to your beckon call. It is unreasonable to suggest AMD can plow untold billions into future designs when the current processors are struggling to keep with demand, and therefore make money.

It would likely take Intel to totally lose the plot and launch something that completely fails in the market, or a bad design unnoticed until launch in order for AMD to stand any chance of catching up -- and thats assuming AMD actually got things in order in the first place.

I'm sorry, but Intel are too refined at this game now. They're confident enough to show you ahead of time how good something will be. They never disappoint either, with results in the wild (as far as I recall) always on par with the slides leaked.

I just don't see Intel slipping up, and I certainly don't see them slowing down now.
 
I totally agree. The situation was totally different. In essence Microsoft "forced" you to use their browser, at the very least in order to download and then install another one to replace it. It should have been a optional extra from day one.

Huh? it's their OS. why wouldn't they use their browser as default their browser?? Thats like saying you are forced to use MS OS, by having to ...install windows. I have never understood this mind set. It's like saying that upon buying a Vuaxhall, you are forced to use a Vuaxhall steering wheel...or at least use it to get to the parts store to get an after-market wheel. If we are going to complain that MS doesn't pedal other companies wares, then why not complain about paint or media player not having Corel or AVS as an option?

I view it the same. At the end of the day AMD's misfortune has nothing to do with Intel. The blame rests squarely on them, and them alone.

true, but not relevant when it comes to how the legal system works. you gave a good example yourself..



It is totally unreasonable to expect Intel to be in any way to blame. It certainly couldn't be argued it was their fault for having a better product. I'd be saying the same thing if the fortunes were reversed

"reasonable" has nothing to do with it, and that is my point,
I think you guys are dismissing the legal system on this one. weather or not intel's doing, or they have anything to do with it, the system does not like monopolies regardless of how they came into being..

The rise of 64-bit ARM might mitigate the monopoly argument in any case, although it does give me cause to wonder how the x86 market would be seen with regards the Intel/AMD x86 cross-licence (VIA being pretty much a non-entity) if said licence is not transferrable in any potential AMD buy out. I'm assuming that the licence is voided as an (at least 4+ years) x86-64 licence + AMD's IP would exceed AMD's market cap and company value (esp. if you add in the GPU IP).

I agree and mentioned this earlier. it may be that because of this that AMD tanking may not even rise to the level of monopoly anyway.
 
Huh? it's their OS. why wouldn't they put as default their browser?? Thats like saying you are forced to use MS OS, by having to ...install windows.

Which is also valid, to a small degree anyway. There aren't many places I can purchase a laptop that doesn't come with Windows installed. Its not exactly easy getting a refund if I don't want Windows either. I appreciate its a minority, but its truth none-the-less.

I see where your coming from though, and looking at it from that angle I can see your point.

True, but not relevant when it comes to how the legal system works. you gave a good example yourself..

"reasonable" has nothing to do with it, and that is my point,
I think you guys are dismissing the legal system on this one. weather or not intel's doing, or they have anything to do with it, the system does not like monopolies regardless of how they came into being..

I know law is complex, but I still don't see how the situation could have been avoided on the part of Intel. Its not like industrial espionage (least I don't think so!) played a part in their downfall.

I just don't see it being morally right, legal or ethical to force a company to break apart because of the actions of its main competitor.

Though I clearly don't claim to understand commercial or business law either.

P.S. If I did, I'd be applying for a lawyers job at Apple -- who by the way probably have the best job security of any legal department in the world! :haha:
 
It is totally unreasonable to expect Intel to be in any way to blame.
Well, they certainly aren't in danger of selling at below cost.
AMD should consider themselves at least fortunate for having the GPU income they do as well as something thats competitive with Nvidia for now.
That is something, although Sandy Bridge/Ivy Bridge and AMD's own A6/A8 series are surely eating into the bottom end. The midrange market is AMD's to lose...unfortunately average selling prices and a high BoM are what's keeping AMD graphics from posting significant profit. The high end /pro cards are more favourable to Nvidia ( higher ASP, larger market presence, similar BoM), so 28nm GCN will be fairly crucial to AMD's plans -both in relation performance, and their commitment to OpenCL.
I have high hopes for ARM. Okay it remains to be seen if it could compete anywhere near the enthusiast price-point of Intel, but in the budget and general consumer market the higher efficiency processors are likely in my opinion to make considerable head waves
The beauty of ARM is that it's driven by the licencees. Qualcomm, TI and Samsung are all major players with significant experience. Add in Nvidia, Intel/Marvell, STMicro etc. and you have a very healthy competitive enviroment
I personally think they're screwed.
Pretty much I would think. One of two things happened...
1. AMD are waiting to be carved up and reshaped. The Abu Dhabi royal family/Mubadala/ATIC already own GloFo..not inconceivable that they would want to re-unite AMD and their ex-fabs.
2. The AMD BoD are incompetant. No CEO works in a vacuum and has carte blanche on the running of a company, so I don't buy laying all of AMD's woes at the doorstep of Dirk Meyer. The BoD either misread the projections of themselves, their foundry partner and Intel's ability to keep upping the performance bar, or they read it correctly and disregarded the findings. Option three being that the BoD haven't heard of strategic business planning
At best the influx of people (if you want to call it that) will arrive and they'll immediately be thrown to the drawing board. It is unlikely they'll be able to continue from where the designs were left by those that worked on them before.
Maybe they are relying on promotion from within. In which case why not award promotion on merit when it presents itself rather than a blanket approach which see's everyone promoted at the same time. If AMD is the kind of company where you need a systematic purge to change the old order then it doesn't reflect well on the BoD/Politburo
It would likely take Intel to totally lose the plot and launch something that completely fails in the market
Hence the AMD-centric forums jumping on any perceived Intel misstep such as Ivy Bridge's intro date slippage...Selling more Sandy Bridge CPU's I don't think is greatly hurting Intel's position.
...or a bad design unnoticed until launch in order for AMD to stand any chance of catching up -- and thats assuming AMD actually got things in order in the first place.
Intel actually increased marketshare in Q1/Q2 when they were embroiled in the SATA controller recall/B3 replacement...not a good advertisement for AMD's product stack.
I'm sorry, but Intel are too refined at this game now.
Yup. Like the old adage "Money makes money". Intel's R&D probably exceeds AMD's entire expenditure. More to the point they have enough breathing space to take their time getting the next big thing right...and of course if they decide to slow their process tick-tock cadence then every CPU release has a greater length of time in the marketplace -and of course the attendant ROI increases.
Intel could conceiveably (I don't think they will...at least for SB) keep Sandy Bridge in the market for another year. The process is bought and paid for, and releasing a 2800K/2900K at 3.8 - 4.0GHz while dropping prices on lower SKU's would likely be enough to keep the majority of people happy

AMD's problem would not just amount to getting the products out (and in a timely fashion), but overcoming market inertia. AMD's image took a fair old beating over Llano/Brazos/Zambesi/Interlagos lateness, lack of availability, and to a degree, performance. While your average desktop builders might be a fickle bunch easily swayed by rhetoric, brand allegience and shiny packaging, OEM's -especially those in the enterprise sector- tend to place a very high premium on time to market, availability and promised performance, and given that 1.OEM's tend to stick with a supplier until the supplier screws up, and 2. OEM's are where the money is at, it's difficult to see how AMD will regain any portion of the market.
 
I know law is complex, but I still don't see how the situation could have been avoided on the part of Intel. Its not like industrial espionage (least I don't think so!) played a part in their downfall.

I just don't see it being morally right, legal or ethical to force a company to break apart because of the actions of its main competitor.

Oh I agree completely! it should be used as motivation and opportunity for someone to step up and offer competition, not a penalty for the successful corporation.

Though I clearly don't claim to understand commercial or business law either.

I don't claim to either, just enough to know what happens from watching what has happened to those that have been deemed a monopoly in the past, and that definition is suspect as well :)

@ Chef,
I will predict that in addition to hiring people that the tech community has not heard of in the engineering dept, they will also recruit a big name away from someone ...(and probably have to promise the moon and a bunch of control) and that will be the indicator of where they are going...as opposed to any slide show they are going to offer.
...and for the record, things started to go south when they canned Mario Rivas. :)
 
I will predict that in addition to hiring people that the tech community has not heard of in the engineering dept, they will also recruit a big name away from someone ...(and probably have to promise the moon and a bunch of control) and that will be the indicator of where they are going...as opposed to any slide show they are going to offer.
Well, I hope you're right. Someone in Sunnyvale needs to grow a pair and move away from powerpoint presentations and into a valid strategy.
Even if they get the brains, they still need the brawn (R&D cash) and most importantly some people that set their sights on what the competition will be doing as well as a solid Plan B

.
..and for the record, things started to go south when they canned Mario Rivas. :)
That's another way of saying "when Randy Allen took over". Randy would make Larry, Moe and Curly seem competant
 
I can see why Global Foundries would be sad....is that why you called them "Global Froundries"??

:)
 
AMD just needs to be bought out by IBM.. and get it over with.

Perfect Fabs & engineers.
 
Back