The headline needs to be corrected -- Did you mean 7% of the company's workforce is to be cut instead of 7K jobs? From the source: "Reduce global headcount by 7 percent, largely expected to be completed by the end of Q4 2014"
Consoles are helping a lot, but their computing division is still far too weak. Man, I feel like AMD has been on life support for nearly a decade and it's getting close to time to pull the plug.
Yeah, but then, without the graphics division, AMD will be left with nothing good to sell.
And then intel will squander ATI, because they lack any competence in writing some proper graphics drivers that will work and harness the full potential of the GPU.
Sale of fabs to Global foundries put them at a 2 gen lag versus Intel, but their Buldozer architecture is the nail in the coffin. I'm sure they saw their mistake 3 years ago but it takes time to develop a new architecture, so for the time being they make what they can with their dual module crap.
Feh! They've always been behind Intel in process technology. I remember when they were a year behind, and Ruiz said they were working to get to six months. Instead, they fell to eighteen months. This is nothing new, it's just continuing the path set earlier.
What was amazing is the time (over a decade ago) when AMD was still two generations behind Intel's manufacturing technology, but still producing technologically superior, higher performance, and less expensive chips.
AMD did have some manufacturing breakthroughs, though. Copper interconnects and integrated memory controller...and of course 64-bit x86 extensions.
You guys have strange memories. Around a decade ago AMD was first to 90nm, and SOI was superior. How were they 2 generations behind Intel in manufacturing tech? In addition they also had the superior architecture... And it was all down hill from there.
The problem was that they were almost too successful for their own good. They awakened a slumbering giant. In the short term Intel held them at bay via questionable business practices, back when AMD had the technological upper hand. Meanwhile Intel buckled down and poured money into their fabs as well as R&D, to catch up and pass AMD. Intel has done a pretty good job of keeping their guard up ever since.
AMD has done pretty well overall in the graphics market and their APUs are competitive for the money. I have hope that Keller's team will create something interesting for 2016.
You really are desperate to avoid blaming AMD. This had nothing to do with intel and everything to do with AMD buckling down on the utter failure of an architecture that is bulldozer.
A lot of AMD's "firsts" actually come from the DEC Alpha line of processors. The IMC first appeared on the DEC Alpha 21064/6, introduced way back in 1992. Ditto with the double-pumped FSB that gave Athlons their edge over the Pentium II/III.
During those early days, Intel was just catching up to RISC with the Pentium. Jerry Sanders gave AMD an excellent jump start by bringing people from DEC (Dirk Meyer!) on board and incorporating a lot of technology that Intel was just starting to develop on its own.
The Bulldozer architecture is incredibly good, I can't see a single reason that would prevent me from recommending it to even demanding desktop users.
Price? Great. Performance? Better than anyone needs.
But the market has spoken, the new FX line has been ridiculed by idiot nerds and their tag-alongs (who don't know anything anyway)
And AMD has blinked. The so called APU architecture is the future, but when Bulldozer was introduced, APUs were not ready, but in 2015 they probably will be.
In reality I don't know what AMD is planning, but I suspect that getting rid of Rory Read was a good move, he was getting the company nowhere.
If you were to look at "good enough computing", perhaps, but the competition has moved on. There are strong points to the architecture, and I maintain my belief that an octo-core Steamroller would be a very strong multithreader, but they'll only really be reaching performance parity with Sandy Bridge over the next year or so, some four years too late.
Unless they have an ace up their sleeve, I don't think their graphics division will be profitable much longer either. The channel has indicated that since Maxwell, AMD's 290 and 290X sales have almost completely stalled. AMD taking a month to act on a price-cut didn't help, either.
I hope they survive. NVidia AND Intel need competition, even if its weak competition...because it was the lack of competition that brought us the Pentium 4's Netburst catastrophe and NVidia's initially weak, overprice Fermi cards...the last time ATI was really dominating graphics with the 4000/5000 series.
AMD were providing plenty of competition when the P4 came out, and the 5000 series came out before Fermi meaning ATi were on top at the time. No, definitely not a lack of competition; the more recent CPU launches by Intel are more indicative.
Their graphics division has been keeping them afloat. Without ATI they'd really have nothing. No high end GPUs that still pull in money, no APUs which are their biggest CPU sales. They'd just have crappy CPU cores.
Without the ATI purchase, AMD would have more cash to keep up with Intel on the CPU side and could have had both ATI and nVidia competing to provide integrated graphics, even on the same chip as SoC vendors do. Particularly as Intel started pushing their integrated graphics aggressively AMD might have ended up holding two good cards in a three-way race by not having their own solution and offering a platform for both ATI and nVidia to launch at Intel, potentially also boosting their fab volume. By committing to ATI, they naturally invited Intel and nVidia to join forces and crush both sides of the company. Still it was 2006 and many things looked different back then and a whole lot could have happened since, it's pretty much all speculation.
They are keeping the lights on and 93% of its workforce employed. They have things they can do. They just don't. There are better processes out there. No reason they need to be stuck on 28 nm. They would do better resurrecting their previous architecture and fabbing at these smaller geometries than bulldozer.
Samsung also is using 20nm on their SoCs. In particular Exynos 5430 and Exynos 5433 (now called Exynos 7 octa) are build in 20nm. In fact they were doing it prior to Apple.
What? AMD has no market power, along with no money. Apple bought 30 million x86 chips last year, and 250 million ARM chips of their own design. How can AMD compare to that?
paying off debt... I believe AMD is trying to put as much money as possible into paying off debt before profits, or, sometimes when it puts them into losses/worse losses.
AMD isn't "stuck" lower process nodes still aren't ready for parts that aren't super-high yield or super-high-margin iirc.
So, you're saying that 22nm isn't ready? Aren't you a bit behind the times? And 20nm is obviously ready.
AMD's technology was always poor. I remember going back decades, that it was poor. They often announced chips that were leading edge, but couldn't produce them until Intel surpassed them, and AMD had to sell them at a discount when they first appeared, later on.
Cost/yield is part of the reason they didn't jump on these new processes right away. The other is that you have to design/redesign to target a new process, so when you're building a chip you have to choose whether to gamble on a new process. Sometimes you gamble and you lose, the process isn't ready for primetime and you suffer.
It's also easier to start working with smaller geometries when you're dealing with tiny SoCs. The process will have to mature a bit before AMD adapts its much larger chips. That being said they should bring their smaller cat-core SoCs over to smaller processes ASAP.
you don't need to be competitive in the high end to make money. Brazos is what kept AMD's lights on with Bulldozer and everything else going on. Would Bulldozer having matched/exceeded the hype have been better than Brazos was? Maybe, but, maybe not.
AMD's almost anytime you see an x86 AMD semi-custom or embedded win from AMD, that is only possible because of their CPU division being the only game in town besides Intel with an x86 licence (well, there is Via...)
I did not realize Intel was selling atom's in the embedded and semi-custom markets?
Markets where you often need a specific piece of specifications, if the GPU is not up to snuff in an atom, it will never be bought there. If AMD can be modified for a lower cost to add something the company needs, it will be done.
Not sure why you think Atom is targeting Embedded and semi-custom, which is where AMD will likely end up making its money on x86.
Atom is also to slow to target the lower end APU market. The other place wordwide where AMD is likely to make quite a bit.
Given I believe AMD is one of the movers in ARM server chips that has the most experience in servers and working alongside companies that make them, I would give them pretty good chances to take a large chunk of that profitable market. Atom won't do anything there. Core architecture won't do anything their either.
Intel is also now in the semi custom game to a degree (though they hate to admit it). They carefully ensure that their bed and butter is still bringing in the dough though.
<quote> He added Intel's new Grantley Xeon CPUs are seeing strong uptake, and that the chip giant now has 35 custom Xeon SKUs for various clients. Facebook, eBay, Microsoft, and other Web giants have been avid buyers of custom Intel server CPUs. <end quote>
Intel wants Atom variants to fight the ARMy not AMD. But basically due to contra bribes, acceptable performance and better power characteristics, it basically has killed Mullins and any other Tablet SOC that AMD can field in that area. Further they have specific low power Core i SKUs for real performance + acceptable power. So AMD is sandwiched between a
a) Lower performance + Lower power sucking + heavily contra bribed SKU platform that runs both Android and Windows x64 b) Higher performance + similar (may be slightly) higher power sucking non subsidized SKU platform that runs both Android and Winx64 with partners like Asus doing dual booting etc.
And Rory Reed was adamant in pooh-poohing anything but Windows 8.x for AMD and see where it led it into.
With respect to ARM server business, currently Applied Micro(ApM) is shipping its X-Gene which is its own implementation of ARM V8 ISA aka ARM 64 custom core. While AMD's Seattle would be a ARM Cortex A57 derivative. It will be 2016 before AMD would have an answer for that. And if the ApM custom core turns out to be faster that Cortex A57, then AMD is again fighting on the price front and likely losing money.
BTW if you look HP's Moonshot servers which originally was supposed to include AMD x64 (and possibly ARM64), now has the following http://www8.hp.com/us/en/products/servers/moonshot... m4xx - ApM XGene ARM64 m8xx - TI Keystone II ARM32 Cortex m7xx - Low power Xeons x64
So AMD has its a**e handed by HP (which is even till date largest AMD client) by excluding both its x64 as well as ARM64 products. One of the reasons could very well be that AMD basically is now competing with HP, DELLs and IBMs after Sea Micro acquisition. So no magic hat thinking for AMD. It is going to be bloody for longer time.
If Bulldozer even came close to matching, not the hype, but AMD's statements about performance, they would have been much better off. As it was, they were avoided. It's been a continuation of their slippery slope downwards.
They had a two and a half year respite when Intel got cocky with Netburst, and their newly industry leading chip processes. But Intel turned that around, and AMD went back to become an also ran.
They've gotten past the issues their fabs brought, ATI's over-valuation, and, their APUs are finally reaching into areas where they can prove what they do.
Will they grow massively? I HIGHLY DOUBT that. But, I think they will be able to grow bar no shenanigans from competitors.
Testbug00 is a frequent poster on SemiAccurate and the posters that charlie allows to post there (I.E. Charlie bans those who do not follow his doctrine) all tote from the AMD fanboi bible that AMD can do no wrong and Intel/Nvidia can do no right.
Nice Ad hominen. While we're digging through everyone's online histories let's take a look at why you seem to have a problem this poster and a site he frequents.
Hmm... Oh look at that, that last post you made was removed for trolling on that forum. What's trolling, oh wait, trolling's making pointless Ad hominen attacks just like you've done here. QED HighTech4US is a troll. Bring home the cows and leave this guy out to pasture because he's exactly the kind of commenter that's ruining Anandtech.
Turning around? Last year (2013) from Q3 -> Q4 they went up 9% in revenue. It's supposed to be a good quarter with Christmas sales and all. Now they're predicting a 13+/-3% decline in revenue. Their CPU/GPU business continues to trend down and losing money, semi-custom revenue is a spin-off of their general purpose technology. You can see their long term business dying, even if they're propped up by short time business right now.
AMD's revenue is going to drop simply because they don't have any new CPU nor GPU families to launch until Q1 2015, with the exception of little Carrizo over the next two months. Desktop Carrizo is due during H1 as is Pirate Islands, which is too late to make a difference to this year's financials.
Mullins and Beema simply haven't had the success they deserve; when your only competitor sells their competing products at a loss to combat you, it's hardly surprising.
While hindsight is 20-20, the main reason for AMD buying ATI is now clear. To serve Sony and Microsoft for their SOCs for PS4 and XB1 respectively. That part (has) worked well, but the "Fusion" part of applying the synergies to the PC market took far more time and caused AMD to hemorrhage money with a better CPU architecture (Core).
AMD has a plan, and its not to sell ATI. ATI is at the corer of its plan. When it comes to chips for PC's, AMD's is weak. The old bulldozer is just too slow in too many benchmarks for most people or companies to want to buy. I know I have the first Intel chip in my system in almost 20 years right now. So whats the CPU plan? The K12... If the K12 chip is as good as AMD is hoping it is, when it is paired with the next version of the GPU AMD is working on, they might just be able to compete with Intel again. I say might because I have yet to see any benchmarks on the K12. I have no doubt that AMD can boost its GPU speeds. Lisa Su, the new CEO at AMD is an engineer with an impressive background. She has pretty much been running the show for a while anyways, so lets give her some time to show us whats in the works. I believe that she is going to be able to put AMD's PC division back into the running against Intel, and turn that half of the company into a money maker.
She is a pretty good pick. But what can she do, technically? If they haven't been able to do anything for years, it's not likely she will be able to do it now. And she' son in an engineering position, she's management.
What I mainly do not understand is, why right after they saw what kind of fiasco Bulldozer is, they did not start complete revamp of the architecture, but tried to optimize the Bulldozer. That optimisation is going forward, yes, but in such tiny steps, that Intel easily increases lead their i3 /facepalm.
You seem to missread their financial results. They have a big operating income of nearly 4.5% of their Revenue, and have similar results for years now. Don't be confused by the fact that their Net Income is lower. That is low by design, because you need to pay taxes on it. Much better to hand out your profits in the form of debt interest, i.e. tax-free profits.
Where do you guys come from? No company makes lower profits so that they can pay lower taxes. Debt interest? What are you talking about? You need to learn something about economics. They don't pay out "debt interest". There is no profit on debt. Depending on how it's is structured, they may get some taxes back on the debt interest they are paying, but they are still paying that interest. It has nothing to do with profits.
As a matter of fact, most companies will make debt in order to lower taxable profit. Here is an article on Apple doing it openly last year, and commenting on it:
What you are missing is this: The banks don't care if they get dividends on stock or interest on debt. Either way they are getting their return on investment. And the companies don't really care if they pay interest or dividends either, except for one detail: interest is paid before taxes, dividends are paid after taxes.
So yes, every large company will take debts it does not really need, just for the purpose on saving taxes.
Apple didn't want to pay taxes on bringing foriegn earned cash back into the US to do a stock buyback. This is a very different situation. Unless AMD has their loan from an affiliate, paying a ton of interest is not good for AMD. That's still money out the door that can be used for operations or investment.
It's just a pity that their reasonably good discrete GPU business (Radeons plus Fire Pro) is now formally inside this unfortunate "Computing ..." division. I'm wondering what may happen to their Radeons in the longer term...
LISA SU, I HAVE THIS AWESOME IDEA I WANT TO RUN BY YOU:
If getting rid of 7% of workforce results in $85 million savings in 2015 then perhaps you should cut 100% of workforce for $1.2 billion savings ... Or better yet, cut 1000% of workforce for 12 billion savings (and chump change). Look at me, I'm CEOing!!!
The Q4 is very very very disappointing. Maybe the reason Rory left early and maybe GTX970 and the cancellation of 285X had something to do with it. I think the board felt that after Meyer and the lost of mobile market, plus the bulldozer fiasco, the company needed to move faster to the next face before becoming NON-competitive also in the GPU business.
Quote: "two new Semi-Custom SoC designs" What are those?
AMD stated these won't happen before 2016 and then they stated a number like $3 billion in revenue but gave no time frame so it may be $3 billion over 10 years or only $300 million per year.
Meanwhile 2015 will be a bloody year for AMD financially.
AMD is profitable - that's most important. Second, there are plenty of new products in the pipeline: - Carrizo L/Carrizo Q4 2014/Q1 2015 - Radeon 3xx 20nm - 1H 2015 - ARM Opteron - 1H 2015
The biggest miss is the lack of big CPU competitive with i3/i5 with AM3+ compatibility. There are millions of PCs with no upgrade path for Phenom X4/X6. FX is too weak and power hungry.
The no upgrade path on AM3 is what had me just leave that platform.
I finally said enough was enough and built an Intel Socket 1150 Haswell rig using a Core i5-4690k. I'm very happy :)
AMD still has my business as I put a Radeon R9 270X 2GB in the aforementioned rig, which was a replaced for a Radeon 6850. I love their chips for HTPC and general purpose office machines, because well FM2+ is really cheap with the right APU.
You left the AM3+ platform because there's no upgrade path and went to 1150 which has absolutely no upgrade path either? By the time you're going to want to upgrade, the socket is going to be dead as a doornail. Good grief.
But good that you're happy with your i5 4690K, that's a fine CPU. It's just that the FX8350 isn't noticeably worse.
AM3+ compatibility is overrated. With all those new SATA, PCIe and networking standards, you want new mobo anyway. Especially when AM3+ has like 10 years old roots.
Easy to say, but they badly need a step change in single thread performance - and a process shrink from their fab partners. Kinda hard to compete with 14nm (Broadwell) when your stuck on 28nm. [Assuming all else being equal; thats a 1/4 the area!!]
Needing to differentiate themselves from Intel in some way or form; perhaps AMD should stick to quad core or lower, meaning a relatively higher area budget per socket for each core and also allowing a proportionally higher transistor budget for the decode logic where they painfully lag behind.
If Jim Keller manages to pull off anything of note, then he truly will have earned his crust.
Well, I think it really has all to do with Intel having new product, while AMD's AM3+ socket is old, and the refresh chips aren't enough reason to invest in what most people already think of as a dead socket.
The APU's are neat chips, but considering my Phenom II 955 BE can outperform those chips on some benchmarks, and that the internal GPU's are nice, but not quite nice enough to properly replace a discrete GPU for gaming, it makes you wonder, what market is AMD aiming at with the APU's?
It's the only platform that has some modern chipsets currently from them. The Intel Z97 and X99 have all the latest and greatest features and there is nothing even close to those in the AMD camp.
I have always valued AMD for their performance to price ratio's versus Intel's, but at this point, I'm in the market for a new PC and there is nothing in the AMD camp to interest me at all, not even GPU's considering Maxwell just launched.
The AM3+ socket is basically dead. The features and PCIe 2.0 just aren't going to cut it for me if I'm going to invest in a new platform, I WANT the newest of the new features and hope it lasts for another 4 to 5 years, like my current PC did.
The APU's, there is nothing there to work with, since I game, and the graphics in those APU's aren't going to be enough for games I want to play and have high settings the way I want.
The graphics cards, with the R series had some nice cards, but even then, I was willing to look at Nvidia because the R series high end stuff chewed up more power than I'm willing to deal with.
Now with the Maxwell release, there is zero reason to even buy the high end R series, when you can get a 970 that has damn near the same performance, or better, than the R290 and has much better efficiency which will equal a lower energy bill each month.
It's no wonder Intel had a record quarter... when AMD really doesn't have anything to offer the PC desktop market at the moment.
Sure you can argue the price differences and for budget gamers or people not needing all the newest features AMD does have plenty of that to offer, but it seems that segment isn't enough to keep the money from going Intels direction.
I just hope AMD does manage to come out with something by next year, because without the competition, Intel would be able to charge whatever they want (although, I think they already do) without AMD at least nibbling enough to keep the prices in somewhat of a check.
"AMD will then adjust their real estate footprint to accommodate the smaller workforce, which could mean additional infusions of cash from the sale of buildings."
700 people leaving isn't enough to free up any buildings. What is happening is that they'll rent office space to other tech companies. I know this both because of the numbers and because I know a company that's going to rent a floor from them.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
81 Comments
Back to Article
webdoctors - Thursday, October 16, 2014 - link
It's 700, not 7K layoffs LOL. 7K would be almost the entire company.RussianSensation - Thursday, October 16, 2014 - link
The headline needs to be corrected -- Did you mean 7% of the company's workforce is to be cut instead of 7K jobs? From the source: "Reduce global headcount by 7 percent, largely expected to be completed by the end of Q4 2014"Hrel - Thursday, October 16, 2014 - link
Consoles are helping a lot, but their computing division is still far too weak. Man, I feel like AMD has been on life support for nearly a decade and it's getting close to time to pull the plug.Maybe sell off ATI to Intel.
meacupla - Thursday, October 16, 2014 - link
Yeah, but then, without the graphics division, AMD will be left with nothing good to sell.And then intel will squander ATI, because they lack any competence in writing some proper graphics drivers that will work and harness the full potential of the GPU.
Wreckage - Thursday, October 16, 2014 - link
"The Computing and Graphics segment revenue decreased 6% from last quarter and 16% year-over-year. "Da W - Friday, October 17, 2014 - link
Sale of fabs to Global foundries put them at a 2 gen lag versus Intel, but their Buldozer architecture is the nail in the coffin. I'm sure they saw their mistake 3 years ago but it takes time to develop a new architecture, so for the time being they make what they can with their dual module crap.melgross - Friday, October 17, 2014 - link
Feh! They've always been behind Intel in process technology. I remember when they were a year behind, and Ruiz said they were working to get to six months. Instead, they fell to eighteen months. This is nothing new, it's just continuing the path set earlier.Samus - Friday, October 17, 2014 - link
What was amazing is the time (over a decade ago) when AMD was still two generations behind Intel's manufacturing technology, but still producing technologically superior, higher performance, and less expensive chips.AMD did have some manufacturing breakthroughs, though. Copper interconnects and integrated memory controller...and of course 64-bit x86 extensions.
dragonsqrrl - Friday, October 17, 2014 - link
You guys have strange memories. Around a decade ago AMD was first to 90nm, and SOI was superior. How were they 2 generations behind Intel in manufacturing tech? In addition they also had the superior architecture... And it was all down hill from there.Alexvrb - Friday, October 17, 2014 - link
The problem was that they were almost too successful for their own good. They awakened a slumbering giant. In the short term Intel held them at bay via questionable business practices, back when AMD had the technological upper hand. Meanwhile Intel buckled down and poured money into their fabs as well as R&D, to catch up and pass AMD. Intel has done a pretty good job of keeping their guard up ever since.AMD has done pretty well overall in the graphics market and their APUs are competitive for the money. I have hope that Keller's team will create something interesting for 2016.
sonicmerlin - Saturday, October 18, 2014 - link
You really are desperate to avoid blaming AMD. This had nothing to do with intel and everything to do with AMD buckling down on the utter failure of an architecture that is bulldozer.JlHADJOE - Sunday, October 19, 2014 - link
A lot of AMD's "firsts" actually come from the DEC Alpha line of processors. The IMC first appeared on the DEC Alpha 21064/6, introduced way back in 1992. Ditto with the double-pumped FSB that gave Athlons their edge over the Pentium II/III.During those early days, Intel was just catching up to RISC with the Pentium. Jerry Sanders gave AMD an excellent jump start by bringing people from DEC (Dirk Meyer!) on board and incorporating a lot of technology that Intel was just starting to develop on its own.
dragonsqrrl - Friday, October 17, 2014 - link
"They've always been behind Intel in process technology."Uhh no they haven't. Am I the only one who remembers the 90nm Athlon 64's?
atlantico - Friday, October 17, 2014 - link
The Bulldozer architecture is incredibly good, I can't see a single reason that would prevent me from recommending it to even demanding desktop users.Price? Great. Performance? Better than anyone needs.
But the market has spoken, the new FX line has been ridiculed by idiot nerds and their tag-alongs (who don't know anything anyway)
And AMD has blinked. The so called APU architecture is the future, but when Bulldozer was introduced, APUs were not ready, but in 2015 they probably will be.
In reality I don't know what AMD is planning, but I suspect that getting rid of Rory Read was a good move, he was getting the company nowhere.
sonicmerlin - Friday, October 17, 2014 - link
Is this sarcasm? Or do you honestly believe the Bulldozer architecture is "incredibly good"? I just... you can't be serious.silverblue - Saturday, October 18, 2014 - link
If you were to look at "good enough computing", perhaps, but the competition has moved on. There are strong points to the architecture, and I maintain my belief that an octo-core Steamroller would be a very strong multithreader, but they'll only really be reaching performance parity with Sandy Bridge over the next year or so, some four years too late.przemo_li - Friday, October 17, 2014 - link
Linux radeon driver is quite good this days :)Even to the point where AMD is ready to reuse some of that code for their Linux Catalyst efforts.
Hope that layoffs spare floss folks.
Samus - Friday, October 17, 2014 - link
Unless they have an ace up their sleeve, I don't think their graphics division will be profitable much longer either. The channel has indicated that since Maxwell, AMD's 290 and 290X sales have almost completely stalled. AMD taking a month to act on a price-cut didn't help, either.I hope they survive. NVidia AND Intel need competition, even if its weak competition...because it was the lack of competition that brought us the Pentium 4's Netburst catastrophe and NVidia's initially weak, overprice Fermi cards...the last time ATI was really dominating graphics with the 4000/5000 series.
silverblue - Friday, October 17, 2014 - link
AMD were providing plenty of competition when the P4 came out, and the 5000 series came out before Fermi meaning ATi were on top at the time. No, definitely not a lack of competition; the more recent CPU launches by Intel are more indicative.atlantico - Friday, October 17, 2014 - link
While none of us have secret information on AMDs near and far future plans and capabilities, there is one third party which does: Apple.Apple has been moving all its new products to AMD GPUs. Including most recently Apple's new retina 5K iMac and the MacPro.
Apple knows something we do not and it votes with confidence for AMD.
meacupla - Friday, October 17, 2014 - link
pretty sure it's a deep discount offered by AMD for apple to buy their cards.tipoo - Thursday, October 16, 2014 - link
Their graphics division has been keeping them afloat. Without ATI they'd really have nothing. No high end GPUs that still pull in money, no APUs which are their biggest CPU sales. They'd just have crappy CPU cores.Wreckage - Friday, October 17, 2014 - link
Without ATI they would have $5.4 billion.Kjella - Friday, October 17, 2014 - link
Without the ATI purchase, AMD would have more cash to keep up with Intel on the CPU side and could have had both ATI and nVidia competing to provide integrated graphics, even on the same chip as SoC vendors do. Particularly as Intel started pushing their integrated graphics aggressively AMD might have ended up holding two good cards in a three-way race by not having their own solution and offering a platform for both ATI and nVidia to launch at Intel, potentially also boosting their fab volume. By committing to ATI, they naturally invited Intel and nVidia to join forces and crush both sides of the company. Still it was 2006 and many things looked different back then and a whole lot could have happened since, it's pretty much all speculation.Mayuyu - Thursday, October 16, 2014 - link
I think they should just close the cpu division. They're not even close to being competitive.meacupla - Thursday, October 16, 2014 - link
Then they would have no APUs, which brought them their console deals.Flunk - Thursday, October 16, 2014 - link
Which are not really that lucrative or the company would be doing better.eanazag - Thursday, October 16, 2014 - link
They are keeping the lights on and 93% of its workforce employed. They have things they can do. They just don't. There are better processes out there. No reason they need to be stuck on 28 nm. They would do better resurrecting their previous architecture and fabbing at these smaller geometries than bulldozer.Mayuyu - Friday, October 17, 2014 - link
Seems like only Apple is using 20nm for non-Intel made chips so far. AMD enough money or market power to buy up all the 20nm capacity like Apple did.TylerGrunter - Friday, October 17, 2014 - link
Samsung also is using 20nm on their SoCs. In particular Exynos 5430 and Exynos 5433 (now called Exynos 7 octa) are build in 20nm.In fact they were doing it prior to Apple.
Flunk - Friday, October 17, 2014 - link
Samsung owns their own fabs so it's not a good example.melgross - Friday, October 17, 2014 - link
What? AMD has no market power, along with no money. Apple bought 30 million x86 chips last year, and 250 million ARM chips of their own design. How can AMD compare to that?testbug00 - Friday, October 17, 2014 - link
paying off debt... I believe AMD is trying to put as much money as possible into paying off debt before profits, or, sometimes when it puts them into losses/worse losses.AMD isn't "stuck" lower process nodes still aren't ready for parts that aren't super-high yield or super-high-margin iirc.
HighTech4US - Friday, October 17, 2014 - link
Quote: paying off debt... I believe AMD is trying to put as much money as possible into paying off debt before profitsAMD has not been paying off debt. In fact they expanded their long term debt with their last refinancing.
All AMD is doing is pushing out when the major portion of the debt becomes due with the hope that somehow they can right the ship.
AMD has done reverse mortgages on their properties in order to get cash to run day by day operations.
Now along with the 7% employee cuts AMD will be selling some property for cash.
None of the above is good and usually if this was done by an individual it is a sign that BK is coming sometime in the future.
melgross - Friday, October 17, 2014 - link
So, you're saying that 22nm isn't ready? Aren't you a bit behind the times? And 20nm is obviously ready.AMD's technology was always poor. I remember going back decades, that it was poor. They often announced chips that were leading edge, but couldn't produce them until Intel surpassed them, and AMD had to sell them at a discount when they first appeared, later on.
Alexvrb - Friday, October 17, 2014 - link
Cost/yield is part of the reason they didn't jump on these new processes right away. The other is that you have to design/redesign to target a new process, so when you're building a chip you have to choose whether to gamble on a new process. Sometimes you gamble and you lose, the process isn't ready for primetime and you suffer.It's also easier to start working with smaller geometries when you're dealing with tiny SoCs. The process will have to mature a bit before AMD adapts its much larger chips. That being said they should bring their smaller cat-core SoCs over to smaller processes ASAP.
testbug00 - Friday, October 17, 2014 - link
you don't need to be competitive in the high end to make money. Brazos is what kept AMD's lights on with Bulldozer and everything else going on. Would Bulldozer having matched/exceeded the hype have been better than Brazos was? Maybe, but, maybe not.AMD's almost anytime you see an x86 AMD semi-custom or embedded win from AMD, that is only possible because of their CPU division being the only game in town besides Intel with an x86 licence (well, there is Via...)
HighTech4US - Friday, October 17, 2014 - link
The low end market that AMD played in is going away as Intel decided to attack it with the low cost Atoms.Next year when RockChip gets to go full out with the Atom expect AMD's former playground to be under full attack.
AMD losses in the CPU division will accelerate.
testbug00 - Friday, October 17, 2014 - link
I did not realize Intel was selling atom's in the embedded and semi-custom markets?Markets where you often need a specific piece of specifications, if the GPU is not up to snuff in an atom, it will never be bought there. If AMD can be modified for a lower cost to add something the company needs, it will be done.
Not sure why you think Atom is targeting Embedded and semi-custom, which is where AMD will likely end up making its money on x86.
Atom is also to slow to target the lower end APU market. The other place wordwide where AMD is likely to make quite a bit.
Given I believe AMD is one of the movers in ARM server chips that has the most experience in servers and working alongside companies that make them, I would give them pretty good chances to take a large chunk of that profitable market. Atom won't do anything there. Core architecture won't do anything their either.
Flunk - Friday, October 17, 2014 - link
Actually Intel is now offering embedded and semi-custom chips. They're even offering their Atom core as IP to integrate with other designs.rocketbuddha - Friday, October 17, 2014 - link
Intel is also now in the semi custom game to a degree (though they hate to admit it). They carefully ensure that their bed and butter is still bringing in the dough though.http://seekingalpha.com/news/2031045-intel-talks-s...
<quote>
He added Intel's new Grantley Xeon CPUs are seeing strong uptake, and that the chip giant now has 35 custom Xeon SKUs for various clients. Facebook, eBay, Microsoft, and other Web giants have been avid buyers of custom Intel server CPUs.
<end quote>
Intel wants Atom variants to fight the ARMy not AMD. But basically due to contra bribes, acceptable performance and better power characteristics, it basically has killed Mullins and any other Tablet SOC that AMD can field in that area. Further they have specific low power Core i SKUs for real performance + acceptable power. So AMD is sandwiched between a
a) Lower performance + Lower power sucking + heavily contra bribed SKU platform that runs both Android and Windows x64
b) Higher performance + similar (may be slightly) higher power sucking non subsidized SKU platform that runs both Android and Winx64 with partners like Asus doing dual booting etc.
And Rory Reed was adamant in pooh-poohing anything but Windows 8.x for AMD and see where it led it into.
With respect to ARM server business, currently Applied Micro(ApM) is shipping its X-Gene which is its own implementation of ARM V8 ISA aka ARM 64 custom core. While AMD's Seattle would be a ARM Cortex A57 derivative. It will be 2016 before AMD would have an answer for that. And if the ApM custom core turns out to be faster that Cortex A57, then AMD is again fighting on the price front and likely losing money.
BTW if you look HP's Moonshot servers which originally was supposed to include AMD x64 (and possibly ARM64), now has the following
http://www8.hp.com/us/en/products/servers/moonshot...
m4xx - ApM XGene ARM64
m8xx - TI Keystone II ARM32 Cortex
m7xx - Low power Xeons x64
So AMD has its a**e handed by HP (which is even till date largest AMD client) by excluding both its x64 as well as ARM64 products. One of the reasons could very well be that AMD basically is now competing with HP, DELLs and IBMs after Sea Micro acquisition. So no magic hat thinking for AMD. It is going to be bloody for longer time.
melgross - Friday, October 17, 2014 - link
If Bulldozer even came close to matching, not the hype, but AMD's statements about performance, they would have been much better off. As it was, they were avoided. It's been a continuation of their slippery slope downwards.They had a two and a half year respite when Intel got cocky with Netburst, and their newly industry leading chip processes. But Intel turned that around, and AMD went back to become an also ran.
testbug00 - Friday, October 17, 2014 - link
If anything, AMD is just turning around...They've gotten past the issues their fabs brought, ATI's over-valuation, and, their APUs are finally reaching into areas where they can prove what they do.
Will they grow massively? I HIGHLY DOUBT that. But, I think they will be able to grow bar no shenanigans from competitors.
HighTech4US - Friday, October 17, 2014 - link
Quote: If anything, AMD is just turning around...You and others said the same thing when Rory came on board and we all see how that turned out.
AMD is in a slow death spiral and at the very end all that may remain is the Consoles and a company trying to sell their IP.
testbug00 - Friday, October 17, 2014 - link
No, I don't think I said anything when Rory came on board. I will say it now that I see results happening.AMD is not in a slow death spiral, if their ARM efforts flop AND their new x86 chip flops, yes, I agree.
However, I don't think both will, the x86 chip, maybe... The ARM chip, I doubt it.
melgross - Friday, October 17, 2014 - link
What results? You sound like a BlackBerry guy. Something will magically happen to save the company.HighTech4US - Friday, October 17, 2014 - link
Testbug00 is a frequent poster on SemiAccurate and the posters that charlie allows to post there (I.E. Charlie bans those who do not follow his doctrine) all tote from the AMD fanboi bible that AMD can do no wrong and Intel/Nvidia can do no right.You can check out Testbug's posts here: http://semiaccurate.com/forums
Testbug's posts concerning Nvidia are always so full of bile that I sometimes wonder if he is actually Charlie with a different ID.
GreenToTheGills - Saturday, October 18, 2014 - link
Nice Ad hominen. While we're digging through everyone's online histories let's take a look at why you seem to have a problem this poster and a site he frequents.http://semiaccurate.com/forums/member.php?u=3046
http://semiaccurate.com/forums/search.php?searchid...
Hmm... Oh look at that, that last post you made was removed for trolling on that forum. What's trolling, oh wait, trolling's making pointless Ad hominen attacks just like you've done here. QED HighTech4US is a troll. Bring home the cows and leave this guy out to pasture because he's exactly the kind of commenter that's ruining Anandtech.
melgross - Friday, October 17, 2014 - link
Grow? They keep shrinking. Why do you think they are having another round of layoffs?Kjella - Friday, October 17, 2014 - link
Turning around? Last year (2013) from Q3 -> Q4 they went up 9% in revenue. It's supposed to be a good quarter with Christmas sales and all. Now they're predicting a 13+/-3% decline in revenue. Their CPU/GPU business continues to trend down and losing money, semi-custom revenue is a spin-off of their general purpose technology. You can see their long term business dying, even if they're propped up by short time business right now.silverblue - Friday, October 17, 2014 - link
AMD's revenue is going to drop simply because they don't have any new CPU nor GPU families to launch until Q1 2015, with the exception of little Carrizo over the next two months. Desktop Carrizo is due during H1 as is Pirate Islands, which is too late to make a difference to this year's financials.Mullins and Beema simply haven't had the success they deserve; when your only competitor sells their competing products at a loss to combat you, it's hardly surprising.
rocketbuddha - Friday, October 17, 2014 - link
While hindsight is 20-20, the main reason for AMD buying ATI is now clear. To serve Sony and Microsoft for their SOCs for PS4 and XB1 respectively. That part (has) worked well, but the "Fusion" part of applying the synergies to the PC market took far more time and caused AMD to hemorrhage money with a better CPU architecture (Core).Mark_gb - Friday, October 17, 2014 - link
AMD has a plan, and its not to sell ATI. ATI is at the corer of its plan.When it comes to chips for PC's, AMD's is weak. The old bulldozer is just too slow in too many benchmarks for most people or companies to want to buy. I know I have the first Intel chip in my system in almost 20 years right now. So whats the CPU plan? The K12... If the K12 chip is as good as AMD is hoping it is, when it is paired with the next version of the GPU AMD is working on, they might just be able to compete with Intel again. I say might because I have yet to see any benchmarks on the K12. I have no doubt that AMD can boost its GPU speeds.
Lisa Su, the new CEO at AMD is an engineer with an impressive background. She has pretty much been running the show for a while anyways, so lets give her some time to show us whats in the works. I believe that she is going to be able to put AMD's PC division back into the running against Intel, and turn that half of the company into a money maker.
melgross - Friday, October 17, 2014 - link
She is a pretty good pick. But what can she do, technically? If they haven't been able to do anything for years, it's not likely she will be able to do it now. And she' son in an engineering position, she's management.ppi - Friday, October 17, 2014 - link
What I mainly do not understand is, why right after they saw what kind of fiasco Bulldozer is, they did not start complete revamp of the architecture, but tried to optimize the Bulldozer. That optimisation is going forward, yes, but in such tiny steps, that Intel easily increases lead their i3 /facepalm.ShieTar - Friday, October 17, 2014 - link
You seem to missread their financial results. They have a big operating income of nearly 4.5% of their Revenue, and have similar results for years now. Don't be confused by the fact that their Net Income is lower. That is low by design, because you need to pay taxes on it. Much better to hand out your profits in the form of debt interest, i.e. tax-free profits.HighTech4US - Friday, October 17, 2014 - link
Quote: Don't be confused by the fact that their Net Income is lower. That is low by design, because you need to pay taxes on it.No it is lower because of the very large interest payments on their huge long term debt.
melgross - Friday, October 17, 2014 - link
Where do you guys come from? No company makes lower profits so that they can pay lower taxes. Debt interest? What are you talking about? You need to learn something about economics. They don't pay out "debt interest". There is no profit on debt. Depending on how it's is structured, they may get some taxes back on the debt interest they are paying, but they are still paying that interest. It has nothing to do with profits.HighTech4US - Saturday, October 18, 2014 - link
What are you babbling about.AMD has over $2 billion in debt. Interest has to be paid on that debt. That interest is subtracted from gross income resulting in lower net income.
ShieTar - Saturday, October 18, 2014 - link
As a matter of fact, most companies will make debt in order to lower taxable profit. Here is an article on Apple doing it openly last year, and commenting on it:http://www.bloomberg.com/news/2013-05-02/apple-avo...
What you are missing is this: The banks don't care if they get dividends on stock or interest on debt. Either way they are getting their return on investment. And the companies don't really care if they pay interest or dividends either, except for one detail: interest is paid before taxes, dividends are paid after taxes.
So yes, every large company will take debts it does not really need, just for the purpose on saving taxes.
vicbdn - Saturday, October 18, 2014 - link
Apple didn't want to pay taxes on bringing foriegn earned cash back into the US to do a stock buyback. This is a very different situation. Unless AMD has their loan from an affiliate, paying a ton of interest is not good for AMD. That's still money out the door that can be used for operations or investment.TiGr1982 - Thursday, October 16, 2014 - link
It's just a pity that their reasonably good discrete GPU business (Radeons plus Fire Pro) is now formally inside this unfortunate "Computing ..." division. I'm wondering what may happen to their Radeons in the longer term...Arnulf - Friday, October 17, 2014 - link
LISA SU, I HAVE THIS AWESOME IDEA I WANT TO RUN BY YOU:If getting rid of 7% of workforce results in $85 million savings in 2015 then perhaps you should cut 100% of workforce for $1.2 billion savings ... Or better yet, cut 1000% of workforce for 12 billion savings (and chump change). Look at me, I'm CEOing!!!
OWAIT ...
yannigr2 - Friday, October 17, 2014 - link
"two new Semi-Custom SoC designs"What are those?
The Q4 is very very very disappointing. Maybe the reason Rory left early and maybe GTX970 and the cancellation of 285X had something to do with it. I think the board felt that after Meyer and the lost of mobile market, plus the bulldozer fiasco, the company needed to move faster to the next face before becoming NON-competitive also in the GPU business.
HighTech4US - Friday, October 17, 2014 - link
Quote: "two new Semi-Custom SoC designs" What are those?AMD stated these won't happen before 2016 and then they stated a number like $3 billion in revenue but gave no time frame so it may be $3 billion over 10 years or only $300 million per year.
Meanwhile 2015 will be a bloody year for AMD financially.
yannigr2 - Friday, October 17, 2014 - link
Thanks for the info. Just read an article elsewhere where the author was talking about only 1 billion in 3 years starting from 2016.HighTech4US - Saturday, October 18, 2014 - link
Wow that is even worse.fujiyama - Friday, October 17, 2014 - link
AMD is profitable - that's most important. Second, there are plenty of new products in the pipeline:- Carrizo L/Carrizo Q4 2014/Q1 2015
- Radeon 3xx 20nm - 1H 2015
- ARM Opteron - 1H 2015
The biggest miss is the lack of big CPU competitive with i3/i5 with AM3+ compatibility.
There are millions of PCs with no upgrade path for Phenom X4/X6. FX is too weak and power hungry.
jimjamjamie - Friday, October 17, 2014 - link
The Thuban core is the best AMD has ever built in my opinion.creed3020 - Friday, October 17, 2014 - link
The no upgrade path on AM3 is what had me just leave that platform.I finally said enough was enough and built an Intel Socket 1150 Haswell rig using a Core i5-4690k. I'm very happy :)
AMD still has my business as I put a Radeon R9 270X 2GB in the aforementioned rig, which was a replaced for a Radeon 6850. I love their chips for HTPC and general purpose office machines, because well FM2+ is really cheap with the right APU.
atlantico - Friday, October 17, 2014 - link
You left the AM3+ platform because there's no upgrade path and went to 1150 which has absolutely no upgrade path either? By the time you're going to want to upgrade, the socket is going to be dead as a doornail. Good grief.But good that you're happy with your i5 4690K, that's a fine CPU. It's just that the FX8350 isn't noticeably worse.
http://cpuboss.com/cpus/Intel-Core-i5-4670K-vs-AMD...
testbug00 - Friday, October 17, 2014 - link
except, of those "millions" of PCs I doubt over a hundred thousand would upgrade. And, a hundred thousand is being very generous.Not worth the cost. 20nm Radeon's you say...
ppi - Friday, October 17, 2014 - link
Do you remember Radeon 5870 story? First on new node with GTX 480 far far away.They no doubt wanted to pull that off again, but the process tech did not click in this time.
ppi - Friday, October 17, 2014 - link
AM3+ compatibility is overrated. With all those new SATA, PCIe and networking standards, you want new mobo anyway. Especially when AM3+ has like 10 years old roots.Atari2600 - Friday, October 17, 2014 - link
Easy to say, but they badly need a step change in single thread performance - and a process shrink from their fab partners. Kinda hard to compete with 14nm (Broadwell) when your stuck on 28nm. [Assuming all else being equal; thats a 1/4 the area!!]Needing to differentiate themselves from Intel in some way or form; perhaps AMD should stick to quad core or lower, meaning a relatively higher area budget per socket for each core and also allowing a proportionally higher transistor budget for the decode logic where they painfully lag behind.
If Jim Keller manages to pull off anything of note, then he truly will have earned his crust.
silverblue - Friday, October 17, 2014 - link
They fixed the decoder with Steamroller; they just don't have the execution resources per thread (correct me if I'm wrong).sonicmerlin - Friday, October 17, 2014 - link
AMD plans on competing with Intel by releasing yet another Bulldozer CPU with a lower IPC than their K series Athlons.silverblue - Saturday, October 18, 2014 - link
Eh? You might want to provide a link.PCjunkieXL - Friday, October 17, 2014 - link
Well, I think it really has all to do with Intel having new product, while AMD's AM3+ socket is old, and the refresh chips aren't enough reason to invest in what most people already think of as a dead socket.The APU's are neat chips, but considering my Phenom II 955 BE can outperform those chips on some benchmarks, and that the internal GPU's are nice, but not quite nice enough to properly replace a discrete GPU for gaming, it makes you wonder, what market is AMD aiming at with the APU's?
It's the only platform that has some modern chipsets currently from them. The Intel Z97 and X99 have all the latest and greatest features and there is nothing even close to those in the AMD camp.
I have always valued AMD for their performance to price ratio's versus Intel's, but at this point, I'm in the market for a new PC and there is nothing in the AMD camp to interest me at all, not even GPU's considering Maxwell just launched.
The AM3+ socket is basically dead. The features and PCIe 2.0 just aren't going to cut it for me if I'm going to invest in a new platform, I WANT the newest of the new features and hope it lasts for another 4 to 5 years, like my current PC did.
The APU's, there is nothing there to work with, since I game, and the graphics in those APU's aren't going to be enough for games I want to play and have high settings the way I want.
The graphics cards, with the R series had some nice cards, but even then, I was willing to look at Nvidia because the R series high end stuff chewed up more power than I'm willing to deal with.
Now with the Maxwell release, there is zero reason to even buy the high end R series, when you can get a 970 that has damn near the same performance, or better, than the R290 and has much better efficiency which will equal a lower energy bill each month.
It's no wonder Intel had a record quarter... when AMD really doesn't have anything to offer the PC desktop market at the moment.
Sure you can argue the price differences and for budget gamers or people not needing all the newest features AMD does have plenty of that to offer, but it seems that segment isn't enough to keep the money from going Intels direction.
I just hope AMD does manage to come out with something by next year, because without the competition, Intel would be able to charge whatever they want (although, I think they already do) without AMD at least nibbling enough to keep the prices in somewhat of a check.
nofumble62 - Saturday, October 18, 2014 - link
If there is any talent left at AMD?barleyguy - Sunday, October 19, 2014 - link
"AMD will then adjust their real estate footprint to accommodate the smaller workforce, which could mean additional infusions of cash from the sale of buildings."700 people leaving isn't enough to free up any buildings. What is happening is that they'll rent office space to other tech companies. I know this both because of the numbers and because I know a company that's going to rent a floor from them.