Is this the end for AMD?

[Tsavo]

Could AMD stand to generate more x86 revenue just by holding Intel's revenue hostage with licensing fees?

I'd wager Intel covered all their IP bases with their settlement with AMD.

What's telling in the fact their GPU division makes next to no money and they can't deliver a competitive CPU now, because the costs associated with both design and process will only get massively more expensive with each iteration. They can't afford to deliver now, and they sure as hell won't be able to deliver in the future.

What worries me is that they have nothing on the vine that can carry them into the future on the CPU side of their APU business for even the low end, five years from now.

As a few posters wrote in this and other threads, it's probably a really good time for AMD to find someone to buy them, and I do not disagree. AMD has had at least a decade of actions by a really bad Board of Bystanders and some largely incompetent executive staff and now they are paying the price for it at a time when they can least afford to.

 

[IlllI]

1) AMD will completely remove itself from high-end x86 CPU designs,
2) AMD will completely remove itself from high-end GPU designs, and,
3) The savings from #1 and #2 will be redirected to Fusion/Bobcat/low-power devices with a focus on low-power all-in-one CPUs with embedded graphics, and/or possibly a development of a new CPU architecture designed by AMD to compete in smartphones/tablets.

if this happens i fear intel prices will skyrocket. only competition (small as it is) is keeping them from basically pricing whatever they well please.

i for one am not looking forward to that kind of future.

but i dont think amd has the resources to keep it going on that path, so you are probably right about your predictions.

 

[RussianSensation]

Are "successful" CPUs, only those CPUs that beat Intel? So by that metric, Phenom II was in fact rather successful, in my mind.

Depends on what you mean by "beat Intel". AMD has consistently lost CPU market share, relinquished average selling prices (due to less the competitive CPUs). Maybe you were happy to get a Phenom II for $130-160, but that's certainly not winning. Winning is being able to successfully sell a $180-300 CPU at 40-60% profit margins. Even then, look at the majority of users - they consider AMD's CPUs to be competitive only on the low-end. Phenom II X4 is barely better than Q6600 from 2007, while the X6 is essentially an X4 with 2 of its cores unused in half the apps for most people. Not sure how that's winning tbh. On top of that, the X6 has a huge die size compared to Core 2 Duo/Quad/Lynnfield/Sandy Bridge, etc., and yet AMD is "giving these" away for $160.

I am pretty sure most of the gains in AMD's CPU market share came from Brazos and Llano platforms. By that account, Phenom I, II and Bulldozer were all failures since none of them met their target goals of gaining back market share while returning AMD to higher profitability. With Intel's market share now 95% in servers and > 80% overall in PC microprocessors, AMD's wall is very high to climb with 3 consecutive high-end CPU generations failing. It would be one thing if AMD had small market share but extremely high profitability in those segments (like Apple in the smartphone market).

I thought that AMD's GPU strategy, was the opposite of that. That they wouldn't try to create a bigger, hotter, faster GPU than NV, rather, they would build smaller, cooler-running, and scalable GPUs, so that they could take the lead by doubling-up on their GPUs on one card for a flagship.

In theory this likely was an excellent idea, but the small die strategy clearly didn't result in a more profitable AMD as was originally thought. By pricing AMD's high-end GPUs so low ($299-379), AMD was now forced to lower the price for all other segments below that.

Recall: X800Pro vs. 6800GT - $399: Battle of the Mid-High End
$540 X800XT PE vs. 6800U: Battle of High-End

^ Now, the "equivalent" of the X800Pro (then a mid-high end card at $399) is the HD5850 that sold for $269 and HD6950 2GB that had an MSRP of $299 but now that sells for $230-240. Wow! Giving up so much profitability.

Now, 9800XT, X800XT/XT PE, X1800XT, X1900XT/X, X1950XT/X, were all high-end chips that sold for $500-650.

Quick Case Study from a Business (not Gamer's) Perspective: Did the Small Die Strategy Actually help ATI/AMD's GPU division to improve profitability?

9800XT ($499) = RV360/380 dies size = 210-mm^2 (best I could find)
X800XT ($499), X800XT PE ($549) = 257-260 mm^2
X1800XT ($549-599) = 263 mm^2
X1900XT ($549), X1900XTX ($649) = 314.5 mm^2

vs.

HD2900XT = performance was a let down (so no point in even discussing this card)
HD3870 = 192 mm^2 (performance was a let down, and prices had to be lowered). Die size very close to a $500 9800XT
HD4870 ($299) = 256 mm^2 (priced at almost half of X800XT/X1800XT despite similar die size!! failed)
HD5870 ($350) = 334 mm^2 (failed at achieving a small die)
HD6970 ($370) = 389 mm^2 (failed at achieving a small die)

AMD's Cypress/Cayman chips are not small by AMD's historical standards and yet were priced significantly lower.

AMD's Evergreen (HD4000) series was similar to historical die sizes, but was sold for half of what ATI would normally sell its GPUs.

So basically ATI's small die strategy under AMD has actually resulted in: "Larger die sizes, and lower profitability due to lower Average Selling Prices, while eroding ATI's brand value on the high-end by conditioning high-end gamers to pay $300-350 for a high-end AMD GPU, not $500-600"

If I were to take my "gamer"/"hardware enthusiast hat" off for a second and critiqued AMD's GPU small die strategy/new GPU price positioning -- they are now basically selling just as expensive to manufacture GPUs as ATI did in the past, but selling them for nearly half the price!! I would call it a HUGE fail.

AMD is selling 2x HD6970 (389mm^2 dies) in the form of an HD6990 for $700 today, and yet X1950XTX with a smaller die size sold for $649.

Keep this strategy up long enough and you'll be going out of business.

 

[boxleitnerb]

if this happens i fear intel prices will skyrocket. only competition (small as it is) is keeping them from basically pricing whatever they well please.

i for one am not looking forward to that kind of future.

but i dont think amd has the resources to keep it going on that path, so you are probably right about your predictions.

No, customer demand is dependent on affordable prices. They may rise (rather, progress will slow down) but definitely not skyrocket.

 

[Tsavo]

No, customer demand is dependent on affordable prices. They may rise (rather, progress will slow down) but definitely not skyrocket.

+1. They know who butters their bread, and it isn't the 990x users.

 

[bunnyfubbles]

if this happens i fear intel prices will skyrocket. only competition (small as it is) is keeping them from basically pricing whatever they well please.

maybe 10 years ago it would happen, but not now with x86 under assault by ARM, Intel will have to fight for relevancy as ARM devices become more than fast enough for the average person's computing needs

really the only thing that could potentially suffer egregious price increases would be higher end parts (ie platforms like socket 2011)

heck, proof is in the pudding, AMD hasn't had a truly competitive part in the past several years, yet intel hasn't increased their prices. The only downside that has thus far happened is intel took away overclocking from their budget parts due in part to the lack of competition. Granted, recent news of IvyBridge suggests we'll get some degree of overclocking back.

 

[Revolution 11]

Wait, I always thought that AMD was barely profitable on their GPU division because the finances were mixed with other divisions. Are they really doing that bad in high-end GPUs or is it just complicated accounting practices?

 

[MagnusTheBrewer]

maybe 10 years ago it would happen, but not now with x86 under assault by ARM, Intel will have to fight for relevancy as ARM devices become more than fast enough for the average person's computing needs

really the only thing that could potentially suffer egregious price increases would be higher end parts (ie platforms like socket 2011)

heck, proof is in the pudding, AMD hasn't had a truly competitive part in the past several years, yet intel hasn't increased their prices. The only downside that has thus far happened is intel took away overclocking from their budget parts due in part to the lack of competition. Granted, recent news of IvyBridge suggests we'll get some degree of overclocking back.

It depends on how you define "competitive." The vast majority of computer users, including a large number of forum members, are absolutely fine with 'good enough.' The idea that AMD will disappear or relegate itself to some small niche is foolish at best and wishful thinking by Intel fanbois at worst.

 

[Tsavo]

The idea that AMD will disappear or relegate itself to some small niche is foolish at best and wishful thinking by Intel fanbois at worst.

They are already niche in servers and headed that way on the desktop. That the costs of design and process will only increase and keep increasing while their ASP remains stagnant to decreasing while their market share declines doesn't automatically mean that they will vanish, but doesn't exactly bode well for the future.

Things will look better for them if their upcoming respin of BD fixes its awfulness, but that seems like a pretty damned tall order.

 

[cbn]

What I'm curious about is the scenario where AMD just pulls out of x86 across the board and tells Intel to sit on it and rotate when it comes to the x86 IP that Intel needs to license for their existing products.

What degree of exhorbinant licensing fees could AMD extract from Intel in such a scenario? Given the volume of Intel's revenue that is dependent on having access to AMD x86 IP, could AMD stand to generate more x86 revenue just by holding Intel's revenue hostage with licensing fees?

Wow!

These are some really interesting ideas!

EDIT: I wonder how much of this would still stand, though, if AMD decided to building its own Custom ARM Cores? (Surely AMD would need to license some Intel patents to pull that off)

 

[cytg111]

in 5 years, give or take ? Sure ..
AMD will be relevant again when the cycle repeats.

1. Intel holds superior product, no competition
2. Future Intel products remain there, in the future, no competition, milking the market.
3. The giant dozes off, with just one bullet in barrel (haswell)
4. AMD intercepts and surpasses with xxxx product.

That might take 5 years. Investing in SB or IB now is problary the best CPU investment you are going to make this decade, cause it will last you min. 5 years.

 

[MagnusTheBrewer]

They are already niche in servers and headed that way on the desktop. That the costs of design and process will only increase and keep increasing while their ASP remains stagnant to decreasing while their market share declines doesn't automatically mean that they will vanish, but doesn't exactly bode well for the future.

Things will look better for them if their upcoming respin of BD fixes its awfulness, but that seems like a pretty damned tall order.

You conveniently left out the part where I said, "It depends on how you define "competitive." The vast majority of computer users, including a large number of forum members, are absolutely fine with 'good enough.'"

Folks on these forums and others dearly love to believe they are the 'face' of computing. Gamers and enthusiasts certainly have influence but, they are NOT the sole voice of computing or even the most influential.

 

[sandorski]

lol, so much angst over the failure of BD to deliver. AMD may cease to become relevant for the High-end Hobbyist, such as ourselves/many here, but they have made great inroads into the Mainstream recently. Llano/Zacate have the real potential to make AMD be the success against Intel that they have always wanted. Intel has a great disadvantage in comparison to the APU and the Market where the APU competes is the Market to be in, on the x86 side anyway.

I highly suspect the Enthusiast CPU is quickly a dying breed. It won't be long before the APU/Intel equivalent will be the only choice. Intel will have to either figure out how, finally, to make a decent GPU core or acquire one. If it doesn't, AMD will gobble up Marketshare and once be the darling of all.

 

[Concillian]

None of that tells me any priority on high-end /premium market for desktop CPUs, discrete GPUs. :'(

Put it this way, how can you have a "drastic" or "massive" restructuring plan without changing anything about the business?

Right, but where are the APU designs going to come from? Continual re-hash of the same GPU design is the only option with a small, lean APU-centric GPU division. The discrete graphics portion only made a small profit on the discrete side, but it saved the CPU side's bacon. Brazos nor Llano would not be what they are without the GPU side development, and they are directly responsible for a good portion of the APU side profit.

Perhaps they will try and coast on what designs they have, but that will truly be the death knell for AMDx86 if so, they will have put themselves in a position where they're behind on CPU side, they're behind on OpenCL (behind nVidia / CUDA unless GCN is a home run). One of the main points of APU was OpenCL performance and that isn't there yet.

I think it would be short sighted to slash the GPU side. Much of the RIF was PR / marketing, which is good, technology can speak for itself. A64 didn't need massive marketing, it needed to have a strong technical foundation compared to competitors. The GPU side is the only place they have demonstrated that's even possible in the last few years.

Also, you don't have a 10% RIF, then announce a week later you're going to do it again. I don't see it. They're going to enter a market where they can leverage their successful side. They will enter a market where competitors are on roughly equal ground from a foundry / manufacturing standpoint, as this is an area where they have demonstrated they can compete. They know they can't be competitive with Intel when that is as much GloFo competing with Intel's foundries as it is AMD vs. Intel. They want into some market where they can flex their muscle. It might be ARM, it might be Brazos-style servers, who knows? But it will be something where TSMC or GloFo parts are at least on a par with whatever other people have to offer in that arena.

 

[smartpatrol]

You conveniently left out the part where I said, "It depends on how you define "competitive." The vast majority of computer users, including a large number of forum members, are absolutely fine with 'good enough.'"

Folks on these forums and others dearly love to believe they are the 'face' of computing. Gamers and enthusiasts certainly have influence but, they are NOT the sole voice of computing or even the most influential.

The problem for AMD is, the kind of user who's satisfied with "good enough" has even less reason to choose Bulldozer than a hardcore gamer does.

Why would a user who just wants "good enough" spend the extra money for a Bulldozer CPU rather than going with a Core i3, Phenom II, Athlon II, or Pentium?

 

[IlllI]

No, customer demand is dependent on affordable prices. They may rise (rather, progress will slow down) but definitely not skyrocket.

customer demand wont change. if intel is the only game in town, what other options are there? none. if intel has the entire market to themselves, they can price whatever they damn well please. either the customers pay up or go with nothing. thats what will happen if amd abandons that market.

maybe 10 years ago it would happen, but not now with x86 under assault by ARM, Intel will have to fight for relevancy as ARM devices become more than fast enough for the average person's computing needs

arm cant hold a candle to intel processors. and good enough for what? web browsing? lol.

i know it wasnt clear but my post was actually referring to the upper echelon processors and servers and all the high margin processors that intel makes. not those crappy processors that are inside phones and ipads

 

[boxleitnerb]

customer demand wont change. if intel is the only game in town, what other options are there? none. if intel has the entire market to themselves, they can price whatever they damn well please. either the customers pay up or go with nothing. thats what will happen if amd abandons that market.

No, the point is, that Intel needs to make money to make shareholders happy. If they increase prices by 50%, business will go down the drain as people won't buy new CPUs as often as they used to.

 

[cbn]

if intel is the only game in town, what other options are there? none. if intel has the entire market to themselves, they can price whatever they damn well please. either the customers pay up or go with nothing. thats what will happen if amd abandons that market.

Non-overclocked CPUs have a long lifetime.

If Intel doesn't improve most people will just hold onto what they have.....and be happy for many years.

The big exception to this would be mobile....where smaller form sizes and battery life are still improving.

 

[Nemesis 1]

They make almost no profits from their GPU business even in the best of times. Any new CEO is going to see this as a huge red flag.

AMD abandoned the high-end for CPU's a long time ago, not willingly though.

If they came out and claimed it was a "strategy" to bail on the high-end CPU market then it would merely be an acknowledgement of reality for the past year or two.

The layoffs are telling that they are in GPU.

What I'm curious about is the scenario where AMD just pulls out of x86 across the board and tells Intel to sit on it and rotate when it comes to the x86 IP that Intel needs to license for their existing products.

What degree of exhorbinant licensing fees could AMD extract from Intel in such a scenario? Given the volume of Intel's revenue that is dependent on having access to AMD x86 IP, could AMD stand to generate more x86 revenue just by holding Intel's revenue hostage with licensing fees?

AMD can't stop Intel from using ANY existing x86 IP . If you got proof other wise show it.

 

[dma0991]

I do hope AMD does not go down. If AMD did then the price would go up instead. Hopefully Intel would not wring my wallet dry with IB. :hmm:

 

[cbn]

AMD can't stop Intel from using ANY existing x86 IP . If you got proof other wise show it.

It would be great to find out how these Hardware IP agreements are written.

What conditions and clauses are contained within the documents to prevent one competitor from pulling out of the business and demanding licensing fees from the other competitor?

Also It would be interesting if there are clauses to prevent existing licensed x86 CPU patents from being used for other non-x86 CPU projects.

For example, If AMD wanted to transfer some Intel CPU IP from its own x86 processor line-up to another non-x86 (ARM, etc) processor would it have to offer up additional GPU patents to Intel? Or could this be done without having to give up anything?

 

[gmaster456]

You conveniently left out the part where I said, "It depends on how you define "competitive." The vast majority of computer users, including a large number of forum members, are absolutely fine with 'good enough.'"

Folks on these forums and others dearly love to believe they are the 'face' of computing. Gamers and enthusiasts certainly have influence but, they are NOT the sole voice of computing or even the most influential.

Very well put.

 

[LiuKangBakinPie]

They fired a couple of cleaners. So what? maybe they can get real guys to do the R&D now and the pc shops can look like pc shops again not construction yards.

 

[Kevmanw430]

We do have to realize that, while AMD's GPU division is not making much money right now, they are pouring money into 28nm for the 7xxx series. If they can shrink the die and get good yeilds, destroy NV in performance AND get 28nm parts out before then, they will make money. Also, again, if they get out the parts before NV and they do in fact destroy NV parts, then they can also raise prices for higher profit margins.

Also, if Trinity ends up well (Using Enchanced Bulldozer), then they may again be in a very competitive place with Intel. They would have to make Bulldozer a lot more power efficient, though, but alot of that comes from inefficiencies on GloFlo's 32nm process. If they can sort out the bugs, and get power consumption down, while using a m ore powerful 28nm IGP (w/the 512MB DRAM they may use?), then they will kill Intel in the low to mid end laptop and desktop department. Sure, their CPU part most likely wont be anywhere near as fast, but the GPU should kill Intel's, and the CPU will be fast enough for any mainstream user.

For AMD to be profitable, they need to target the low-mid end consumer market, not the enthsiast market. We make up a small percent of the overall PC users globally.

Also, they NEED a better server chip. Bulldozer probobly performs well in server loads, but the performance/watt is terrible, something crucial in servers.

 

[lol123]

If AMD exits the high-end x86 space including server processors (which I find unlikely), I suspect that IBM will enter it one way or another (either by purchasing the AMD server division or starting to design x86 CPUs themselves). As far as I understand it they still have a valid x86 license. The natural counter to this is that IBM is only mildly interested and only reluctantly involved in the hardware industry, but the question is what they (and other server makers) would think of a situation where the server market is dominated by the x86 architecture (and the market is heading rapidly in that direction) and that architecture in turn has only one player/manufacturer (which would be the case if AMD leaves it).

This raises another question: what would a POWER CPU with an x86 frontend (think of something like AMD K5) look and perform like?