AMD said on Tuesday that it would embrace a strategy of developing processors for a wider range of products outside computers and called on the industry to focus on user needs rather than creating “technology for technology’s sake.” Our Take: In the last year it has been made clear that AMD could no longer outpace Intel in the MHz, speed and power consumption race of PC CPUs. Even the highest end models of the new AMD Opteron and its desktop version, Athlon64, which are to have a modern 64-bit core, according to SPEC benchmarks they will not be the speed leaders by the time they will come out (Q2 2003), bringing AMD to yet another race against Intel’s Xeons/P4s. Oh well, shift happens.
AMD Shifts Focus Away from PCs
2002-11-24 AMD 38 Comments
It’s just the slashdot sensationalism. They’re NOT getting out of the PC business and they’re not going to stop making processors for PCs. The article is FUD.
I don’t believe that they will go out of the PC market completely. The article never said that. The article said that they will shift the main focus away from the PC to other industries.
Like it had happened with Be and BeOS. When Be shifted focus away from BeOS, they didn’t “kill” it overnight. Their main focus at that time was BeIA but they were supporting BeOS for quite some time after that to some of their contractor clients.
Same with AMD, I expect them to try to find another market to dominate, as Intel has made very clear with their roadmap and some samples they had running over 4 GHz, that AMD won’t have much luck competing straight head to head with them. Not even with the Opteron technology, which while it is _great_ technology, it does not add up: It will be expensive-ish, it will be already 1.5 years late when it will come out, and the Athlon64 version of the Opteron chipset for the desktops won’t be as powerful as the Opteron one. By that time, Intel will be selling desktop CPUs as powerful as Opteron itself. And if that was not enough, Microsoft has made NO commitement to support and release Windows64 for the Opteron/Athlon64 (there are betas though) and they haven’t include any such version of Windows on their roadmap.
This already is a burden for AMD, and I see their focus shift as a sensible decision. It is all business and logic at the end. If market conditions does not help something to fly as much as the company needs it to fly, the company will have to find other sources of income. For zealots, this kind of thinking is like “I don’t believe it, AMD will never leave us”, but people who have “burned” from such focus shifts this should be a no brainer that it is going to happen, and moreover, it even makes sense to happen.
> AMD won’t have much luck competing straight head to head with them
I hope people realize that AMD has reached the end of the Athlon/XP technology, the latest 2800+ model (running at 2.25 GHz) was only shipped to very “good” customers (like Dell) and it was manufactured in limited numbers, as their CPUs that they could go that high in the MHz range had to be almost hand-picked.
AMD knew of the K7 MHz/heat limits for a while. This is why they have invested in the Opteron technology, which while it is promising and easily marketable because of its 64-bit core, it has hit AMD’s own limitation of finishing up the CPU, as they are already more than a year behind schedule. When AMD were saying that “our new CPU will kick a$$” they were saying that with the mindset that it will come out in time, at a time that the P4 and Xeons were not running at these high speeds they run today. When Opteron will come out next April (fingers crossed that it won’t slip again), their CPU will be among the best performers in that CPU range, but possibly not the best. And the desktop version of Opteron, the Athlon64, which has lots of stuff “cut” in it (less cache etc) won’t be able to compete with the HyperThreaded desktop P4 CPU of that time.
Because of these facts, I see absolutely no bad reasoning in AMD’s decision to focus shift. Their shareholders should be happy. But their users will probably be upset for a while.
AMD can’t afford to lose money for ever, and competing head-to-head against Intel almost guarantees that. There is very little need for people to but a new computer these days, and so the PC market will continue to grow slowly for the forseeable future. I think the big bucks will be in appliances and handheld devices, a market where AMD has far better prospects. AMD already have a strong flash memory business, for example. Intel has taught people to look for a Pentium in their next computer, but I don’t think consumers care what kind of chip exists in their stereo or PDA.
and IBM and will create the POWER4i needed by both compagnies.
Some have felt that a 64 bit PC market is not there. I’m not entirely convinced though. A 64 bit CPU wil help to forge new OS models of address space use that could bring some significant performance improvements that would be difficult with current 32 bit intel architecture. Eugenia, is it fair to compare apples and oranges? Should you not be doing your comparisons between Opteron and Itanium?
But then again, I dont have to buy AMD.
> shift happens
This is hilarious!
This is an interesting exercise to imagine how this change will cascasde through the industry. This will happen over time as others have noted but it will happen and it will have an effect. Intel will be altered by this the most. They will have less pressure on their margins (means more money available) at a time they are also diversifying beyond the CPU world (even more than in the past).
Less competition in the HW business will generally cascade as higher margins and prices. This will really help some companies and hurt other who will suffer from loss of marginal HW producers retreating from the market. PCs will get a little more expensive on the HW side and this will tend to dampen demand even a bit more.
Higher prices and dampened demand will mean even MSFT will have pressure to lower their prices to keep demand up. MSFT has always enjoyed the position of watching others struggle to cut prices while they keep their margins untouched. This “may” change a little. Cheap HW has benefited MSFT at the expense of one of the big selling points of products like Linux or BeOS. It was always possible to hide bloated intentially contrived complexity under the forgiving umbrella of increasing HW speeds and falling prices. When this environment stumbles then more efficient SW has a slight opening to make show the benefits of efficient design.
It could also be that MSFT or APPLE purchase AMD. MSFT may feel that the recent messages from the Bush weakened Feds (and soon to be courts) indicate that they could buy AMD under the guise of stimulating competition in HW now that they are viewed as a friendly compliant monopolist <grin>.
What will Intel do with a little less pressure and a lot more money?
So we see that this news story may have many effects not immediately obvious in many markets not obviously connected.
I think that AMD could change their strategic away by choosing Linux or any other alternative operating system, for example Lindows, Beos, etc… that runs well in AMD processors. I have installed in my computers Linux and Lindows and runs very well using AMD processors.
What I mean is that if AMD couldn’t outpast Intel, they could choose another way, thinking in servers and desktop users that like alterative operating systems…
I know that are version of Linux that support 64 bit processors:)
I don’t believe in benchmarks in the first place, and in the second place the Integer processors of AMD is way supperior to Intel ones… So i’m puzzled how the benchmarks of Intel got superior if they didn’t redo them…
The only way for Intel processors be superior then AMD ones is when using the SSE2 instructions, which makes Intel processors a bit supperior.
But realistically, why would anyone want a dual processor desktop with excess of 2GHz???
It would be much better that when a new processor reachs the streets, the software where already prepared to cope with it… but that would only be possible if the instruction set where forced to become standards… and published… but that is asking to much for a trade secret eskisofrenic society… [that is also like asking for software without bugs included].
For starters, a “shift in focus” can mean any number of things. Secondly, “MHz” is a completely irrelevent metric for comparing CPUs – and SPEC numbers based on preliminary findings for a benchmark with a limited scope is just as disengenuous.
This isn’t rocket science here.
What are you basing this off of? Have you seen the scores of hyperthreading? The result is a plus/minus 5%. To conclude that its some great technology is foolish. Your wording is wrong too. Its says “PC”s. A server is not a personal computer, last time I checked. I usually agree with you Eugenia, but AMD has been around for 25+ years, while Be was around for around 10. AMD’s revenue is in the 250 millions per quarter, I don’t think Be ever eclipsed that. Finally BeOS NEVER had a foothold in against MS, while AMD has compaq and sony (laptops) and several other big names interested/using their processors. UNFAIR ANALOGY.
But then again, I dont have to buy AMD.
I hope you know what you’re saying. Even if you don’t like AMD and/or don’t buy their products, if it weren’t for them being involved in the CPU-making business, we’d probably all be paying $5,000 a piece for low-end Pentium II’s. Competition is good!
I think the big thing is that the PC market is going to contnue to be flat. At the same time, as others have mentioned, the market for other devices s really starting to heat up. I think it’s a good decision.
Guys, did anyone read the article? They are still definitely in the desktop processor market, they are just expanding. They will still be making processors for a good long time, and competitively.
Eugenia, I expect more from you. I hope you would have read the article first.
Hyperthreading like most new technology tools is not an immediate boost for all applications. Some apps may not even work properly if it is enabled. Some apps may be tweeked to take advantage of the new tool. Hyperthreading is more useful on the server but it will help user apps eventually. For now it just makes Windows seem more responsive. The moderate boost from hyperthreading coupled with the Mhz advantage will give Intel advantages. And they still have headroom in the Mhz arena.
This makes it very tough for AMD. The real competitor for Intel may now be the Power line from IBM and they are some formidable challengers.
Although they aren’t the speed kings now, they do give a great bang for the buck. And right now with less competition for Intel, there is little incentive for Intel to continue pushing the Moore’s Law when the economy recovers and demand for PCs resume.
Mike, I’m sure Eugenia read the article. You are right, they wil be still making chips for PC’s. But, I think the main thrust in the posts here is that an outfit like AMD would not be making announcements like this unless, as you said, they feel the need to diversify. And, especially so if they do not see their upcoming new processors as superior to or at least equal to Intel’s.
First of all, Dell is definitely not receiving any of AMD’s 2800s. Mainly because Dell hasn’t bought a single AMD product ever.
AMD realizes that the processor market is changing. There are two main factors. First of all, processor performance, for all practical purposes, is peaking. Videocards are where it’s at; with every new chipset, games are able to place even less of a burden on the processor. In fact, I wouldn’t be surprised if in a couple years, games don’t even use the desktop processor anymore; or at least, they only use it for non-multimedia features, and let the videocard processor determine absolutely everything else.
Second, the desktop is becoming less and less important. Email and communications is switching to mobile devices; not just laptops, which use weaker processors that AMD is better at making, but BlackBerries, cellphones, and the like, which don’t even use x86. And pretty soon, Tablet PCs with Bluetooth keyboards will fully replace laptops (really just a change in formfactor, but still).
The combination of Tablet PC, Palm/cellphone/BlackBerry, and a Bluetooth keyboard is enough to fully replace the desktop PC as we know it for everything short of games… and, as is becoming more obvious every day, fewer and fewer A-list titles are being created exclusively for PC. Good games are simply coming out for consoles now, and I doubt it will be long before the PC becomes shut out of the gaming industry entirely. All it would take is for one company to let you use a keyboard with a PS2, and even the most diehard of FPS gamers would be converted (and everyone else is converted already).
The desktop is dying as a platform. And, as the strongest consoles internationally use custom RISC processors, and portable devices use ARMs or similar processors, the x86 is dying with it. AMD doesn’t want to be there when it happens.
Uhmm, just because AMD is having a focus shift doesn’t mean the prices would shoot up. Processor price plays little part in the whole pricing for the average PC. Prices for other stuff would still go down. If Intel was making every part of the system from the HDDs to the optical drives, and AMD was their only competitor, I would have to be worried.
The good thing is when the economy recovers, Intel can have better profit margins (healthy profit margins) so they don’t have to go down to the red everytime the PC market crashes.
>Dell is definitely not receiving any of AMD’s 2800s
This is correct. I was meaning that they shipped the 2800s to big customers only, because they are in shortage of these CPUs.
>Eugenia, I expect more from you. I hope you would have read the article first.
I read it, and I very clearly said that having a focus shift does not mean that you change your strategy overnight.
>Should you not be doing your comparisons between Opteron and Itanium?
The fact that Opteron is 64bit and Itanium2 is also 64-bit does not mean that the Opteron is not set to mostly go against the Xeons and not the Itanium2s. In fact, the highest end Xeons are faster than the Itanium2s on many things.
…[AMD] called on the industry to focus on user needs rather than creating “technology for technology’s sake.”
Is this admission to the status of this industry coming from a company that sees the truth and is now stating it only because they are failing to compete with their rival?
It is a shame that companies only show interest in the truely negative state of things when they fail to take advantage of it. Take away the money and they try to bash the same lousey system they’d gladly have used to their own advantage, if they could.
I do not want to start a flame war but reading your posts seems that Hammer CPU is late and bad.
I SUPPOSE a thing: hammer cpu and chipsets are not expensive to make. It is only a supposition but think about it: hammer has on chip the support for memory and a form of NUMA. SIS already makes 2 united chipsets (735,745) for athlon. In a single inexpensive chip they are able to put athlon north bridge and south bridge.
Now with Hammer they do not have to implement memory controller. So chipsets will cost ever less.
And an Hammer can link to 2 4 8 CPU without additional (and expensive) external logic. I HOPE to see dual hammer cpu+motherboard at a very reasonable prices.
Two real cpu are better than hypertreading!
The ironic part is that Intel can’t just make a version of Itanium with less cache.
Itanium is a more powerful processor with a better architecture than any Pentium. It costs no more than current Pentium 4s to make; notice, for example, how Itaniums are price-competitive with Xeons with as much cache. And though Itanium is slower than x86 processors on 32-bit apps, which it runs in emulation, keep in mind critical mass; if Itanium was mass-deployed in desktop computers, pretty soon all major apps would have native recompiles, and would run far faster.
As long as Itanium is confined to servers, there will be no critical mass. And we’ll all be stuck using inferior x86. Itanium may not be perfect, and it may not hold a candle to Alpha, but it’s far better than the legacy piece of crap we call x86. And remember – the Alpha people are working for Intel now.
AMD is aware that:
a) The desktop computing is fading in the US. It won´t recover when – and even *if* – the American economy recovers.
b) IBM and the Chinese Red Empire are coming to desktop computing.
c) IBM and the Chinese dictatorship have WAY TOO MORE money than AMD.
d) There´s a whole world out there
Maybe… less America, more the rest of the world?
This is just to contradict Eugenia in her 3rd post
2800+ at newegg
>This is just to contradict Eugenia in her 3rd post
General availability NO. Not all of the OEM resellers got the 2800+ CPU. AMD *IS IN SHORTAGE* of this Mhz CPUs. You will find the CPU available on the biggest retailers (like the newegg you linked), but the smaller retailers, or on smaller countries the CPU is not available.
currently intel makes use of dualchannel rambus and pc2700 for a whopping 5.7GB/s bandwidth.
DDR 400 is coming.
clawhammer’s built-in memory channel controller is Single channel, 64-bit controller is only DDR333/pc2700.
intel will have 800mhz DDR bandwidths availablesoon.
athlon 64 will not be able to match the P4, and therefore, will be unsuccessful it is alleged niche.
…are simply coming out for consoles now, and I doubt it will be long before the PC becomes shut out of the gaming industry entirely
except strategie-games 😉
I have heard these things since the birth of the nintendo NES (if not before). It will never happen.
Everyone I know uses AMD processors!
Amd is certainly losing the MHz battle with Intel and more importantly, AMD no longer has the money to be able to keep pace with Intel.
AMD is doing the right thing because, despite all the apple is behind posts on this forum, the mass market just doesn’t care any more. Intel’s 3GHz pentium recieved a yawn the 4 Ghz pc will probably put people to sleep. Intel is trying to drive everyone out of the processor business by investing in equipment that no one else can afford. Intel, though, best start getting some more markets (like DSPs and embedded processors) because the economic returns on those new fabs are already pushing the limit in regards to what can be justified.
AMD better look for new markets as well and frankly apple should be one of them so long as apple plays nice and goes for more market share (which will only happen with lower prices). An alliance with red hat might not hurt either.
Don’t compare RAMBUS to DDR. RAMBUS is only a 8 bit bus and is thus 1/4th the speed at the same clock speed. Intel knows full well that the RAMBUS can’t compete against the DDR and thus uses the dual chanel mode system which rases the price. Also, the latest RAMBUS chipsets allow you to run 4 RAMBUS chanels.
I use a Dual PowersPC S-Star system all the time. It uses a standard memory very simular to the DDR; only they use a 256bit data bus (4 DIMMS) to feed the processors. The system is designed to support upto 4 CPUs and much of the IO overhead is handled by dedicated IO processors. Thus with the 4 RAID controls and 2 1000T network controls, the system could have as many as 10 CPUs access memory and fighting for bandwidth. DDR can handle this type of workload; you just need to use more then 1 chanel.
The bigest question is: Can the P4 make use of that bandwidth? A CPU should never use more then about 25% of the memory bandwidth; this allows other devices (DMA …) to use the memory also. This is even more important when running multiply processors. AMD’s chips may only use DDR333 currently; but they have also determined that they can run 8 CPUs with the limited bandwidth supported by it. Thus a 4 way system should only be using about 50% of the bandwidth of the DDR333 memory. I question Intel’s need for a 5.7GB/s bus to run a single processor.
It has no sense. How much power you need for browsing an Internet, reading or writing e-mails, playing MP3 or even programming? 1GHz? Do not exagerate. The GHz race has no sense. There is only a question of smart marketing.
In fact normal people do not need such as fast processors that fry eggs (high end servers perhaps…). Now, the processor market is simmilar to the car one: you can buy Porsche or Ferrari, but quite well you can drive a normal Ford Mondeo. There will be always people who will search for the newest bells and whistlers. So it seems that the market for super fast processors is finished.
(But on the other hand who knows, may be the next generation Microsoft Windows will require 10 GHz only for sturt up…)
I’m not going to go through all the trouble of quoting everything I want to respond to, but here goes:
Nobody knows how well Hammer will perform (either version) when it is released and Intel has to deal with heat issues before they can ramp up the clock speed any higher. They’ll be increasing the speed of the FSB, sure, but for an 800 MHz FSB to become reality, DDR 400 (or DDR-II) must be a viable memory option before Intel will move on it. It isn’t. Not yet.
Preliminary Hammer benchmarks don’t count. You cannot accurately measure speed on silicon which isn’t final yet.
Samples of P4 CPU’s running at high clock speeds such as 4 GHz are just that: samples. It doesn’t mean anything unless they are ready to be brought to market. It’s hype, people. Seriously.
I’m not saying this is a bad decision for AMD, since the market for high-end components accounts for a small percentage of their sales. But saying that Hammer is doomed to be a second-class part is purely conjecture.
Also remember that AMD has been in a similar situation before … just before the K7 was released. What ended up happening? History can repeat itself, and AMD has outpaced Intel before even though they didn’t have nearly as much money or as many resources as Intel had.
As for consoles overtaking PC’s in the games departmnet:
*shudder* Mario Bros. (and its various spinoffs) were cool … WHEN I WAS 13 … but I appreciate games with depth such as hardcore RPG’s — which is sorely lacking on consoles today. Input devices, while mostly sucky on consoles, have very little to do with it.
A perfect way for IBM to resurrect OS/2(since it only has 5% of th PC market anyway)is to write up a 64-bit version of OS/2 for AMD’s latest chip? What have they got to lose?!?
This story is silly and speculative.
AMD isn’t “shifting away” from the PC processor market, as least not in the same sense that Be, Inc. “shifted away” from BeOS.
AMD has bet the _entire_ farm on Hammer. They would have to be a lot worse than shortsighted fools to risk the whole company on developing a chip like that, only to end up producing a late, outdated device for a market they’re no longer targeting. Get real.
AMD is *expanding* “beyond” the PC market, and they are smart to do so… who knows when, and how much, it will ever recover? But meanwhile, they’ve invested a lot in Hammer and they’re mpt going to just let it wither and die.
Just because Forbes.com insinuates a “retreat” by inserting phrases like “outside computers”, “losing battle”, and “a shift away from the slowing personal computer industry”, doesn’t mean that was the intent or even the underlying meaning of AMD’s annoucement.
Take a look at this page:
Those of you claiming that Hammer is stuck with single-channel DDR 266 are sorely mistaken. The chart *clearly* shows built in support for 200, 266, *and* 333 MHz DDR, and “Future memory technology support as it is defined”. That means DDR 400, DDR II, etc. Come on, DDR 400, has *barely* just been _standardized_!
For you *really* stubborn posters (or Intel employees), take a careful look at the bottom item on the chart, “Opteron (sledgehammer-MP)”. See that? “DUAL DDR Channels”. Dual–they alreay have it ready. If the general consumer market demands dual-channel DDR, you can bet they’ll have a commodity product for it.
Not only that, but ALL hammer based designs can use an *external* memory controller, (in the “northbridge”), just like existing Intel/AMD systems. So you can bet that if AMD doesn’t keep up, VIA will have some sort of KT-blahblah chipset with super-quad-pumped DDR III, etc.
And I’m not going to even bother debating SPEC benchmarks for unreleased (or non-existant) processors. You can bet that when Hammer is released, it *will* be competitive. AMD has been more than competitive since the K6-2. Besides, Xeon is 32 bit, and Hammer is 64-bit. Itanium 2 is almost vaporware and let’s remember that it *doesn’t* have the MHz headroom of the Pentium IV design, at least not with it’s heat problems.
AMD said Hammer will “blow away” Xeon, and I don’t doubt it. I actually expect it to be quite comptetive with Itanium 2.
Enough OSNews sensationalism for one day…