Slashdot Mirror
AMD Announces A Shift In Focus From PC Processors
slughead writes "Forbes Magazine is reporting that AMD will no longer compete with Intel to make faster, smaller, and more efficient processors. Just as Mac users would be worse off if Windows didn't exist, Intel users will be much worse now that AMD will no longer compete. You see, there's this thing called demand, and when there are no competing products in a market, a good or service will always increase the price to the economic equilibrium, unless forced not to by the state (forget that right now, communists!!). In English: you're going to get less new technology, and higher prices on existing technology." On the other hand, AMD is definitely not exiting the chip business -- they're just trying to branch out from chips for microcomputers.
For those too lazy to click though, here's a sample quote:
It says nothing about "not competing with Intel". What a load of sensationalist crap.
Slashdot.
Tabloid News for nerds.
-- kwashiorkor --
Leaps in Logic
should not be confused with
Jumping to Conclusions.
Please read the article. Note that there is
nothing in it about "no longer competing with
Intel." Notice also that Forbes article came out
of Comdex, which was full of AMD CPU demos
for the PC market...
Exactly what I was going to say. Even the submittor didnt seem to rtfa, its clear as day. to quote:
Ruiz brought out executives and representatives from Gibson Guitar Company, George Lucas' JAK Films and supercomputer company Cray Inc. to illustrate the technology that Sunnyvale, California-based AMD was delivering outside personal computers.
All this article really means is that AMD is not going to let its only horse in the technology race be one in PC Processors, they want to branch out and put their products into as many markets as they can stomach/reach.
I agree with parent, read the fuckin article...
I'm a little tea pot.
When there are no competing products in a market, the door is wide open for competition. As the equilibrium price rises (out of lack of competition), the barrier to entry lowers. As the barrier is lowered, competing firms will surface. These firms will fight it out until one "wins" by forcing the others out of business. Then there are no competitors in the market, and the door is wide open for competition...
"God is a comedian playing to an audience too afraid to laugh." -Voltaire
I share some of the concern regarding the effect on competition with Intel, because there's only so much R&D money to go around in any company. However, let's say that they slow down trying to compete with intel over the very fastest chip on the market. People buy AMD because you get more chip for the money/the same performance at a lower price. If they invest enough to keep themselves just a step behind intel in their fastest chip, but still delivering better value, they might be able to sustain similar profits with substantially lower R&D and other costs. This would still keep the pressure on intel.
> And before you flame me for not reading the article, I didn't read the article.
:-)
And now that I have, you can flame me for replying to myself
The article talked not so much about ditching the CPU business, as partnering with other companies on non-desktop-PC applications--Gibson for digital audio workstations (using the MAGIC network protocol, covered here), JAK Films/ILM for video/storyboarding gear, and Cray for a new Sandia Labs supercomputer--the first two of which look more or less like specialised versions of desktop PCs anyway. So presumably you'll still be able to throw together a 1337 Athlon box for your own use, but they may be treating the Dell/HP/whatever market as a lost cause.
Click here if you just like to click on shit.
AMD will put compatibility ahead of sheer speed. The press release mentions embedded devices, but also demos of 64-bit game and database software. AMD is emphasizing that its 64-bit processor has better backward compatibility than Intel's with 32-bit software, even though its 64-bit mode is slower. This looks to me like a bid for industry support for its x86-64 architecture, hardly a concession of the PC market.
May I remind you that Intel is not exclusively in the chip market either. Intel spread to new concepts in computing years ago and are better of for it (e.g. From their site: Consulting Services, Compilers, Performance Analyzers, Threading Tools, Training Center, LANDesk* Software etc...) While most of these are certainly related to the PC chip industry it is not nearly as narrow as AMD.
In doing what Intel did years ago, they are actually increasing their competitiveness. In fact a quick look at Intel's (INTC) financials confirm just that.
Hats off to AMD. for keeping capitalism and competitiveness alive.
I hate people that dont have a sig
Unless the temperature in your house is regularly exceeding 100F
Let me tell you as someone who knows quite a bit about heating and cooling -- that's not entirely true. One or two degrees in the case can equate to one or two more degrees in the case
Because the inside of a computer is generating heat, that generated heat has to go somewhere in order for the system to stay cool. In most cases that heat is dispersed out of the case. The cooler the ambient temperature, the easier that heat flows out. Quite simply, cooler air takes heat off of a heatsink easier than than warmer air, even if the difference is only a few degrees.
The difference between a 72 degree room and a 75 degree room can be enough to take an otherwise rock solid system and turn it into something that crashes non-stop. Given just a little more heat, It may become too unstable to even post all the time.
"Everything you know is wrong. (And stupid.)"
Moderation Totals: Wrong=2, Stupid=3, Total=5.
I think the market fall was due to their credit rating being dropped. (It's mentioned briefly in the article)
The difference between a 72 degree room and a 75 degree room can be enough to take an otherwise rock solid system and turn it into something that crashes non-stop. Given just a little more heat, It may become too unstable to even post all the time.
You would have to have a notrh-bridge cooler on your proc for this to be true!
This is barely true in extreme circumstances... what you are saying is all those cheap people like me who don't have ac and live in 90deg weather HAVE to have crashing computers? WRONG-O>>> Ambient temperature doesn't have as much to do with the cooling as you people think... Hell... AIR temperature has less to do with your computer's cooling then many other effects...
First off... the bigger is better rule does apply to the heatsink... You can sufficiently cool an AMD (which does run cooler then the new p4 btw) without a fan if you have the proper heatsink... The CPU will ALWAYS run hotter then the ambient temperature... You would have to live in Death Valley or the Sahara to have an ambient temperature equal to the normal operating temp of todays 1ghz+cpu's. This means that the air isn't the problem, but rather how much area is being affected by the air. You can immerse a processore in ice and run it, but it will still overheat even if you keep it continually exposed to fresh ice. This is because there is not enoough surface area touching the ice... here is an experiment to try: take a block of ice and a bag of cubes... place them both in a temp controlled room and see which melts first. The block will keep it's internal temperature for longer because it has a lesser surface area exposed, this is the same principle in proc cooling.
secondly the speed of the air has some affect. Notice how the air feels cooler when you are speeding down the freeway rather then stopped? This is because the standing air will absorb the heat and not conduct it. This is why you can use bubble-paper-stuff for a decent insulator... air doesn't conduct thermal energy very well. If you move the air past a thermal surface at a higher speed, that air will absorb the same thermal energy as if it were stagnant, but it would rapidly be moved out of the way and replaced with air that has not been thermally charged.
now I am tired and hope you can make sense of what I typed... it looked good when I typed it, and I could make heads-and-tails when I previewed it...
Erutangis ym si siht.
"The offer curve is flat, so the one that has the most R&D has a higher return ..."
Please do note the link between flat offer curve and the fact if you have more cash, you can also (and always) have more R&D than your competitors. So you can always have a better product if you want (and you DO want to). So you are almost always (almost = if you miscalculate R&D in one period, you just add extra R&D for the next products) selling better products, have better fabs and lead the market, ad infinitum, untill and if some other technology you did not foresee and cannot buy takes over you.
unfinished: (adj.)
Motorola did the same thing a few years ago and that's why IBM is now Apples only source of PPC.
I bet there are a lot of motherboard engineer that will be looking for work when this happens.
AMD is quoted as saying Opteron has been above 50% yield months ago. Further, Opterons are smaller than Athlons (better circuit design, optimized for SOI process) meaning they have lower manufacturing cost. What makes you think Opterons are going to be priced any higher than P4s or Itanics, if they are cheaper to make than Athlons?
ClawHammers (the low end Opterons, which will be released under the name "Athlon XP 64") will almost surely continue where the old Athlons have left off, in terms of price and market focus.
SledgeHammer, will have more cache and ECC enabled, which makes it more appropriate for high end servers -- this is the one AMD will charge a premium for. But, its just a showcase chip meant to win benchmarks, sell to Cray, and open the market for server CPUs for AMD.
Probably got rejected because it wasn't a dupe from last week. (rimshot)
Thank you, I'll be here all week.
I think that a lot of OEMs using AMD processors in the machines that they retail on the high street are making it fairly well known, through adverts and other bits of marketing, that AMDs give higher performance at lower clock speeds. I saw an ad in the local paper this morning explaining it as "using two hands to type quickly instead of using one hand very fast". A lot of the simple - minded people that I talk to have some grasp of the idea that AMDs do more operations per clock cycle, if not totally understanding it. Either that or they are impressed by my new mobile XP2000+ box running RHL 8................. (sorry, had to brag about that one)
I buy plenty of stuff through pricewatch, and with only one exception have I ever had a problem. If you're scared about the reputation of a company go here: http://www.resellerratings.com/
Check out what other people have said about a certain place you're about to buy something from. Also, pricewatch does have feedback where you can send a complaint about a reseller.
Question everything that you've accepted without thinking.
Knocking a chip and its "mobo", because you bought it for something it was not meant to be used for - well guess the egg is on your face!
The mini-itx works great at what it is meant to do, low-cost, very low noise, low power consumption, low heat, small footprint - hmmm....lovely in the office enviroment. Does Internet (mail, web), does Office programs and when the boss is out it plays DVD's like a dream. Works as a server too, with file and print and the low-end mini-itx "mobo" (5000) does not even use a fan - less mechanical parts to cause server failure. Give it plenty of RAM (which is cheap today) - supports up to 1GB and give it a fast HDD - supports ATA 100/66/33 and there are some great PCI-graphics cards (GeForce too) out there - the little box will spin nicely but quietly along. There is also a 1U server rack case available for the mini-itx, several telecoms use it.
For most users the mini-itx is great value in price and able to meet most computer user demands, but most importantly it gives choice - which is what it is all about.
- Kenzai, Master of the Little Penguin. "Long Live BeOS...ehhh, where is everybody going!?"
AMD has never been a CPU only company, just like Intel isn't a CPU only company. You really only hear about Intel's motherboard chipsets and ethernet controllers, but they do a lot of other stuff. Likewise, while AMD may be relatively new to motherboard chipsets, they've been making ethernet controllers for a long time. You never hear about the other stuff, since it mostly goes into embedded/proprietary/special-purpose stuff, so it seems to the casual observer that both companies are totally dependent on the desktop PC.
Under capitalism man exploits man. Under communism it's the other way around.
eWeek has another write up...
While I'm an avid AMD supporter, I have to disagree with you about stability. I realise I may be just unlucky, but I've had quite a few bad mb boards for AMD desktops and my experiences with it as a server left me with a sour taste initially.
I have a Tyan Thunder K7 based dual MP 1.2 server with a Broadcom based 3com gbit nic. Apparently, if I understand the comments on the lkml, this board at least seems to reorder some pci reads/write (or something). This "confuses the hardware no end", resulting in an unstable system (going down at least once a week).
The tg3 driver had to be modified to adjust for this hardware bug. There's stability for you...
Now it runs rock-solid, but I expected better from Tyan. So the reputation for instability may have some thruth to it.
--
AMD ... said that making semiconductors smaller, cheaper and faster was no longer the key for an effective strategy.
They said it's no longer the key, which doesn't mean they won't invest some resources into smaller/cheaper/faster. I think they are just reacting to a market that is saying "for the most part your last generation of processors were small, cheap and fast enough - I don't need an upgrade right now, thanks".
The reality is that many people with 300MHz machines don't feel compelled to upgrade at all. Least of all to an Athlon XP 2800+. When these people do decide to upgrade to a 1GHz machine, AMD will be in there competing with Intel. It's the high end freaks who will feel AMDs absence. They just aren't a big enough market to justify the expense of developing and producing better high end processors.
Education is a better safeguard of liberty than a standing army.
Edward Everett (1794 - 1865)
Couldn't agree with you more as far as AMD underpricing stuff. While I like the fact that AMD has made ludicrously fast chips available quite inexpensively over the last few years, I've always wondered about their pricing.
/OPINION
I've always viewed Intel's strategy as "price the fastest available chip considerably higher than the rest, and that'll pay for R&D." This seems to have held reasonably true over the years, where the fastest chip out there is more expensive than the second fastest by 50% or more. If people really want that much more performance, then that'll give Intel the financial incentive (and more importantly, the finance) to go out and develop something faster.
AMD in the grand scheme of things, hasn't had a whole lot of experience with ground up development in the x86 arena (i.e. very complicated chips) until recently. (Note that this does not apply to other chips, such as the terrific 29xxx RISC processors they made for embedded control, and then dropped.) They started life in the x86 world as a second source x86 manufacturer. (I imagine IBM wanted more than Intel making 8086's, because Intel was pretty small at the time.) They built x86 chips under license from Intel (i.e. they got the masks from Intel and just ground out copies) up to and including the 386 to my knowledge.
Then Intel marketing really started to turn up the heat and realize that there might be something to that whole PC processor business right around the time the 486's were getting popular. The "Intel Inside" campaign started to take off sometime around then. Also with this came the fact that Intel was well established at this point as a manufacturer, and no longer felt the need to license to second source manufacturers, and AMD was sort of left holding the bag and scrambling to do design in the x86 market, something they hadn't needed to do before.
Right around the same time, Cyrix, who started in the PC biz making the notable "FasMath" math coprocessors (because Intel's weren't terribly fast until the Pentium; more on this later), started to make whole x86 procs themselves. With the 486, doing this design wasn't too bad for AMD and Cyrix because the 486 was mostly a logical extension of the 386 (with an internal math-coproc), and added things like more L1 cache (a few bytes vs. 256 bytes in the 386), higher speeds, some other things I don't remember, and clock multiplying (i.e. running the proc at some speed faster than the system bus). In fact AMD and Cyrix were typically bettering Intel's designs at the time due to this fact. (Cyrix, I think, had a write back cache before anyone else did, and AMD built some crazy fast version of the 486; up to 160mhz if memory serves me right!)
What really started to bring Intel forward was the Pentium. Intel, I think, has always seemed to have a gift for getting a feel for what's coming along in the next 3-4 years, and designing ahead of time to meet it. When the 486 came out with an integrated mathco, everybody was like, "only CAD weenies and the like need those; they just add cost!" Even Intel catered to that crowd and made the 486SX, which shipped (at much lower cost), with the mathco disabled, and later versions had it removed entirely. But that decision proved to be fortuitous in the fact that applications begin to take advantage of that power in the 486's lifetime.
In the Pentium, Intel pushed farther by doing 2 big things to make the chip faster:
1. Making it superscalar (multiple exectition units that could perform operations simultaneously, for the most part)
2. The performance of the mathco really got good in the Pentium
Enter now AMD into the Pentium forum. Now they're beginning to get caught with their pants down. The stopgap solution to counter the performance advantage of the Pentium was just to make really fast versions of the 486. They would do a performance rating of equivalence to a given Pentium. I think the 133mhz version was equivalent to a Pentium 75, and the 160mhz version (anybody have one of these?) was I think equivalent to Pentium 90. After a while, it became obvious that this strategy wasn't going to work. So after much apprehension, AMD rolls out their home brewed competitor to the Pentium, the AMD K5, and in a nutshell, it ends up totally sucking.
In all fairness to AMD, it wasn't that bad a chip (I still have one in a rsync backup machine) but they made some short sighted design considerations, and fixed relatively minor problems while leaving larger ones untouched. Notably, they decided to spend their transistor budget (I'm talking physical transistor count on the chip, not money), which was slightly less than Intel's, but not a whole lot less, on fixing the 'u'/'v' pipeline interdependency problem (one of the pipelines on the original Pentium couldn't support every instruction the other could) and making considerations to make mixed 16/32 bit code (think Windows 3.11 and Windows95) run faster, instead of making their FPU go really fast. They sort of banked on the fact that the FPU performance still wasn't terribly important compared to integer execution speed. Which was probably true when they were developing the chip.
But all of a sudden the internet got big, and people wanted to do really FPU intensive things like 3D gaming, playing MP3's, and digital video. The fact that they had a shitty FPU and the fact that they were manufacturing on a process that was old, and consequently couldn't get the chip to clock as fast as Intel's offerings began to hurt them. The K5 was not going to cut it. Which brings us to the next important part of AMD's corporate strategy, buy your way out of trouble!
Lucky for AMD, a small company called NexGen was working on a comptetitor to Intel's x86's, but their offering, the NexGen 586 (anybody have one of these?) was not doing so well. They were also working on a fairly impressive chip that was basically RISC inside but chomped up x86 instructions into RISC sized bites, sort of like the Pentium Pro (and P2, and P3). This chip was the NexGen 686. AMD liked it so much that they thought buying NexGen out was a good idea, and the NexGen 686 became the AMD K6. Because NexGen had applied a modicum of thought to the design, it turned out to be relatively extensible too; over the course of its life it added a 100mhz bus speed (Super Socket 7) and some onboard L2 cache (64k for the K6-II, 256kb for the K6-III). While it still wasn't as fast as Intel's offerings, AMD could offer it quite a bit less expensively than Intel's stuff. They even managed to get the 3dNow! stuff going and sold a chip with inferior FP performance to the gaming crowd. An impressive feat.
But by the time the 450-500mhz chips were coming out, AMD was once again in trouble. Intel had figured out with the Celeron-A how much faster having an onboard full speed cache could make a proc, and more importantly they could get themselves into a much less expensive design to manufacture (Slot 1 was hella expensive to make). AMD's K6 was beginning to show it's age.
The solution to this, from what I understand, was basically to buy portions of the Alpha processor design (and maybe some members of the design team) to put the Athlon together. This theory seems to hold some water in the fact that the original Athlon (and maybe the later chips too) used the Alpha EV6 bus for I/0, so obviously there must of been some resemblance in design.
What AMD chose to do here is to crank out Athlons as cheap as they possibly could to try to sway consumers, PC manufacturers, and maybe some Wall Street analysts on their side and beat Intel in the raw numbers game. What they should have done was raise the price of their high end chips somewhat and started putting money in the bank for R&D, and get prepared for the time when the Athlon would start to fall behind. We're now beginning to see this, I think, with AMD having to shift away from the clock speed battle (probably a smart move) in the Athlong XP performance rating system. More recently, the setbacks with the Athlon XP 2800+ and faster seem to imply that AMD is starting to reach the limitations of the Athlon design.
OPINION
I wonder if the Clawhammer (and other x86-64 designs) are going to be all that good because, clearly, AMD has less of an R&D budget to play with. My interpretation of their recent annoucement is that it is possibly some spin-doctoring to buffer the fact that when the Hammer comes out, it's not going to be in the same class as Intel's and will not to be able to compete. From a marketing/PR standpoint, and perhaps a relations with investors standpoint, this may make some sense because investors are less likely to have their faith shook if they think that competing with Intel on the desktop market wasn't part of the game anymore, as opposed to x86-64 failing while in open competition with Intel. Customers are also less likely to lose faith with AMD as well, and simply see them once again as a budget manufacturer of x86 chips, as opposed to one that offered better chips.
If AMD does focus less on making fast x86 processors to compete with Intel, the hobbyists will be disappointed, but I applaud AMD for having the business acumen to make that decision and try to stay in business doing what they can do, as opposed to trying to go for broke in a CPU performance "pissing contest" with Intel. In the early to mid '90s, AMD cancelled production of the very good (and reasonably well selling) 29000 series RISC chips because the cost of supporting them (making compilers, etc.) was too high to make any money from the chip. Maybe if AMD had priced the chips just a bit more they could still be in business of making 29xxx devices, and maybe, just maybe, if they had charged $5 or so more for each of their procs, they would have had the R&D funding to go after Intel, and keep up in the clock speed dept.
Perhaps this will teach a lesson to all those wishing to compete. It will also fuel the fire of fanboys on both sides of the equation while those who go for logic and reason will continue to laugh at the stupidity of both sides. Pick what works, not what your 'team' goes for... idiots.
I don't think the Forbes article was telling the whole story. From bits and pieces I've seen around the web, it seems like AMD is just going to get the Athlon scaled up a little more and shift focus to the ClawHammer. Who cares about a 5GHz P4 if it can't run 64-bit apps which might be common in two years? A lot of people who buy computers today thinking they're hot stuff are going to be very pissed off in the near future when they start having compatbility issues, even if software developers do both 32-bit and 64-bit versions of their apps. If Hammer is on time for next year (Yeah, right!), I suspect we'll see mass migration to 64-bit apps by '05. A lot of people probably don't anticipate upgrading for another five years or more.
The culture of the far east doesn't have what it takes to forge ahead into brave unknowns and consistently produce risky and excellent designs. That culture excels at honing, or incrementally improving existing designs and products. It's as if the whole culture was run by committee on every scale.
I don't know. You may be painting with a rather wide brush. Culture varies from place-to-place in Asia. Koreans, for example, are known to be somewhat impulsive and brash. IOW, "risk takers".
Table-ized A.I.