Follow Slashdot stories on Twitter


Forgot your password?

AMD's Secrets Revealed 119

Techman writes "Three days ago AnandTech brought you a glimpse of Intel's plans for 2001. Now they're following up the coverage with AMD's roadmap for the next year and on into 2002. Does AMD have what it takes to continue their incredible winning streak, or will AMD return to the state they were in before their recent success?"
This discussion has been archived. No new comments can be posted.

AMD's Secrets Revealed

Comments Filter:
  • Trust me, if you wanna play Quake3 at 1024 using a 2MB S3 Virge you'll need more than 650 Mhz.
  • by Anonymous Coward on Wednesday November 22, 2000 @08:45AM (#606945)
    AMD have an impressive roadmap on paper. But then anyone can make anything look good on paper (Especially .coms!), so i'm not going to simply say "Good oh, good old AMD thrashing Intel! Yay!"

    AMD have always had poor luck with their own processor designs, and seem to stuck following in the footsteps of Intel when it comes to inovation. They've created some compitition in the short term with their cheaper CPU's, sure, but thats only been in the last year or so.

    The future is a little more murky. Their x86-64 architecture isn't compatible with Intels IA-64 architecture. Aside from the fact that the x86 instruction set is crap anyway, Microsoft have no plans to support x86-64, instead choosing their long standing partnership with Intel and focusing on the IA-64.

    Now, i know that there are people working on Linux, *BSD ports to x86-64 etc. but that is not going to carry AMD into the next decade by any margin. Without industry standards and support for their architecture, they are going to die, pure and simple. We all know how poor Microsoft are at supporting anything other than Intel as well, so don't hold your breath for them (And others) to support two diferent systems.

    At the moment, it's AMD 1, Intel 0. But thats going to change pretty quickly, and thats a fact.

    T. Lee
  • Don't you see where this is a really good thing?

    Intel is going to develop all of these instructions and then AMD can just implement them in one of their future chipsets.

    They arent "betting" a whole lot on these instructions either, its just something for an edge, even a small one which can mean everything in a highly competative world


  • You're bitching about a ~$47 for 600Mhz cpu not being cheap enough when you're using SCSI for your disk subsystem? Are you nuts? ;-)

    With regard to the MB price difference, I can sort of see your point but the delta only looks to be 20-30 dollars... (I'm looking at for general prices and (<--been a satisfied customer there several times) for specifics.).

    (Oh, sorry, the retail box with fan & heatsink Duron at 650Mhz from essencom is a whopping $75... heh heh On pricewatch the lowest price is 47 for a back-o-the-truck-special duron 600)


  • Could be issue latency to account for the frequency domain changes. They may have to throw in an extra clock every N clocks to keep the PLLs happy.

  • Hmmrmm... am I the only one that noticed low end systems supporting AMD processors are expected to give up on AGP in favor of UMA architechtures?

    This will be a great thing for cheap systems, if they can get the bandwidth there. Of course, it probably means we'll be stuck with the video solution that is provided with the chipset (like the blasted CRM graphics on my 5-year old UMA-architechture O2), but it still will be interseting to see how this unfolds.

    I haven't looked at a processor roadmap for a while (heck, since Intel announced Celerons with cache!)... does anyone know if Intel is planning the same core architechture move? I recognize that the i740 graphics and successors were a step in this direction, but are there plans for the future?

    ...More Powerful than Otto Preminger...
  • Um, Itanium []?

    It does run x86 code, but is designed for IA-64 (NOT X86). Curiously, most of the /. crowd is bashing Intel for this because it won't run x86 code as fast as a Pentium 4. The Hammer series from AMD is a 64bit version of x86, with all the problems that traditional x86 chips have. It may work out in the short term, but I would much have a 3rd generation IA-64 bit chip than a 3rd generation Hammer chip.

    I just pray that whithin 10 years Intel releases an IA-64 based chip that is NOT backwards compatible. For God's sake, it's time to cut the cord...

  • we'll see what happens after SUN buys them. I can easily imagine SUN using the SledgeHammer in their servers as opposed to US III. Solaris will run on x86-64 before it runs on ia-64.

    One of the reasons why I think that SUN will buy AMD is because in order to be successful at making processors, you have to sell lots to pay for development, production, new fabs and so on. So how long will SUN keep making their own processors (ps: I know that TI fabs their CPUs but they still develop them)? It would make sense for SUN to acquire AMD because they seem destined to introduce a cheap and good workstation that sells lots. And since the Sledgehammer will have the server goodies (good bus, scallability, tons of cache, etc), it will happily fit in a server line. They have already stated that Solaris will be ported to x86-64.

    Intel and AMD will produce lots of fast processors for cheap. SUN knows that eventually people will see that a x86-64 or ia-64 trounces the US-III with no problem. They will need real performance and a good price. Intel and AMD don't make toys like they used to (except in terms of FP which is due to the x87 isa and not the hardware, and which will be replaced in both ia-64 and x86-64). The best net profits come from hardware support anyway and not development + sales.

    Sorry if what I said was a bit incoherent, there is a lot on my mind at the moment. But I do think that before next summer, SUN will have bought AMD because of what I said and more. The only two unknown variables that I can think of right now are MAJc and SUN's pride towards using the Alpha bus, which I was told but can not say whether or not it's true, which is somewhat not as good as SUN's own.

  • well established 64 bit market.

    So, where can I get one of these 64bit machines for $2K that will run Windows ME?

    I expect Intel/AMD are entering the 64bit market in order to bring it to the home user. That 64bit market is nonexistent, and won't exist until they create it and leave the 32bit home user market behind.

    -- LoonXTall
    "Microsoft is the only vendor as of yet that has a fix...."

  • by Ih8sG8s ( 4112 )
    I seriously don't expect a chipset to be able to set its multiplier to a non-integer setting.

    Chipsets have been capable of doing that for years.

    For example: 100 x 4.5 = 450MHz

  • It's not like there isn't a lack of non-x86 processors out there. If a non-x86 processor was going to kill x86, why wasn't it the Alpha/PowerPC/MIPS?

    This is why I think Intel has just sunk billions into what will become Just Another Chip, the non-x86 Itanium. It will join the Alpha, MIPS, Power, Sparc, and others in a crowded server-chip market.

    AMD, on the other hand, is going to make a fortune on the x86-64 platform, dominating the NT-or-Linux server market running a mix of x86-32 and x86-64 software. Intel will desperately adapt the Pentium V to handle AMD's 64-bit extensions, and probably run afoul of a judiciously-placed AMD patent. AMD will wind up getting royalties on every Pentium V, and eventually overthrow Intel as the #1 chipmaker.
  • Give it a rest. Don't be such a heavy. This is a forum for information exchange. We can't all be experts in every branch of comp sci.

    BTW you missed on the interpretation of developer: the author wasn't referring to software developers, but chip developers/manufacturers.

    So take your own advice and someone please mod this inappropriate rant down.

  • To be honest, I'm not a Troll, just in a bad mood today...besides, quite a lot of my recent attempts at being funny got moderated as being interesting or informative for one reason or another. And I'm not stupid, just awake for a very long time...
  • Penguin Computing already sells AMD Linux boxes. If VA doesn't want to sell you the box you want, Penguin will. They're fast, helpful and ship with a shitload of extras - like a mountain of docs. I highly recommend them.
  • by Anonymous Coward
    Wrong, actually hardware engineers and scopes display data in the tie domain where you use nanoseconds. In most cases everything is based on the a multiple of the PCI clock to start with, which has a 30 nanosecond period. A multiplier by 4 for the FSB gives 7.5 nanoseconds for 133.333 MHz if the crystals were precise enough (they are often on the low side).

    But instead of giving the clock period, marketeers give the frequency because it increases at every generation. It would be hard to sell something with ever lower and lower figures. When you see things like memory chips, they are rated as PC100 or PC133, but actually the chips are marked with the access time. Of course it is the stupidity of Joe user which makes many vendors able to sell 3-3-3 memory when 2-2-2 is much better (and there latency is very important).

  • I think AMD recent success is dues mostly to Intel's woes. One comment I saw recently (can't remember by who) said that "Intel just keeps hitting parked cars." Quite apt really. It's time Intel spent more time developing chips that people want, rather than pushing stuff that people don't want (e.g. Rambus). I for one am pleased that AMD are currently doing well.
  • Microsoft have no plans to support x86-64, instead choosing their long standing partnership with Intel and focusing on the IA-64

    It's weird cause I actually had heard the exact opposite a couple of months ago. Microsoft had prefered the x86-64 because it was more backward compatible with the current CPUs architecture.

    Guess that would have to be double-checked.

    "When I was a little kid my mother told me not to stare into the sun...
  • Cheap, tight, rubish laddie, we transcend even our dreadful rep. You should see my system, cost me about £250($400 I think), and that includes upgrading the mobo once and the CPU twice. (K62-500, 128Mb PC100, 5Gb HDD (3+2), SB128 PCI, Ext v90 Modem, Dual-Head (8Mb RagePro PCI + 32Mb Geforce2 Mx + a spare monitor I don't know what to do with), Dual Boots win98 and Mandrake 7.1)
  • I have a friend who works designing the CAD systems for intel and according to his opinion way too much of Intel's design is done by hand, and by hand I mean on computers but edited by hand, instead of the computer coming up with the design.

    I see where I am unclear, when I say CAD I mean having the computer come up with the logic and the circuitry, I know they use computers to actually design but its humans doing the design, not a program.

  • They already did. Doesn't the 760 chipset support DDR and RDRAM???
  • hahah, mod this up! If you read the article you'll get it
  • Second, if you're building a "low end" processor why the hell are you throwing SCSI at it? If it's low end SCSI doesn't/shouldn't even cross your mind...

    Okay, okay, the thing is I do a lot of CD burning so transfer from the harddrive to the cd writer is the most important step, not raw processing power. I got a scsi drive basically free so instead of spending the extra 70+ whatever dollars on duron, I went celeron and put those 70 bucks into a controller. That's all.

  • by Anonymous Coward
    Its the strong dollar vs. the euro. Durons are manufactured in Dresden, Germany. Not only are they a great value in the U.S. when compared to the celeron, but even more so in Europe. In an AMD system purchased in Europe almost every part needed can be purchased from a company based outside the U.S. (I'm thinking Taiwan here), ignoring the strong dollar problem. In an Intel based system, you've always got that damnable processor priced on the greenback :)
  • You forget that those quartz crystals the clocks are based on don't resonate at a nice, cleanly divisible rate for the sake of our convenience. Because of this neither the period nor the mhz rating is ever entirely precise.

    Engineers can measure and compare all they like but those crystals will never be perfect, which is why the numbers are approximated. Since the difference in rated and actual frequency is negligable it's generally considered acceptable to do so.

    I'm fully aware of vendors' dirty pool in their marketing schemes, but how exactly does that apply to actual frequency calculation?
    Where can the word be found, where can the word resound? Not here, there is not enough silence.
  • Now a little over a year later they've managed to look less like a Geo ("It'll get you there and back, but don't expect too much.") into more of a Chevorlet ("Quality, full featured automobile at a fair price."). Now with their recent development of dual processor capable motherboards they may be approaching something on the lines of a Pontiac ("Good performance, quality machine, low cost."). While Intel is starting to look more and more like an Oldsmobile ("A nice comfortable car for fuddy-duddies.")

    Hey, don't knock Olds...back in the day, they developed most of the new stuff (automatic transmissions, front-wheel drive, etc.) that eventually appeared in other GM divisions' products. Especially in the era of the Rocket V8s, Olds engines were a Good Thing (TM) to have, used to some extent by everybody but Chevy (an Olds engine would've been too expensive to put in a Chevy, but Pontiac, Buick, and Cadillac used a few).

    If I had to compare processors to cars, the P4 would be a Cimarron [] (a J-car from Cadillac is still a J-car) and the Athlon would be more like a 442 [] (lots of power for a not-unreasonable price). :-)

    Don't knock "your father's Oldsmobile." :-) (Hell, mine isn't all that different from his...a '77 vs. his '73.)

  • Didn't the quakes take out memory warehouses, and not chipset fabs?
    Where can the word be found, where can the word resound? Not here, there is not enough silence.
  • I wonder when semi companies will have to switch to alternate materials such as GaAs for chip manufacturing, and I wonder what the cost impact will be. Also, how about the guys making boards, will they eventually have to switch to alternate materials to reduce signal coupling, etc?
  • Intel keeps constantly betting on extra SIMD instructions like SSE, but must've realized already that only a small niche of developers is actually going to support them if not directly payed to optimize their software.

    a). Adding SSE2 was directly responsible for the 79% speedup of P4's SPECfp score over P3/Athlon. The instructions are superior to the X87 instructions which the other ones use. This is "innovation"

    b). AMD is copying SSE2 on the Sledgehammer project.

  • A Celeron system is MUCH cheaper than a Duron system, because an integrated chipset is available for it. AMD didn't even have an integrated chipset on their ROADMAP until a few months ago. So, Intel can afford to charge more for the CPU and still have the lower system price. This is a massive competitive advantage for Intel. Indeed, AMD's share of the sub-$1000 market has deflated from 50% to 5% in the last year.
  • If it wasn't for games, they'd just use Macs, DUH! Go into any major magazine company or such, and see if you can find a PC that is used for design/layout. I bet that the percentage is less than 20%, and that's playing it safe.
  • If AMD's run so much hotter, they should just condone the use of alternatives to air cooling. That would help promote more mainstream acceptance, and make heat just a little bit less of an issue. . .
  • I have just bought my second Nissan 300zx Turbo (sigh) - used one ...

    I care less about a name, so I won't pay premium for BMW if some other company comes up with equally good product for less.

    Returning to the topic of AMD - don't you think that one of the reasons for AMD not being used in the corporate environment is the corporate obsession with DELL, the only Intel-only big PC manufacturer?
  • Um maybe I'm mad but I was under the impresion that Durons were manufactured exclusively in Austin on Aluminium. Dresden only runs at 50% capacity right now and produces T-birds on Copper. If anyone knows better and can tell me where to get Dresden Durons I'd be a very happy man. Durons clock higher then equivalent T-birds due to having more voltage overhead to play with and I'm sure a Duron with copper interconnects would clock higher still. As to why Durons sell better here (I'm in Scotland), search me. Could be that most folks here bought celerons to OC them and Duron is just so much better for that.
  • Hey, that seems to work for Microsoft, and they are the biggest software company on the planet.
  • Perhaps that's true, but all the motherboard + CPU combos in the range you mention had integrated sound and video, which I try to shy away from because 1) they're frequently incompatible with linux, and 2) it's much more difficult to repair, replace, or upgrade those integrated components.
  • Unfortunately, giving up on x86 binary compatability is still considered to big a risk for PC/MB producers.

    Why is it unfortunate? It would be a huge error to waste billions of lines of code written for x86 just in order to get some theoretical improvement in speed.

    And in any case if you really need serious performance why not get an Alpha? Few people do actual computations and for most it is really unimportant whether the architecture is x86 or not.

  • 1 P4 1.5Ghz is faster than 1 1.2 Ghz Palomino. Fortunately, AMD can call in the infantry and add a second Athlon...
  • The biggest problem with the P4 is its extremely long pipeline (I won't mention the girly-man FPU, 'cause SSE2 may fix that). This makes it unsuitable for any application that has hard-to-predict branches. What happens is that the P4's branch predictor can't work so well, and coupled with the immense penalty for a branch mispredict, the P4 dies horribly. This problem is not just related to Word, etc. Compiles take much longer on the P4 compared to PIII, Athlon. And, for all you gamers out there, AI just kills the P4 (those pesky branches again). With modern games having ever more complex AI's, this will be a significant problem. (And for all who point to the P4's Q3 numbers, take a look at the review on Ace's Hardware.) If Athlon continues to scale, and AMD brings Clawhammer out on time, Intel will be in trouble.
  • First off, it isn't really 266MHz, it's merely two 133MHz pipelines. However, I think that the P4 is four 100MHz pipelines, so it's become a convention. Damn.

    How can they claim that both the 200 and 266MHz FSB CPUs are the same speed? That simply violates the FSB x multiplier rule (1100 / 266 = 4.135338345865; 1200 / 266 = 4.511278195489). I seriously don't expect a chipset to be able to set its multiplier to a non-integer setting.

  • Intel keeps constantly betting on extra SIMD instructions like SSE, but must've realized already that only a small niche of developers is actually going to support them if not directly payed to optimize their software.

    I ask myself: why bother? Don't they see that playing their cards on such proprietary instructions is a very bad thing?

    I remember that not a lot of people complained when Intel decided to introduce Slot1, and I still don't know why. It looked perfectly clear that Slot1 wasn't necessary at that time and Intel was doing it to screw AMD and Cyrix over. (Only Intel's processors would be Slot1, and you wouldn't be able to use both on the same motherboard).

    It didn't work quite like they expected, even with their quite large market share. At this moment the situation's even worse, and I can't see a very good future to the ones who decide to fragment the market like this.

    Not only this is bad for Intel, it's bad for us (developers AND consumers) as well. AMD also occasionally releases new instructions, even less successful than Intel's.

    I, as a developer, wouldn't want to write my software 3 different times to make them compatible with (a) old MMX-only machines; (b) AMD's processors and (c) Intel's processors.

    People in general, as consumers, wouldn't get software that's optimized for anything because developers aren't _willing_ to optimize their stuff.

    The bottomline: everyone loses.

    Note that I don't blame Intel for this practice. The new instructions are good for marketing and they NEED to get out because they are intelligent solutions to coding problems. AMD thinks the same way. The problem is we need compatibility.

  • I'll bite []

  • Cooling can be silent it just takes some effort. And the heat isn't due to bad design, it is due to the fact that above-zero conductors tend to have resistance. That is if it is a design flaw, it is inherent in everything eletrical.
  • That happened with MMX, but it can't happen forever.

    AMD implemented MMX because Intel owned the place and because there were absolutely NO previous DSP instructions in the x86 family.

    Now Intel doesn't own anything more, and AMD wouldn't bother to implement their instructions for 2 reasons:

    a. They already have their equivalent ones.
    b. They don't wish to bow down to Intel in any way.

    I wouldn't downplay SSE's importance either, because _every single article_ mentions it as a point where Intel prevails. On the marketing side, SSE is great, since Intel pays a couple of companies to develop some products with SSE extensions and shows off how fast the processor is. [Since we're discussing marketing, why exactly do you suppose Intel is pushing their processors' clock instead of striving for a better processor/MHz ratio? That's part of the reason]. On the technical side it's also good, but as I've said, it's bound to fail.

  • I need a roadmap to navigate this article.
  • IA-64 [], Intel's 64-bit architecture is not based on x86/IA-32. Starting with the Itanium and going to McKinley. Supposed to be for servers, but so was every other chip. I would guess in a few years we will be seeing quite a few IA-64 desktops....hopefully.
  • I dunno. If they can get x86-64 on the market before IA-64, then people will start buying them even if they just run the same old 32-bit software, as long as it is able to run that software faster than 32-bit processors. That could lead to enough of an installed base that Microsoft will have to consider it. Microsoft has always utterly depended on the legacy. Do you think anyone would bother with Win2000 if it couldn't run any Win9x applications? Would they bother with it if they couldn't try it out on their existing hardware? No way. If x86-64 can get a large enough installed base before IA-64 hits, then AMD will become a part of the legacy that Microsoft needs.

    I think the Microsoft and its herd will follow whoever happens to be the price/performance leader aming 386-compatables. If that turns out to be x86-64, then MS won't hesitate to stab Intel in the back.

  • The core from the Athlon was technology developed by a small company. I forget the name something like BroadBand (thats not it I forget) but basically they acquired that company and then used computer aided design to generate the rest. Unlike Intel, which still does the outdated method of designing by hand.

    okay so they didnt really steal it, aquired it then.

  • The AMD K7 whomps the PIII and in some instances, the P4, while being cheaper.

    The K7 to my knowledge doesn't have a serial number on it that other people can happily query for.

    Whereas it's possible for the P4 to be faster at new software compiled and built with new extensions, etc. The Athlon is faster at what's running now and appears to be a better PIII than the PIII.

    SMP + Slightly upped MHZ = sweet right now.
  • There's also the issue of Rambus, inc's shadily-aquired patent on computer memory. They're suing everyone who makes DDR or DDR chipsets for royalty fees, which is probably putting a crimp in its deployment.
    Where can the word be found, where can the word resound? Not here, there is not enough silence.
  • Your analysis appears to be 2 years behind the times. Oops.

  • Good points.

    Resorting to the worn-out automobile line of analogies, it looks to me like AMD needs to spin off a differently named company to sell the "quality" line of (really AMD) processors. Pull an "Acura" in other words.

  • It's because the Duron isn't that much cheaper than a Thunderbird* of equal frequency in North America. At least two people I know decided to just shell out the extra $15-20 for a Thunderbird* soon after they came out. The price gap is probably even less now for the slower chips.

    * Or "Athlon with enhanced cache", but personally I think that's a long-winded and silly name.
    Where can the word be found, where can the word resound? Not here, there is not enough silence.
  • by AFCArchvile ( 221494 ) on Wednesday November 22, 2000 @08:30AM (#606996)
    Does AMD have what it takes to continue their incredible winning streak, or will AMD return to the state they were in before their recent success?

    It all depends on how soon Intel will release a DDR chipset, and whether it resembles the stability of the 440BX or the follies of the 820. If Intel can make a nice clean DDR chipset compliant with the DDR standards (as they have promised to do), then Intel could have the fastest desktop x86 setup out there. Of course, they already do, but cost is prohibitive and the demon Rambus has not yet been dispatched to the pits of hell. A price drop and a DDR chipset for the P4 could very well be the tolling bell for the Athlon.

  • AMD's site, as well as Anand has been saying that AMD re-released the new Athlons (1, 1.1, and 1.2 ghz) with the new 266Mhz front side bus. My question is, where are they? I'd like to pick one up, but I have no clue where to get a hold of one...not even Pricewatch has any info.
  • AMD steals/buys technology, they haven't had an original design on a product yet.

    Maybe the future will be different but so far AMD has failed to create its own designs.

    AMD has a good plan, take other companies great designs and use them as their own. It works well if you have designs to buy, but its hard creating a company when you dont have the ability to innovate. They got the Athlon design from some small company that they purchased, the name escapes me.

    The p4 will come out in newer revs with faster clockspeeds, DDR support, .13u technology, and copper interconnects and when it does it will kick ass, the NetBurst technology will really help when it gets up to 2Ghz speeds. On the lowpower market we have the low power p3 and on the server market we have the Itanium that will become widely accepted. AMD looks like its in a good position now, but it should go down dramatically in the future years.


  • As with Wal-mart, the Athlon was nice 'cause of the price.


  • by Kryptonomic ( 161792 ) on Wednesday November 22, 2000 @08:32AM (#607000) Homepage
    The article mentions that AMD Duron isn't selling that well in North America, whereas in Europe the sales are "incredible". Why is that?
  • by ahg ( 134088 ) on Wednesday November 22, 2000 @11:02AM (#607001)
    What difference would it make if AMD or Intel introduced a totally new processor (non-x86 compatable) or if you just had a look around at what other options are already available today?

    The PowerPC is very nice processor. IIRC, IBM has released open "PC like" MB designs based on the PowerPC - with an invitation for anyone to manufacture them. Any takers so far?

    Unfortunately, giving up on x86 binary compatability is still considered to big a risk for PC/MB producers. Even us open source OS users recognize that there will be that occasional binary driver or commercial software that we need to run and it's only available for x86.

    Ever see NetBSD on a G4 cube? I did at Comdex/LBE and it's nice, reportedly much faster than equivalent Pentiums but expensive...

    Better stuff is already out there, but it's the old cliche of BetaMax vs. VHS - the product with mass market appeal will continue to reign.
  • there's more to it than that, you have to realize that we make sure EVERY componant on a motherboard and system is FULLY functional in Linux before we aprove it for use.

    If only they tested the damn second serial port on the new 2130's...

    I had to use one of those motherboard serial cables that waste a slot (quite valuable in a 2U). I found one in a box 'o' cables and it worked. Then we told VA (we asked for two serials in the first place) and they sent us a box of them, all different. None of them worked. I can get the winmodem in my notebook working, but not a simple external USR 28800?

    Tested, huh?

    Maybe I'm angry because my thumb is still a little scarred from removing a hard drive from one a couple of weeks ago. Ouch.
  • by DAldredge ( 2353 ) <SlashdotEmail@GMail.Com> on Wednesday November 22, 2000 @08:33AM (#607003) Journal
    When will VA Linux start selling AMD based systems???
  • Going where the market is seems to be a good business move to me. AMD also makes flash RAM which is a good thing to make these days. I don't see them losing their price/performance dominance in the processor market anytime real soon, either.

    I'm also questioning their reasoning in the 64 bit market. However, AMD and Intel are BOTH new players in the well established 64 bit market. I'm looking at their entry into that market more from a perspective that it'll push prices of established 64 bit processors down. AMD would have been smart to have licensed MIPS or Alpha 64 bit technology, which they could have been producing 6 months ago. Sledgehammer's going to have to be pretty damn impressive to convince me to move to it. But the 64 bit playing field is the level one AMD can use to compete against Intel.

    And keep in mind that Microsoft has traditionally not done well in the 64 bit market. Perhaps the Itanium will give them the incentive they need to move over. Right know all we know about Windows for IA64 is that it boots, which it also did on the Alpha (And we all know how well THAT went.) Perhaps AMD's betting that Windows won't do well on IA64 and that sledgehammer will give all those Windows addicts a place to turn.

    Someone had to put all that chaos there!
    ______ "Our 'n about"

  • What if you don't play Quake? You neglect the fact that media apps in consumer space are becoming more and more popular. Low-end 3D renders, video editors, web-media apps will all become popular soon due to the content revolution. Guess what a GeForce2 will do for Premiere? Nothing! For all of the apps mentioned, a fast CPU is absolutely critical. Also, you forget that CPUs have to keep up with the bloat of the OSs. If it weren't for the KDEs, Xs, GNOMEs, Win2Ks, and Office 2000s in the world, much fewer people would need faster CPUs. Win95 killed my 486, and Win2K and Linux killed my PII300. It is impossible for people to resist buying faster CPUs given the fact that the entire software industry is against them.
  • The author seems to be of the opinion that Intel is somehow in the advantage here, but as long as the P4 doesn't carry SMP, the Palomina should come out victoriously on this one. Seems the P4 will be covered with horseshit.
  • "Intel processors aren't defective. They just act that way." by Henry Spencer.

    From the fortune program that came with my Linux distro (red-hat).

  • by rigau ( 122636 )
    Maybe Motorola and AMD need to get together and share some technology. Let's hope that means that Athlons will be cooler and PowerPCs will move beyond 500 Mhz. But see how the universe loves a perverse outcome and how people are ussualy more hard headed about keeping their bad decissions we will probably get an Athlon that cant get beyond 500 Mhz and a PowerPC chip that can cook the thanksgiving turkey in ten minutes flat.
  • so what was your sig anyway?
  • Granted, but IMHO 90% of desktops with 1.2Ghz chips in them sold next year, will be used for 2 things..

    Quake and Email..

    Followed by..
    Napster/Gnutella (hopefully)/ICQ/AIM/etc

    None of which need massive numbercrushing power..

    Followed by a small black dot on the horizon which is all the applications combined that need more than 650Mhz!

    Spend your dimes on RAM/HD/NERF toys..
  • Are you sure you wouldn't want a million Exanode beowulf cluster of 9.99 x 10^(100 x 10^(100 x 10^99999) Exahertz Processors, 2 x 10^1000000 way SMP? That would be a lot better, IMHO.

    Well, yeah... but, don't you think that might be overkill?

  • They're extremely useful for 3D graphics and signal processing and can speed those up quite a bit. I'd still rather see some dedicated signal processors on add-on cards, but that would cause a whole lot of new bus traffic.

    Anyway all those software synthesizers and cool multimedia stuff can really benefit from floating point SIMD, since it can basically speed up common floating point operations by 200-300%. e.g. digital filters usually mostly use multiply and add and that's what SIMD speeds up. SSE2 seems to do to double floats the same that SSE did to single floats.

  • by Nidhogg ( 161640 ) <> on Wednesday November 22, 2000 @08:54AM (#607013) Journal
    Alright I've seen the question answered already. Price, performance, and the ability to meet their demand. This is all true.

    However the question I have is... is either company looking into and/or developing something not based on x86 architecture?

    Isn't there going to be some point where the developers have to sit back and realize that they've done all they can with it?

    I realize that this will potentially cause huge problems both in hardware and software but I think both companies would be doing themselves a favor to start planning for that day now.

  • However the question I have is... is either company looking into and/or developing something not based on x86 architecture?

    Well, let's look at the P4's biggest weapon: SSE2. That is NOT x86. Sure, it's ostensibly x86 because it's in an x86 chip, but that's it. SSE2 will require completely new compilers. The way I see it, when you have to re-write significant portions of a compiler, you've just made a serious architecture change.

    And pay more attention - Intel's Itanium is most *DEFINETLY* not an x86 processor.

    Isn't there going to be some point where the developers have to sit back and realize that they've done all they can with it?

    Okay, most developers won't give a crap - 90% of them don't write compilers or use ASM, they use something on top of that. All you *might* need to worry about is performance. So, to sum up, most developers won't care if their stuff is running on x86 or ia64 or x86-64; so long as they can simply re-compile their application.

    Good lord, don't talk about what you don't know about. And don't insult anybody/anything unless you're damned sure that you're right.


    'Round the firewall,
    Out the modem,
    Through the router,
    Down the wire,
  • First of all, 99% of programmers are idiots who don't know the first thing about computer architecture... they say things like "oh, pointer access... oh if-statement... those just cost 1 unit of processing" No they fucking don't... get rid of those damn pointers and make my job easier for a change... that's where the important optimization etc whatever can be done... babble sorry sleep deprivation...
  • The problem is that x86 compatibility means everything. Asking people to dump their existing software, tools and so on for a technically better CPU is a tough thing to do, especially when you have a competitor selling high performance x86-compatible CPUs.

    My guess is that future processor innovations will retain an x86 emulation or dynamic translation capability so that the huge investments in x86 code don't get lost. AFAIK this is exactly the kind of thing IA64 is supposed to do -- half-limp in x86 mode or fly in native mode (if you believe the marketing...)

    It would seem really unlikely at this point that Intel, AMD or anyone else in the x86 space would jettison that technology for a totally incompatible processor, even if it offered logarithmic increases in performance. In fact, x86 compatibility is so important Transmeta has designed a CPU around emulating it (bit of an overstatement, but you get the idea).

  • I think those bus speeds tend to be irrational so it is 266.666.... And the 0.5-multiplier which has been available for ages is definitely a
    non-integer number

    So that makes it all clear for 1.2GHz tb, but 1100MHz and one gig parts are still strange, as the 1.1GHz would need 0.125 step... well, maybe the new chipset does support those as well.
  • I agree its foolish, SSE instructions have limited use in day to day processing and only appeal to the power gamer crowd, which Intel just barely sneaks by in, I would much rather have my entire system performing than 3 or 4 fps in QIII..

    I just dont see how SSE Is that big of a deal, and AMD can always enahnce SSE (3DNow) etc, its all basically the same thing.


  • by _Quinn ( 44979 ) on Wednesday November 22, 2000 @09:35AM (#607019)
    Oddly enough, the IA-64 will run legacy x86 code in emulation. Which means that even if AMD's x86-64 chips (the Hammer series, IIRC) are slower at the same price than then Intel's IA-64, they'll have the speed advantage for running the your software for a year or two. Quite frankly, the IA-64 is not shaping up to be a desktop chip any time soon anyway -- Intel's releasing the Pentium 4 for a reason.

  • Ever hear of Itanium? IA64? also, AMD is doing sledgehammer, which, while x86 compatible, is getting kind of different from "normal" x86 procs, so I think you can consider that one too... as far as needing to go away from x86, I don't really see the need... wilamette has a trace cache, and once you have one of those, who cares about how long the inital decode takes... you spend all your time executing instructions from the trace cache anyway...
  • So in conclusion, you're saying two things:

    1. There will never be dual-proc P4 motherboards.
    2. You know what a non-existent chip's (the Palomino core has not been released) performance will be.

    Get back to me on that, if you could.
  • 1 P4 1.5Ghz is faster than 1 1.2 Ghz Palomino. Fortunately, AMD can call in the infantry and add a second Athlon...

    Without the use of the silly P4 optimizations that intel is trying to get pushed through, 1x1.2GHz Thunderbird with PC133 SDRAM is faster than 1x1.5GHz Pentium IV with PC800 RDRAM at everything but playing Quake III. It remains to be seen if Whistler will be successfully built with Intel's optimizations; If not, the Pentium IV will not be as successful as Intel may hope.

  • by asolipsist ( 106599 ) on Wednesday November 22, 2000 @09:42AM (#607023)
    "At the moment, it's AMD 1, Intel 0. But thats going to change pretty quickly, and thats a fact." How quickly we forget the past. The short-term future does look like it's shaping up to be AMD's x86-64 vs the Intel IA-64. Most of the posts i've seen favor Intel's chances in this bout. However if you look at any tech industry, you'll notice a startling trend towards jerry-rigging the past (x86-64) instead of adopting some shinny new standard (IA-64). The best and clearest example of this is the IBM PS/2 Microchannel architecture vs the Clones. Obviously the clones won, even though IBM had exponentially more market power. The AMD vs Intel situation is similar, but AMD is in an even better position than the Clones of the 80's in that they are of similar market power to Intel. Both AMD and Intel know this, they are both worried about the future and will weedle as much as possible to advance their respective architectures. In a way Intel is further up the creek than AMD. Intel HAS to outright win with IA-64, its their only hope as far as their stock price evaluation goes (which is inflated, like MS, based on the fact that they have CONTROL) Obviously this situation is nowhere near resolved, its chaotic (highly dependant on minute conditions), and anyone that says they have the definitive opinion (a multi-billion $ question) cleary does not understand the complexity or the historical context involved. The standard is at least 75% of the battle, the one that emerges with the winning standard basically wins, how this happens is anything but simple (think ms, think cisco, think ie, etc...)
  • ("hey! we were the first to 1GHz! It's SO much faster than 900MHz!").

    First, I'll take your capitalized "so" as sarcasm and reply...why yes, yes it is 100MHz. That would be 100 Million cycles a second. That is a lot. lets see that would be 1/9th or about 11.1%. Now lets see when you had a 66MHz cpu and your buddy had a 166MHz cpu you were completely blown away. Sure it's a different ratio, but it is still a sizeable step. It's not like they have produced a 933MHz cpu and claimed it made a difference.

    Second, if you're building a "low end" processor why the hell are you throwing SCSI at it? If it's low end SCSI doesn't/shouldn't even cross your mind.

  • Where the heck were you checking prices? At pricewatch it took me 10 seconds to see that the cheapest "Motherboard w/CPU+Fan Celeron 400" is $149, while "Motherboard w/CPU+Fan Duron 600" starts at $129.
  • Right. My old socket7 motherboard ran the 166mhz chip in it at 2.5 x 66mhz.

    The mhz ratings on chips are approximate, to boot. The only part of the equation that's exact is the multiplier. The bus speed always varies a little bit from its rated value, and even if it were exact you aren't going to get nice clean numbers (as in multiples of 33 or 100) by multiplying bus speeds like 66 or 133mhz.

    The Celeron 333 was actually 66 x 5 = 330mhz for example, and that doesn't even take into account that the bus isn't actually running dead-on 66mhz.

    AFCArchvile, some research before posting maybe? Not that it matters I suppose, people who know even less keep modding you up anyway. :-p
    Where can the word be found, where can the word resound? Not here, there is not enough silence.
  • Man go back to bed and get a little sleep. Star Wars, created by George Lucas first came out in 1978. Battlestar Galactica didn't appear until the like the 1990's and then it was more like a rip off of Star Trek.
  • If it wasnt for Q1/Q2/Q3 does anyone really think these chips would be getting so fast. For the real world, if you want to process 2X as many email, Buy 2 computers! Then you have double redundency as well... Thats why Sun boxen are still the choice for many serious apps, despite having only 450, or 650 Mhz chips... And i read somewhere that the ENTIRE Space station formely known as the ISS runs on a handfull of 386SX chips!
  • Are you sure you wouldn't want a million Exanode beowulf cluster of 9.99 x 10^(100 x 10^(100 x 10^99999) Exahertz Processors, 2 x 10^1000000 way SMP? That would be a lot better, IMHO.
  • The links to tech documents are fake. I can't see any 92% failure rate; the document doesn't exist. Read the links before moderating up some moronic troll.
  • by rkent ( 73434 ) <rkent&post,harvard,edu> on Wednesday November 22, 2000 @08:42AM (#607033)
    Hrmph. I don't know what I think about AMD succeeding because they "executed their roadmap." Most of that stuff was milestone junk that's actually trivial in the long run ("hey! we were the first to 1GHz! It's SO much faster than 900MHz!").

    What I think's impressive is that they're actually meeting their demand. That was always AMD's biggest problem (besides the fact that every product was a direct Intel knockoff ;) - they could never turn out enough chips. Seems like they finally turned that around with the Athlon.

    However, the article did raise at least one important point - for a "low end" processor, the duron's just not cheap enough! I'm assembling my own computer right now, and I chose to go with a celeron because the Duron chips and compatible motherboards were just more, and I preferred to put the extra money into some things that really enhance performance for me, like a SCSI card and more RAM.

    Of course, as soon as DDR is budget priced, I see myself putting together a different system altogether...

  • It was a bit of a paper launch to counteract P4. They should become available soon however.

    266 Mhz front side bus does not make much difference in terms of performance however. Mostly the increas is due to DDR memory.

  • by lizrd ( 69275 ) <adam@bu[ ]us ['mp.' in gap]> on Wednesday November 22, 2000 @08:57AM (#607035) Homepage
    Well, reputation still means a lot in the computer industry. AMD has managed to get itself a reputation for making budget chips. This problem is compounded by PC makers when they select AMD chips. If the PC maker chose the AMD chip because it was lower in price, then that was probably the driving force in selection of other parts as well. Up until a year ago if you were looking at a box that wasn't a genuine "Intel Inside" box chances are that it was a piece of shit.

    Naturally, this applies somewhat differently to the /. kind of crowd. It is possible for someone to build their own box that has an Athlon chip in it that is still a quality machine. I just built my own machine and it has an Athlon 800 chip in it and a real 3Com NIC and a video card with it's own RAM. This however isn't the usual case when you buy a preassembled AMD machine. These are often the budget models that have a winmodem, shared video RAM and all manner of other shortcuts. This reflects badly on AMD and when combined with their lack of support for dual processors rules them out in the high end mass market.

  • Er, were you comparing Celerons and Durons clock-for-clock? Duron's got a 100 MHz FSB vs. 66 for Celerons, so the Duron gives better performance at lower chip speeds when using that more RAM you bought.
  • by lizrd ( 69275 ) <adam@bu[ ]us ['mp.' in gap]> on Wednesday November 22, 2000 @09:56AM (#607052) Homepage
    AMD needs to spin off a differently named company to sell the "quality" line of (really AMD) processors.

    I think that's what they have done to a certian extent. When they launched the Athlon they wisely decided not to refer to it as K7. Not only did they change the name associated with their CPU but it was also the fastest chip on the market when launched. Now a little over a year later they've managed to look less like a Geo ("It'll get you there and back, but don't expect too much.") into more of a Chevorlet ("Quality, full featured automobile at a fair price."). Now with their recent development of dual processor capable motherboards they may be approaching something on the lines of a Pontiac ("Good performance, quality machine, low cost."). While Intel is starting to look more and more like an Oldsmobile ("A nice comfortable car for fuddy-duddies.")

    My apologies to those of you who aren't from the US. I realize that these auto brands aren't exported very often and you might not be all that familiar with them. Also, I apologize for the slogans, it should be quite clear why I am in computer programming rather than advertising.

  • GaAs is much harder to work with then silicon. The cost would be greatly increase by using GaAs. I don't believe anything but very high end systems will ever use GaAs.

  • The only thing that really counts, in the end, is what is your overall performance, and what do you have to pay to get that performance.

    Reviews on anandtech, tom's hardware, and hardocp all indicate that the 1.5 GHZ P4 is essentially equivalent to a 1.0-1.2 GHZ Athlon. A 1GHZ Athlon can be had for under $300 : a 1.5 GHZ P4 is running at a touch over $1000.

    The socket 423 that the P4 fits in is going out by early next year: OTOH, there don't appear to be any plans to phase out socket A any time soon. This leads me to think that the Athlon will be more upgradeable.

    At the very least, you can start with a Duron now and get an Athlon later- the same cannot be said for the celeron/p4 scenario.

    RDRAM is more expensive than SDRAM and has marginal performance benefit. DDR-SDRAM will reduce this benefit to nothing in real world tasks.

    For those of you that buy on brand, buy Intel: if you buy based on real numbers, buy AMD.

  • The IA-64 is a completely differnet CPU.. It's actually one of the most novel ideas I've ever seen in silicon.. This doesn't mean it's practical or will be any faster than a good ole fashioned post-RISC processor.

    The x86 comes in through a translation layer (kind of like the PowerPC's translation of the old 68K ops).

    In fact, Intel changed their entire strategy so that they could make the market ready for the IA-64.. The main issue is that we wanted to have high band-width memory that could work with many many simultaneous memory requests.. We wanted deep pipelines, and speculative out-of-order operations. We wanted to make heavy use of instruction caching.. These are all the sorts of technology Intel has been trying to force feed the industry (See RDRAM).. The reason for this was that Italium (IA-64) is a super-pipelined, multi-way, parallel-op processor that can do all of it's memory prefetching at the beginning of every function call speculatively (meaning that it can even survive accidental illegal-memory references, much like with C++ exeptions.. Fast on the norm, slow on the rarely found case of an exception).

    Intel needed motherboards in place that demonstrated to the public the ability to handle large CPU's with heavy power loads. Multi-path memory structures that work best in a pipelined manner where latency wasn't even an issue (most memory accesses in IA-64 aren't needed for many many instructions).. Additionally, they needed to address the huge CPU to memory disparity. The ability to handle concurrent operation (a la SPARC). In short, Intel has basically taken ideas from just about every CPU manufacturer that I've ever heard of.

    Thankfully, the industry has benifited by some of Intel's exuberance. We've had the war of the memory architectures which has given us DDR-SDRAM. We have massive transistor busses.. We have competition with Alphas and SPARCs (since their turf has officially been shared). We have competing ideas for how a high-end server should look (see AMD's K8 line).

    For better or worse, IA-64 is here today.. Even if nobody buys it 2 years from now. :)

  • by Pink Daisy ( 212796 ) on Wednesday November 22, 2000 @12:09PM (#607057) Homepage
    Resting on their laurels while adressing some problems with the Athlon line. It's not going to give them any new wins, but if Intel doesn't do anything good with the P4, it could work. The P4 certainly isn't going to win anything on its own, but if it gets good compiler support and 3rd party chipsets to drop the price (ie. ditch Rambus and use DDR), it would likely be far superior.

    Second is the 64 bit stuff. x86-64 is one of the lamest ideas to come out of a high tech company since, and everyone who so much as looks at the name knows it. The only way it could go anywhere is if the competition was, what, several years late, having difficulty ramping to high clock speeds, requiring major new compiler technology, and probably many more things as well. Since they've met the first few requirements, I'll hold off on judgement.

    They are counting on Intel to screw up. From the past, this may work. It's not really a good thing to count on. Innovation often fails too, but I hope they don't just go into a holding pattern with the Athlon. That was a big win for them; to do well they have to follow it up with others now.

  • Until someone can put together a 400-way SMP of 9.5 PetaHertz processors, and do it for under $2500, I'll just keep on waiting... of course, I'll need to cryo freeze myself for several hundred years at least, but that's beside the point.

  • by Anonymous Coward on Wednesday November 22, 2000 @08:43AM (#607059)
    Well, call VA, and ask. Know what they will say? "we have some plans to do it, but not right away." They have been saying that for YEARS, back when they were "VA Research" not "VA Linux."

    The truth is, Intel was one of VA's original financial backers, and Intel is still DEEP into VA's pockets. They have an "Intel Deal" and that's about all there is to it.

    VA will tell you, "there's more to it than that, you have to realize that we make sure EVERY componant on a motherboard and system is FULLY functional in Linux before we aprove it for use. We have done a lot of software engineering for Intel systems, and it will take a lot of time to do it for AMD systems." But, that's bullshit. With all of VA's staff and resources, and the whole open source community porting to every bit of hardware under the sun, VA can't do an AMD system... Yea, ok... whatever...

  • Quake? What about 3D Studio, Premier, CoolEdit, Photoshop, etc? Just because all you do is email and Quake, doesn't mean that other people don't find a real use for all this power. Your 'leet Sun systems still cost an arm and a leg. Small time workstation users love the fact that they can get workstation power in a $4000PC. In fact, I'd go so far as to say that the pro workstation market is going to go through a serious shrink af consumer hardware catches up. All ready, the GeForce2 Ultra is faster (in workstation apps, not games) than everything except the WildCat 42XX series cards. Even SGI is using NVIDIA chips (their "VPro" is a modded Quadro) in their low range workstations. Non-PC workstation (low-midrange) manufactuers are either going to have to innovate, or get the hell out of the way.
  • They were working when I checked them out before. And suck failure rates are really not unusual for high end chips (which is why they much more expensive than lowew end chips even though the performance in difference in not that great) Phil Garnier

"I don't believe in sweeping social change being manifested by one person, unless he has an atomic weapon." -- Howard Chaykin