Become a fan of Slashdot on Facebook


Forgot your password?

Athlon Reviews 179

Since the NDA was lifted early this morning, several sites have released reviews of AMD's new Athlon chip (coming in 500, 550, 600, and 650MHz versions). The first was Bill Henning's CPUReview site. He reviewed the Athlon 600 and has several nice things to say about it. Next up is The Upgrade Center's review, and two more submitted by kimmo, the first at Ace's Hardware, and the second at AGN Hardware. Next, Magnetism submitted a link to Tom's review. Finally, as submitted by pmmay, the ZDNet review. To finish, an article at the SJ Mercury that discusses AMD's strategy for the chip market (thanks to Greg Miller for that one). Update: 08/09 12:31 by J : Thanks to The Evil Dwarf from Hell for links to the AMD Benchmark Page, which even has SPECint and SPECfp scores, and to an anonymous reader for the Ars Technica review.
This discussion has been archived. No new comments can be posted.

Athlon Reviews

Comments Filter:
  • check it []

    wherever your socks go (you know, the ones that don't come out of the dryer), I bet you'll find a whole lotta HTML tags that disappear from comments.

  • Again you forgot the Geramn Computer Mag c't. They have a very positive review (Athlon's gonna kick some butf) and very good benchmarks. Check it out here [].
    Oh and do not forgot to pop in your babelfish - it is in German ;-)
  • Finally, I can have a really high end x86 system that has no wintel parts. The Prophecy is complete.
  • The appropriate c't article via Babelfish (German->English) is here:

    Athlon article []

    Unfortunately Babelfish translates only half of it...
  • I don't really mind the locking at this point, as I'd rather see the K7 core get out there and get proven before people start pushing its clock rate.

    Remember, the P6 core had a good workout in Real Life before it got pushed into common OC-land.

  • It really doesn't make much sense to complain about something you've really no God-given right to do in the first place. I'm sure you can still up the bus speed if you're hell-bent on overclocking a 550 MHz CPU, but, at that speed, who the hell needs to overclock?

    - A.P.

    "One World, One Web, One Program" - Microsoft Promotional Ad

  • Wow. These benches are awesome looking, but as I sit here in awe, I wonder, how much are they gonna cost? AMD has always been known for its great pricing, but now with this new chip, will they stray from their roots? I want to build a system soon, but if the processor is too expensive, it won't be an option for it. Let's hope AMD keeps it reasonable, even with its intel-crushing speed.
  • Yeah, and who'll ever need more than 640k of RAM?

    Or whatever the number was...
  • I'm sure you can still up the bus speed if you're hell-bent on overclocking a 550 MHz CPU, but, at that speed, who the hell needs to overclock?

    You know, this reminds me of those quotes ... like the one that says "640k ought to be enough for anybody" and "there might be a world market for ... about 5 computers" .....

  • It's now a good thing that competition in the x86-cpu area has revived. Unlike video cards and such, the end-user who buys thier computer will have a choice for their cpu. Computer vendors were stuck with selling intel because the end-users only knew/wanted intel. MAybe it will be like the 486 old days. Who knows?
  • Ok I wanted to see them
    They aren't worth the dowload all you see is graphics comparing Athlon and PIII. Basiccly on each graphics the Athlon Beats the Intels processors.

    The quake Movie is dispointing they should have take shots of the same demo scene running on both Hardware ...

  • >You know, this reminds me of those quotes ... >like the one that says "640k ought to be enough >for anybody."

    I don't see a connection at all, actually. It's not like you can overclock it to 4ghz. There's no apparent requirement for overclocking a machine 100mhz, and if there was-- I'm sure a few hundred dollars would be well worth it.

  • That's nowhere near being exactly the same connection. Conventional memory wasn't exactly a selling point that was changed from one version of dos to another. Constants versus variables.
  • Apple keeps touting the G3 as a "Pentium Toaster", saying it beats the pants off of the PIII at the same clock speed. Now AMD seems to be saying that they can do the same. Are there any benchmarks out there doing G3 vs. K7^H^HAthlon? Methinks that could get interesting....
  • on this page [] there's a link to all known review, with rating, very cool!
  • Finally, I can have a really high end x86 system that has no wintel parts. The Prophecy is complete.

    To me its not the end - its just another bringe on the same road because the athlon is still tied up to the original 80386 design wich was tied up to the 80086 design which was derived from the 4004 (Intels firts production processor).
    I think the x86 line will be dead and the wintel marriage too when another architecture like the ARM [] or the PowerPC [] will make standard machines bought by normal end users .....

  • by cesarb ( 14478 )
    OK, we'll need to do some things to add Athlon support to Linux:

    a) Kernel - we need to add the Athlon to the cpuid lists so it won't come with the CPU type blank

    b) gcc - we will have to edit the, i386.c and i386.h so gcc will know the instruction delays when you use -mcpu=k7 (generates faster code)

    c) gas - we will have to add the new K7 instructions to it

    d) rc5des (offtopic) - we need k7 cores

    e) Xfree86 - we need optimized drivers for K7 (and for other archs, AFAIK we do not have any CPU-specific optimization in the drivers

    Maybe we would need some SMP changes too, and UDMA and AGP support for the new chipsets.

    Any more changes?
  • Do the other benchmark sites have fair benchmarks? I only looked at AMD's site and it seemed as though every benchmark was made with the disclaimer, "This motherboard is not commercially available." Obviously, AMD wants to show its processor in the best light, but it would be nice to see comparisons of computers real people can build, not just the engineers at AMD.

  • Usually people will pay much more money for even a small increase in performance. That's why I am going to buy the Athlon 650. And that is why clustering is so hot.
  • <flamebai>Ugh. Get a new religion</flamebait>

    Well for starters, the G3 can't get clock speeds as high as the PIII. Second of all, of course Apple would say that. You think they'd say "yeah, we suck alright." ? And it's only faster at one lousy thing, bytemark tests, for those of you who believe in bytemarks. Non-Macophiles the world around and even some of them know the benchmark means nothing in the real world.

    <flamebait>And now this is where I've managed to hit some conditioned reflex and Mac users will start coming out of nowhere to defend their precious lifestyle. Go cool out, I'm not dissing on Jesus or Muhammed. Don't give me any of that RISC vs CISC crap.</flamebait>


  • hmm if this is Intel IOAPIC then maybe I can get a DUAL AMD motherboard soon.. 2 600Mhz CPU's shure would scream.. it would be nice to see kernel compiles in a minute or two
  • "Well for starters, the G3 can't get clock speeds as high as the PIII."

    Actually, I'm not really impressed by the G3 and its "steamrolling pentium toasting" glory. Nevertheless, clock speeds mean absolutely nothing out of context. Now if the typical G3 instruction takes as many cycles as an average PIII instruction, then the statement that they have lower clock speeds would be significant. whatever ;p
  • "OK, we'll need to do some things to add Athlon support to Linux:"

    I'm not a FSF bigot, but shouldn't this really be "support GNU" or "support POSIX" or "support UNIX"? but whatever.

    Please please please make Linux, FreeBSD, FOOnix, KILLER on AMD chips. I am currently a pathetic windows junkie (although I've gone cold turkey on Linux before...mostly I'm lazy and want to play my old games). If *nix is killer on AMD, then WHEN I DO switch to AMD (most definately my next chip - Intel bugger off ;) then I can ALSO switch to *nix (yay).

    Make it and I will come.
  • Actually, the easiest and quickest (and most pronounced) change would be to add the Athlon MTRR (Memory Type Range Register, or the 'fastvid' thing for some of you ... :) control code into the kernel. The other optimizations can all be done with compiler tweaks (as you suggest). The MTRR change should be easy though -- apparently the Athlon MTRRs are compatible with the P6 versions. And as for me -- I'll be glad to code it whenever someone gives me an Athlon...
  • People've been saying that for years, if not decades, and it hasn't happened yet. Both the architectures you mention have standard machines for normal end-users out... Corel's Netwinder, in ARM's case, and, as for PPC... well... can you say "Apple Macintosh"?

    The problem is that massive base of x86 application code out there, with which any new platform must be binary-compatible in order to stand a chance. With luck, free-source software will change this, by allowing source-compatibility rather than binary-compatibility to be the key.
  • I don't have any personal experience with an AMD chip, but from friends that have had them, they never really liked then as much as Intel cpu's. Since I heard about the K7, I have been reading every artical I could find on it. What is said on tomshardware is just restating what was said by the firing-squad when they did thier first comparisons of the P3-550 and the K7-600. And then recently with thier P3-600 (and the oc'ed version P3-660) vs the K7-600.

    But the thing is, will big corperations be willing to use/trust another processor maker for new servers. With server market being run by Intel, Sun, and DEC, will there be enough room for AMD to pop up and show off its stuff. I personally will not get the K6 untill they implament the .18 micron and have the Viper chipset for it. Another factor for not getting a K7 now, is waiting for the commercial release of a new, faster type of ram that can advantage of the fast EV6 bus better.
  • One of the reviews said that the current MTRR won't work (they had to disable it). They also said AMD had contacted the MTRR maintainer (!), so the support will probably be added soon.
  • I haven't looked, but if the Quake 2 benchmarks are in seconds, then lower is better. The typical method for determining Quake2/3 speed is in FPS (Frames per second). The idea is to run a demo and get the highest FPS (or least amount of seconds). Frames per second is an equation of how many seconds it takes to run the demo. The AMD Athlon wins in the Quake2 benchmarks.
  • The K7 200mhz system bus (with support for 400mhz on the way) and its mesh design could result in absolute Intel thrashing x86 SMP boxes with 4+ CPUs. There are tons of "ifs" involved (if they keep the money flowing, if the bigs boys wanna try it, etc.), but the core architecture is a huge leap forward for x86 SMP.

  • Try one of these:
    LinuxPPC []
    Yellow Dog Linux []

    As far as I can tell (I haven't used either), LinuxPPC is a general purpose distribution, along the same lines as Red Hat, Open Linux or Debian GNU/Linux. Yellow Dog is more targetted for the server market.

  • The K7/Athlon can be overclocked. Check out the Kryotech info towards the bottom of this site:
    AGN Hardware: Athlon Review []

    They're looking at having an Athlon system running at 1GHz by November. My take is AMD has no problem with overclocking, as long as it's clearly labelled as such.

  • Any time you can get an increase in CPU speed directly proportional to the increase in cost, you should jump on it (as in this case). Typically, high end processors demonstrate hard core diminishing marginal rates of return. Thus, the ten percent increase in performance you get by stepping up to the latest Xeon might increase your outlay by 50+%. Those $13,000 PA-RISC workstations, to take a different example, are NOT ten times as fast as a solid $1300 system. They'd be lucky if they could clock in at twice the speed. There's simply no way to say: "Ok, you can buy 1 point of SPECInt for $20."
  • But Intel still has the monopoly on the APIC standard. The APIC standard is what Linux uses to parallelize the processors. AMD implements another standard. The fastest AMD will capture slower frame rates than a dual Pentium III.
  • > rc5des (offtopic) - we need k7 cores

    On a related note, has anyone seen rc5/des benchmarks in any of the reviews? Though I've read many of the reviews, because there're so many, I haven't had time to read them all...

    I know that there were some scores rumored to be about 1.6 Mk/sec, I'd prefer some verifiable numbers :).

    Alex Bischoff

  • Seems that Tom's Hardware's ad server is down.. and of course in IE that when the ad fails to load after the page it flips to a stupid IE error message instead of just not displaying the ad. The only fix is to be quick on the stop button.. FUN...

    Anyway... I am going to build a machine this month with a athalon 500 ... wish me luck... I cant wait...!!!!

  • -- If 7-11 is open 24 hours a day 7 days a week, why do they have locks on the doors?

    in case the person running the place has to take a wiz
  • Don't you really mean it will be like the NEC V-20 old days?

    The '486 is relatively new.
  • a) there are no commercially available athlon motherboards

    b) there are no commercially available athlons

    the reason these reviews just all showed up is that they are old (well, not 1-2 days old) reviews that can only just now be printed due to the AMD NDA being lifted. the systems that these people are reviewing are the AMD (essentially reference) motherboards, with processors still warm from the chip fab.
  • I'm just glad to see that Intel has got a real competitor (by the looks of things, anyway) for higher end servers and workstations. This is like a wake up call to them. They are having a helluva time with RAMBUS and Merced has been delayed forever. The Athlon/K7 is very scalable (again, by the looks of things) and it looks like the Camino will not be enough to catch up to it.

    I don't know what will come of this, but I do know consumers are going to be winning big. We're going to have some cheap CPU's that are very good in the near future. Games are (hopefully) going to take advatange of this power. Maybe we'll have some on Linux too. I would just love to be able to put together a great system that does not involve two companies I simply do not like very much.

    Oh yes, and it was nice to see the Linux Benchmarks on Ars :-)

  • Never said they'd have to. "Free source" doesn't mean "no binaries". What it means is that people who do know enough to install packages from source can install them on their platform of choice, making tweaks as necessary to get them to compile... and then roll the results up into an binary RPM or some such that ordinary users can learn to install.
  • Yes there are some of them in some of the reviews, and it has IIRC 1+Mk/sec (while the K6 is at ~ 0.5Mk/sec)
  • Heck, if it becomes fast enough, maybe you can compile two kernels at once! Or four! It's such an exciting prospect. End users all over the planet will discover how fast they can compile their kernel and their daily productivity will soar.
  • All good stuff, mind you, but I can't help but remember back when I bought an Atari 600. It came on the market bigtime. The thing is, the project was cancelled a few months later. So all that was ever avaiable for the machine was the keyboard/cpu thing. No software, no peripherals, no support of any kind.

    That could happen again with the K7. The CPU by itself is just a fancy piece of metal and ceramic (plastic??). Since this isn't a plug-in replacement, it will require chipset and motherboard support. It will require foundry capacity to produce it. Intel has monstrous foundry capacity to compete with. Some would even say their production capacity is one of Intel's main strengths.

    It's important to remember that David required backup from God to take down Goliath. AMD doesn't have anything like that behind them.

  • The review here [] says that they plan to add a marketing department. If you remember, that's what brought down NirvanaCo, Dilber's employer-for-a-show. Hope they know what they're doing :)
  • Reading at the end of the Tom's article there have been heat buildup problems and the related stability problems on the P3-600. Suprising? Not to me.. Intel is worried... the pushed the core of the 600 up .05v (from 2.00v to 2.05v) isn't that the trick i did to my old 586's and ppro's when i upped the core clock? Wow.. intel using overclocker tricks. Have they run out of stuff from R&D to get their cpu's running faster? This is tight... I have a P3-500 running freeBSD as my main e-mail/dns server and it has occaisional hiccups. Hope to see better form the athalon when I get one in production. WOOHOO
  • Some of us like the CPU to run correctly and meet the published specifications. You can't guarantee that with an overclocked CPU unless you have a multi-million dollar chip testing machine and the appropriate test software. In the real world, you don't want some luser overclocking his computer, resulting in a system that crashes or produces questionable results.
  • There is but one true benchmark--Quake.
  • RAM and overclocking aren't similar. Saying "who needs to overclock when you have 550mhz?" is similar to saying "who will ever need more than 640k of RAM?" in that a reasonable answer to either question is "In a few years, you will."
  • It's not multiplier locked, however it isn't as simple as changing a jumper on the motherboard anymore (and isn't as easy for chip remarkers to relabel a chip either).

    In order to OC the sucker you have to take the processor case off and figure out how to change the settings via the goldfinger pins.

    No, I don't know how to do this. But I'm sure some enterprising individual will figure it out in the near future and let the world know ;).

  • I have a difficult time accepting Apple's tripe concerning G3/4 performance, as they only compare using ByteMark, an irrellevant benchmark, and magazines tend to compare using (mainly / only?) Photoshop tests - which is better optimized and originally written for Macs.

    Good luck finding spec marks on Apple machines (spec is a bit setup dependent - OS, caches, compiler, RAM etc), I haven't seen one in a loooong time. Motorola puts them out for the PPC chips and they seem to match PII/IIIs ok, but not anything significant to write home about.

    I've heard that the same program (Lightwave?) run natively on Celeron 300 vs. native G3 400 nets a big margin favoring the Celeron (cache speed a huge help?).

    It does seem interesting that the 21264 Alphas deck even the K7's 2x on spec(Int|FP) same clock, so apparently there is still room for improvement all around.
  • Well, you can already see current prices on pricewatch.. Chips should settle down eventually to just above their lot-of-1000 prices, which range from $249 (500mhz) to $849 (650 mhz)
  • I came across a Web site quite recently which compared different platforms - all running Linux. In that comparism a 300 MHz G3 was almost as fast as a Pentium II or III with 400 MHz. So maybe Apple's "twice as fast" may not be that correct, but 30% faster is not that bad either. If I just could remember the URL...

    Talking about new generatios of processors: Intel only relys on higher clock speeds - a P III at 150 MHz would not be faster than a Pentium Pro at the same speed - it would be slower. A G3 at the same clock speed as a PPC 601 or 604 should be 20 to 40 % faster. That's what I call development...
  • The current mobile P2 has more than 22million transistors (for the on-die 256K L2 cache. Sound familiar?)
  • Actually, this makes it alot harder to remark chips... Before, all they needed to do what slap a new label on the chip.

    If it can be done by software:
    Now they'll have to put it on a board, boot, run some software, turn off the power, take it out of the board, then relabel.

    If it has to be done via the goldfinger connector:
    Now they'll have to take off the cpu case, put a dongle on the goldfinger connector, put the case back on, then relabel. And if you ever take off the case you can tell it was not running at the rated speed.

    There are dozens of other variations that I can think of on what has to be done, but regardless, it just increased the amount of time needed to remark processors exponentially.

    For the person who wants to be on the bleeding edge of speed (and as a result) is always monkeying with the internals of their machine, taking the case off of the cpu and putting it back on is nearly trivial...IMO.
  • Mac always gives benchmarks in Bytemarks. Most other benchmarks do not agree with the Bytemarks, including more real-world tests.
  • I saw a version of the "Look Athalon Kicks Butt" article at Netcenter []. I noticed the following in the article:

    [Referring to Intel's counter to the Athalon]
    "A 700-MHz version is due in the fourth quarter, while Intel is planning 667-MHz and faster versions of the Pentium III."

    I find it particularly interesting how Intel convieniently now rounds UP the .666_ instead of the canoical rounding down. Fear the marketing of a 666Mhz chip!

  • I measured 1.4M keys/sec; it's in my review.
  • I can't remember one time Atari did something right under Tramiel control.

    From the 800 to the 512ST to the TT030 to the Falcon'030 to the Jaguar -- all were awesome pieces of hardware for the time they were introduced, but they went nowhere due to the way the Tramiels managed the company; there was no "drive" in the company. It was more of a hobby for them than a business.

    By comparison, AMD has the drive and the awesome hardware to make a run for the "title" as it may be. Intel is actually helping AMD as well, in it's attempts to own the chipset market.
  • First International Computer (FIC), who traditionally uses VIA chipsets has this Slot A [] motherboard available. At least the specs and a picture for it...
    Looks like it beats a good deal of ass :)
  • And how the people creating the site would be
    earning money?. There's no such a thing as a free
  • The transistor count is a result of the integration of a large amount of cache onto the chip itself, not "wasteful" engineering. We're adding memory to the cpu itself here.
  • A G3 at the same clock speed as a PPC 601 or 604 should be 20 to 40 % faster.

    Uhh, not quite. The G3 used the 603's FPU. For floating point operations, the 604 remains the best PowerPC. The G4 will use a FPU based on the 604 along with all the advantages of the G3 (backside cache, etc). Besides, the G3 lacks SMP support. Apple cannot hope to keep Photoshop users happy for long with single CPU systems when the same amount of money will buy you a dual processor NT machine. It's all about what get's the job done.

  • After reading several reviews of the AMD Athlon 600/650 MHz CPU, AMD has cured the problem of the poor FPU unit that plagued the K6 series and then some! (^_^)

    The new pipelined FPU which can process THREE MMX, 3DNow!, and FP instructions per CPU cycle will make the Athlon THE CPU of choice for anyone who has to run any program that requires serious FPU performance, things like high-end games, CAD programs, photo-editing programs and illustration programs.

    In short, AMD has left Intel in the dust with the Athlon's awesome FPU unit. When AMD starts to produce Athlon variants with full-speed L2 cache with 1, 2, 4 and 8 MB of cache RAM, very likely PC133 and RDRAM support, and SMP support, it'll make even the Pentium III variants based on the "Coppermine" technology obselete.
  • All the guy seems to do nowadays is whine. I guess he's upset that he's no longer the only hardware site out there and that he's no longer the best site out there either. He should set up a separate server for hardware reviews and just rename the current one Plus, his Intel and 3dfx brownnosing got on my nerves last year.

    - A.P.

    "One World, One Web, One Program" - Microsoft Promotional Ad

  • Well, being that they've been taking a loss with all their other CPU's which basically matched intels performance, (due to production glitches, among other things) there's a reason to try to make the money back at the high end. If you can make due with a 500 MHz part, that's great. Most people can. If you absolutely gotta have the most powerful x86 CPU around, possibly excluding the Xeon), then you'll need to pay a permium...

    As for locking the clock mult... That is lame. You should be able to buy any processor, knowing that it's rated to work at that given speed, but if you know and want to take the "risk" of overclocking, that should be your own business... I know - that discussion has taken place a multitude of times already.
  • At present rates, in 3 years we'll be seeing 1.5 - 2 GHz chips. Being able to OC a chip from 500 to 600 MHz will be just about useless. That's not really the argument to use... It's what you want to do NOW that matters...
  • Invest? I already did.

    Get a clue, will you? We aren't talking about an air cooled processor. We're talking about a Kryotech system cooled to -40 degrees C. I'd be damned surprised if they couldn't get it to 1 Ghz by November. It's already running at 650 mhz air cooled.
  • Has anyone looked. For instance, if you have a 650 Mhz Processor with 22 Million transistors, you are bound to draw over 50 watts of power, plus the 200 Mhz main buss. add to that say a 7200K RPM UW SCSI disk, CD-Rom, CD-R, Zip, Floppy, voodoo2/3, sound card, modem, NIC, SCSI card, and multiple fans. for such a system, it seams as though a 300Watt Power supply or more would be needed. also, with all those transistors, especialy the .25 micron varity ought to pump out the heat like a space heater. is the standart case ready to handle all that heat, or do we all need to get SuperMicro 750A cases.
  • The full report please... What compiler? What flags?... And what are the peak numbers? Die size?.. Some people probably think Intel is as good as dead ( ;-) ) - no chance. Why? nobody else can manufacture enough processors! When will dual systems be availible?..

    Oh, almost forgot: I'M GOING TO BUY ONE!


    Has it ever occurred to you that God might be a committee?
  • The chip's bus is 200mhz, not the system bus, thats only 100mhz so your PC100 SDRAM still works with it. And if you'd like to add power sucking space heater like periphrials, don't complain. Some of us know how to use cooling fans and external drives.
  • Unfortunately, how many programs out there RUN under the Alpha CPU at a reasonable price?

    For x86-compatible CPU's, the Athlon wins hands down, especially once the CPU starts supporting larger L2 caches and gets PC133/RDRAM support.

  • but that's one large, steamy pile of bullshit you're spewing there, bub. The fact that the G3 does not reach the same clock speeds as a PIII is in no way indicative of it's performance. The PIII runs at such high clock speeds (and insanely high temps.) because the only way Intel can continue to show performance increases is to keep increasing clock speeds. The G3, on the other hand, executes instructions at a faster rate than the x86 architecture, and is therefore able to get roughly similar performance at lower clock speeds. You are, however, correct in that this issue has nothing to do with RISC vs. CISC, as the Pentium and G3 are both equally close to being RISC architectures as they to anything else.
  • Right now only Quake III (AFAIK) supports SMP mode, in which case a dual P6 (Celery/P2/P3) >=400mhz will beat the Athlon.

    Also the K7 will support the same APIC system as the Alpha, not Intel's APIC system IIRC. Also VIA puportedly is implementing an APIC in their next chipset series even for Intel (they're very nice even with 1 CPU.)

    Now, a bi-athlon (or tri-athlon...) box should be really, really cool.

  • Actually Yellow Dog Linux is essentially LinuxPPC with a little added value in the server area. AFAIK it is 95% LinuxPPC. BTW, LinuxPPC essentially is RedHat 6.
  • The key issue remains the same: Underestimating or dismissing future needs. At some point, a 16 MHz PC/AT had excessive CPU power for most common VisiCalc spreadsheets.
  • Yeah, Intel could drop their prices by 70% and continue making a profit. That's *why* AMD needs the K-7. They can't compete with Intel in the low-end desktop market that they currently hold, because Intel is able to drop their prices to undercut AMD, sell the Celeron at a loss, and make up the difference on overpriced Xeons.

    Enter the K-7... performance that smacks the Xeon, Hz for Hz, plus a higher clock, and more scalable to boot... all with a price tag an order of magnitude lower. The K-7's not aimed at the low-end desktop market, it's aimed to take the high-end x86 server market, where all the money is, away from Intel in one fell swoop. With a little help from Compaq (Compaq == DEC && K7_mb == Alpha_mb), and if the new Dresden fab can keep up with demand, they just might be able to pull it off, too...

    And didn't Kryotech already have a GHz K-7 prototype running? I could have sworn I saw an article about it here a month or two back...
  • *At the same clock speed* the PPC may be 30% faster. However, the G3's maximum clock speed is not as high as the Pentium's, so it evens out...

    Bah. Doesn't matter... Alpha'll smack 'em both around. Faster than the PPC, Hz for Hz, and with a higher maximum clock than the Pentium...
  • ..I still like the idea of having the fastest x86 uniprocessor machine of the moment. And I do not plan to jump in immediately, I will wait for a faster Athlon, maybe a copper version (I think AMD is going to get the technology for copper from Motorola.. ) and at higher frequencies the gap between Intel PIII and K7 will be more notieable.
    However, your reasoning is OK, and I was doing this way for many years.
  • Tell them you can do that, and you would contribute this to the Linux kernel for free, but you would need an Athlon-based PC. I bet they would be very happy to provide you with what you asked, it would be the best spent pocket money ever.
  • Athlon is the ancient Greek noun for struggle or contest, or for the prize given to the victor. (An athletes is a person who participates in the struggle or contest.)

    Appropriate, IMO, considering AMD's ongoing struggle with Intel, and their apparent ability to seize the prize after a long hard struggle. But then the Greeks knew all about the hare and the tortoise, too, if you get my meaning.

  • Go read the Infoworld article people, IBM is ging to introduce a 750 MHz PowerPC processor-based machine for Linux. We've seen what a 133MHz PPC machine can do to a 500MHz P!!!, haven't we now? It looks like Intel's not going to be the processor to have from now on. Assuming IBM has sensible people setting the prices, that is. Remember Big Blue, we aren't universities with the state exchequer standing behind us!

  • Okay, the original post said a G3 could smack down on a P3 at same clock speed. I bet it could. BUT it can't get to the same clock speed. And as it stands, 400mhz isn't enough to topple a 550mhz P3.

    You may of course now debate that it can. I stand by saying it can't, but of course it could if it had a a few hundred more mhz...

    Sorry if I wasn't clear.

  • Hmpf.. it's a 450 MHz PowerPC 750 copper processor. Comments still stand.
  • True, but what I was trying to say here is that the G3 isn't particullarly RISC-like anyway. What with all the new instructions being added (ie. the vector processing unit in the upcoming G4) saying the G3 has a reduced instruction set is stretching the truth somewhat. OTOH, the Pentium series takes the standard CISC-oriented x86 instruction set and translates them into much more RISC-like micro-ops. These micro-ops are fixed length instructions, so Intel can pull off some seriously black magic optimizations like out-of-order execution. Fixed length instructions and out-of-order optimizations are the kind of techniques that are commonly used in RISC architectures.
  • Well, the HPPA 8500 has 140 million transistors :-) (1.5 MiByte L1 (yes level-one) cache).

    Here are some approx numbers:

    Intel: PII was 7.5 M transistors in a .25 micron process and an area of 131 mm^2. The PIII is 9.5 M transistors but I've got no area for it, I believe it's made in the same .25 micron process (not 100% certain) but you can't just scale the area (different types of transistors take a different amount of area, and if the placement is 'suboptimal'..)

    AMD: K6-III 21.3 M transistors and 135 mm^2 in .25 micron process (not same as intels process) and Athlon shouldn't be too different (in that process).

    IBM/Moto (PPC): PPC750 is 6.35 M transistors 40 mm^2 in IBMs .22 micron process and dissipates 5.5 W at 466MHz. IBMs POWER3 is 160 mm^2 in a .2 micron (should this be the same .22 micron as above?) process.

    SUN (SPARC): USII is 5.4 M transistors and 126 mm^2 in some .25 micron process and dissipates 21 W at 400 MHz.

    COMPAQ (ALPHA): 21264 is 15.2 M transistors in a .35 (!) micron process and takes 302 mm^2 and dissipates 72 W at 666 ( ;-) ) MHz

    SGI (MIPS): The R10k was 6.8 M transistors (no numbers for R12k) and took 298 mm^2 and dissipates ca 30 W at 195 MHz in a .35 micron process (this CPU shipped '96, don't have numbers for any later).

    HP (PA-RISC): 140 M transistors and an unknown area and an unknown heat-dissipation. The process is intels .25 micron.

    Larger (in area) processors are usually more expensive to manufacture because the errors are per-area more than per-transistor. And that means that if you make larger chips your 'yield' (percentage of the made chips that works) becomes lower at the same time as you get fewer chips per wafer... Now compare the 21264 and PPC750... The PPC750 designers weren't incompetent - they had other goals than the Alpha designers..

    Intels processes has traditionally been very well tuned with very high yields.

    The raw data is availible here []


    Has it ever occurred to you that God might be a committee?
  • A better term would be Klinux (K* processor + Linux)

    The Linux box I have is a K6/133...guess what I'm asking Santa for? :)

  • Ok, so I did something wrong - here []. The error of my ways was that slashdot no longer likes single quotes around the URL - need to use double quotes..


    Has it ever occurred to you that God might be a committee?

  • Don't forget that Intel still runs the market.* AMD is going to price and clock these things so that they're just a little cheaper and a little faster than Intel.

    *Intel's got business desktops locked up with the Intel Inside marketing program - IIRC, this is the biggest chunk of the market. Even if AMD produced a chip that was 2x the speed and 1/2 the cost, you wouldn't see a Dell "Optiplex" or a Compaq "Deskpro" with an AMD chip until the vendors could get out from under Intel.

    Furthermore, I don't see much general corporate demand for the new CPUs. They've already got enough CPU for MS Office and Win2000 and standard businessy stuff, so cheaper chips are going to rule the day.

  • I thought people bought Xeons to get the big cache, which is where the big performance gains come for certain types of operations (databases, for example). The K-7, as it stands, doesn't really hold up to the Xeon.

    Admittedly, Intel forces folks who don't need the cache to buy Xeons, because the normal PIII only goes 2-way. So there's some market there.
  • by jab ( 9153 )
    Windows style software administration and maintenance is an expensive headache. Try debian and see what you are missing.
  • Xeon cache goes up to 4M, no? Current K7s don't have it yet, but I from what I've heard, the serious server version will have up to 8M... They probably need the new process for those, though...
  • with a name like detrius --I'm guessing yes.
  • Your right on both points. The pentium is severely hampered by the need to remain backwards compatibale with processors from fifteen years ago. It's spectacularly large, has a heat sink bigger than my AC, and runs so hot you could fry an egg on it. I love my G3, and can't until the day I can get linux fired up on a G4 (although I've been toying with the idea of going with the dual celeron system from
  • Xeon cache goes up to 2MB. And only on the 500MHz part and below.

    The Xeon III 550 only ships with 512 kB, so far.

    Of course, by the time those Athlons with up to 8MB L2 start shipping (prolly not until Q2--these will be on a copper .18u process, as will all K7's starting around Dec/Jan if all goes well), this will have changed. But the 2MB limit on Xeons stays, as far as I've heard...probably all the way up until Foster (the Xeon of Willamettes) comes out, oh, say, early 2001.

    Of course, if Intel finds that its butt is being kicked around in the server market, they'll probably try to get bigger caches on the Xeons, too. The main reason against it is that it's damn hard to fab such a huuuuuuge cache and have it run fast enough (full clock speed for Xeons, and probably for the server Athlons too)--and it would be surprising that AMD would be that far ahead of Intel on a pure manufacturing issue.

    [On the other hand, AMD will be able to crank out K7s at higher MHz than P3s, even with possibly worse fabs, because the K7 design is more superpipelined than the P3, especially in the FPU.]
  • A bigger box of Klinux?
  • As has been pointed out, the "Pentium Toaster" ads only used the Bytemark benchmark, which is extraordinarily old and has very little relevance to the sorts of things CPUs do these days. For one thing, it includes no floating point tests at all, IIRC--and these days, most things the average user (i.e. no compiling) runs into that'll tax his/her CPU are floating point dependant. And furthermore, (also pointed out before), the MacOS hampers performance considerably. And if you want to do any sort of multitasking at all, it hampers it hilariously. Obscenely, even. Check out this article [] at the always impressive Ars Techinca for some ROFL confirmation. To be fair, this was benched before OS 8.6, which allows (gasp!) multithreading...but if I understand correctly, apps need to be rewritten to take advantage of it anyways.

    As for real, cross-platform, general CPU benchmarks, there's pretty much only SPEC, limited as it is. Yes, to some degree it depends on issues of compilers, chipsets, RAM, etc. But it's good enough to at least be relevant.

    Apparently, as far as SPEC95 goes, the G3 is about 14% faster/MHz than a P3 in SPECint, and about 9% slower/MHz than a P3 in SPECfp. Course, the G3 doesn't come anywhere near close to the P3 or K7 in MHz terms anyways.

    And double of course, what really matters is app performance. Here, assuming one stays with the MacOS, we run into some serious problems. Essentially, ClarisWorks (now AppleWorks?) was way more optimized for Mac than PC (duh), and it showed. Photoshop is probably equally optimized for both--and no, contrary to what you've heard, it isn't necessarily faster on the Mac. Look a bit further [] in the Ars article above: turns out that while the Mac wins the Gaussian blur test w/64 megs loses with 128 megs on a 100MB loses on the lighting effects (FP intensive) tests...and, this is the big one, it takes 3 times longer to load the 100 MB TIFF in the first place. Woops. And as for, say, anything made by the Microsoft corporation, don't even bother: to say it's better optimized for PC is the understatement of the year. IIRC, MacOffice97 worked by just porting the relevant Win95 API's and keeping the program itself the same. MacOffice98 might be better...but not by too much.

    Obviously, none of the above applies to K7 vs. P3 discussions--except, of course, for 3DNow(!) stuff, but by now most all video card drivers etc. are quite well optimized for 3DNow, and with AMD having the fastest chip on the market, that'll only improve. In any case, just read all the K7 reviews above, and you'll see that this thing doesn't just chew up the P3 in one or two CPU whups it handily in just about everything. Synthetic benchmarks, Winstone, games, encoding, rendering, you name it. And this is before apps are optimized for it (new 3DNow instructions; 3-issue FPU unit, etc. all could benefit from optimizations).

    [Note: from here on out, I'm pretty much talking out the ass of this page here [] on JC's News. Dunno how accurate it all is, but JC generally knows quite a bit about what he's talking about. And I've read some other stuff that backs him up.]

    Now about the G4...well, it seems that the design goals of the G4 were to get SPECint 20 and SPECfp 20 @450MHz (I've heard this elsewhere, although I don't recall a mention of 450 specifically)--implying that it will run at around 450 on introduction. Now, the K7 at 600 beats both of those marks handily, and indeed if you assume, as JC does, that SPEC scales linearly (course it doesn't, but...) then the G4 is just a hair slower at SPECint (and exactly the same at 500MHz as the G3. Anyone else out there know if the G3 and G4 have exactly the same integer unit?), and a bit faster at SPECfp. Note that I'm not sure if he's using old guesses at the K7's SPEC marks, or real numbers...and I'm too lazy to figure it out right now.

    Now, of course the target goals for the G4 were made back when they said it'd be coming out...well, by now. Instead it's going to ship in "Q3 1999"--where Q3 1999 is read, "January." So we can expect higher MHz on intro than 450.

    Of course, by then, the Athlon'll be at 750 at least. Probably 850. Rumor has it 1GHz. We'll see. In any case, JC goes ahead and pits a projected G4-550 against a (projected?) K7-750...and guess what, the K7 is a hell of a lot faster.

    On the *other* hand, the real wild card in all of this is the G4's AltiVec vector processing unit (for those who don't know, vector processing is the sort of thing Crays (used to?) use--very very good at many things that normal CPUs use floating point to do). On paper, it totally totally kicks ass. Like orders of magnitude faster than SSE/3DNow. And from what I hear, it'll be way easier to optimize for than SSE/3DNow, and waaaaaay easier than MMX (which required assembly programming)--i.e., it might just require a recompile with the "optimize for AltiVec" box checked.

    On the other other hand, with the recent emphasis on nonupgradable machines (with comparitively poor 3D acceleration) in their consumer line, and a reported general lack of attention to gaming among Apple bigwigs (course, this was in some ZDnet story, [] so who knows if it's true), the amazing power of AltiVec might only end up being used in embedded DSP machines by Motorola and IBM.

    On the fourth hand, if I had an iBook I could surf the internet while I was in the bathroom. Draw your own conclusions.
  • It's called Coppermine but will not use copper, still aluminium.. The name is from a river like all other intel CPU codenames. Why they got to that river just now when everybody except intel are talking about their new copper based process?



    Has it ever occurred to you that God might be a committee?
  • Apparently the tech docs for the K7 say that there is support for 16mb of L2 cache.
  • Not nobody... About 1.5-2% ( in case of some ads
    up to 6-7% - really).. And anyway, ads are keeping most sites free.
  • If you look at a diagram of the layout, you'll see that a large portion of the transistors and chip area is in the L1 cache, all 128K of it. High speed L1 cache is just area-consuming and difficult to make, but it's an absolute must for scaling and high speeds. Really, the only difference between a decoupled x86 design (like K6, P6, and K7) and a "true RISC" (like there are any of *those*, laugh) design is the extra decoding and retirement logic. Internally, they're all just high-powered RISC-like machines.

    Saying "and it [an Athlon] only just beats a PIII" is really quite wrong -- in some areas, it completely dusts a PIII ... and in fact, in a lot of areas it dusts everything shy of the HP 8500 and the Compaq Alpha 21264. Perhaps you're referring to per-clock performance -- which is irrelevant, since if you can't (or won't) get to higher clocks, it doesn't matter if you can do twice as much at half the clock, at least not in a performance contest. The Athlon doesn't have all those transistors for no particular reason -- it has them because the design team focused on performance, somewhat at the expense of the compactness of the design. History (see Moore's Law) would say that they are correct in this decision ... the process will shrink to accomodate more transistors and lower the price, power consumption, etc. much faster than spending the extra design time. As the Athlon shows (beating the P3 in integer performance per-clock, and almost doubling the P3's double-precision floating-point performance, among other things), the x86 architecture has a fair bit of headroom left ... there are still many design tricks to pursue, as there are in all the other architectures "out there".

If all the world's economists were laid end to end, we wouldn't reach a conclusion. -- William Baumol