Follow Slashdot stories on Twitter


Forgot your password?
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×

Athlon Benchmarks Out 180

|jasper| writes "on AMD's home page they now have Athlon processors benchmarked against P3 550.. Probably still biased.. but its something " They claim that integer performance is slightly faster, while floating point and 3d is significantly faster (at the same clock speed)
This discussion has been archived. No new comments can be posted.

Athlon Benchmarks Out

Comments Filter:
  • this is incorrect. the 200 MHz bus speed is the chipset-processor bus. the chipset will still have memory bus speeds comparable to Intel systems, so until memory can run faster than Intel chipsets support you won't see much difference there.
  • You really need a P3 or a K7 because they both offer an enhanced internet experience! Surely you wouldn't have noticed that large graphic if you had one!

    (In case there's any doubt, yes, I'm being sarcastic.)

  • Everybody got insulting against the author of this post but I think he has a good point.

    The whole idea of the graphs is for a pictorial display of the results, if they do this they might aswell start at 99%, put p3 at 100% and then the k7 with 150% will look 50 times faster.

    The graphs serve absolutely no good purpose unless they show you the true ratio, and when you have your minimum at 100, max at 150 you can start at 0 without too much wasted space.
  • If I was shopping for a new system.. I'd either get a really cheap and reasonably fast cel or.. really cheap and kickass fast dual cel. The word is making dual cel systems is painless now, with some adapters or something.

    Abit have their BP6 motherboard out shortly (2 PPGA sockets there on the board - plug and play), or you can buy socket adaptors with jumpers to set SMP on (from MSI).

    Intel doesn't seem to like this much - there are rumours now of them revising the celeron to properly disable SMP. If you've been hankering for dual celeron, buy those CPUs now..

  • From what I heard, the boards should look like a standard chipset, from the OS programmer's point of view. So, really no adjustment would be necessary. Someone also noted on the linux-smp list that AMD supposedly said that the SMP support would be compatible (again, from a software standpoint) with the Intel SMP spec. So, in that case Linux would require *zero* modification to run on K7 platforms, SMP variants included.
  • I couldn't agree more. Since 1994 or so the interleaving idea has been used in many situations:
    The pentium FX/HX/VX/TX chipsets required paired chips for (very poor) interleaving... SDRAM had its own dual-circuit interleaving right on the IC so that you didn't have to buy paired RAM anymore... but interleaving 100mhz ram _properly_ (which should be more than possible with an Alpha protocol base) would absolutely increase performance.

    Although the efficiency (latency) is a very important factor in all of this, it seems that a lot of people seem to underestimate the power of raw bandwidth.
    You want an example?
    I've got a PII-350 that I run on a Shuttle 661/P-BX Rev. 1 (also have a Rev. 2 but i prefer the Rev. 1) that is running on a 133mhz bus (467 mhz) VERY reliably. I ran it at 490 (140 bus) on a Rev. 2 board for about 3 hours, but since i had more than one unexplained crash, I couldn't tolerate it.

    It benches in many marks faster than a PIII-500 does, and it costs a whole lot less. Let's not forget that when you put well-manufactured components together, the benchmark results can differ VERY much. I also have a p200mmx at 250mhz (83 bus) that outperformed a Celeron 300A system because it was built on a good motherboard with a good hard drive.

    Moral of the story: Don't be nearly as concerned with how fast your processor is. Far more important is the processor's stability and the stability of the components you put your processor with.

    Interested in a better description of my PII-350 at 467? go to: asmus.html []
  • No, that doesn't make me mad. What does irritate me is that the graphs are redundant. The information is conveyed more understandably (and with no risk of confusing a reader) with a table.

    Spec_int95 Athlon PIII
    (350 MHz, L1=x y z
    L2=y, etc.)
  • The graphs are indeed very misleading, but you couldn't expect otherwise from the actual manufacturer of a product.
    These things are of course worthless, because they only confirm what everybody has been saying about the K7 all the time, which is that it beats Intel's fpu, and thus their 3D as well.

    Heck I'll just wait for Tom and the rest of the world. Does anybody know about the availability of the motherboards though?
  • Well gee look at that, what a confusing table!
    The tabs did not show up, ah well, you get the idea.
  • From k7preview1.htm []:

    The K7's use of the EV6 protocol also opens a door for AMD into the world of multi-processor systems, an
    area that Intel has had a lock on for the last several years. The EV6 implements point-to-point topology. This
    means that if there are multiple processors within the system, each gets a dedicated connection to the
    chipset. Intel based multi-processor systems must share a bus interconnected with the chipset. This is the
    technology that Intel has refused to share licensing for an so has pushed x86 CPU manufacturers away from
    developing multi-processor systems. This technology of Intel's is also limited to the use of only four
    processors within a given system. AMD's development of the EV6 protocol for x86 processors opens
    systems up to use as many as 16 processors provided the memory bus architecture can support it. Even
    though each processor gets a dedicated connection to the chipset is must share a system bus from the
    chipset to main memory. This will step up development of larger and faster memory types than current 64-bit
    access SDRAM. You can be assured that AMD will be developing for a variety of new memory types like
  • Does anyone know where I can find an equivalent of this [] for the Athlon?
  • Assuming that these machines are equivalently configured, AMD's statistians intentionational set the starting x-axis closer to 100% to give the illusion that Athlon is way faster than Pentium III. Don't get me wrong Athlon is faster but the modest 10% in SPECint_base95 looks like a whooping 10% by bumping the x-axis.

  • I really hope that AMD can sell these at a reasonable profit. If not, we probably won't see a K8 or beyond. Any idea of selling price for CPU, motherboards and complete systems?
  • the initial results would say the K7 will be the king of the x86 hill for the next year or so.

    But only if they can get around their historically abysmal early-production woes. Of course, that point has been beaten to death by others, so...
  • And with the monopolistic practices being poked at and sternly watched for, AMD has a chance, if compaq/ibm/dell and the big boys of the market don't buy into the Athlon processors, i could see AMD suing all sorts of people and eventually winning.

    The market is so fragile and based on reports instead of #'s that anyone could take over, and i don't see AMD dissapearing anytime soon.

    Plus 10% increase and higher for use on current equipment is AMAZING. Just think if your PCI SCSI controller and AGP 4.0 Video board ran at 200 MHZ the speed increase would be stronomical. Research and Data Storage companies would kill for that kind of performance and i see that driving sales of high end equipment more then joe schmoe who doesn't now what the hell he's buying in the first place.

    I personally can't wait to get a CPU from them again, i remember my 386DX 40 was the fasted machine around, and i could run DESQview and my BBS no problem, and OS/2 was great on it also. I see its time for AMD to have the fastest processor around and I am looking forward to it.

    also, i could care less if it was amd or cyrix or ibm or intel.. i'm also thinking about a nice 450mhz G3 to add to my assortment :)
  • And the win3d benchmarks make it look like the AMD processors are at least 200% faster than the Pentium III 550Mhz, a closer clook shows that it is *only* 40% faster.
  • If you look at most independent benchmarks, the only place the K6-2 beats even a Celeron was in low-end Winstone (basic windowing functions). The Celeron beat out the K6-2 in almost all other areas.

    The K6-3 was marginally (.1 to 1 point) better than the celeron, save in the FPU, but, since it was a known issue already, it's not worth harping about.

    Also, since I'm too lazy to post multiple articles under this subthread, I'll answer the "If not for AMD, the Celeron wouldn't exist". That's bull. Celeron was initially intended to take the lower end market away from the Cyrix chips, as AMD was competing in the mid-range market. It just so happened, that, once the 128K on-die cache was added that performance in 90% of the benchmarks became almost identical to (if not superior in a couple areas) the P2. If you wish to use that sort of damaged, circular "logic", I can justify saying that Intel is responsible for the K7.

    Also, to the contented owners of the various AMD products I mentioned before. I'm glad that you are satisfied with the hardware you've bought. I, however, have found their offerings substandard for my purposes. Usually I find the reasons for the performance deficits I see in independent benchmarks. It's something you don't normally see in manufacturer-run benchmarks, since you don't have any sort of clue what environment the tests were run under.

    Plus, NOBODY, AMD or Intel or anybody else for that matter, is going to come out and go "Yeah! We're 20% faster here, here, here, and here. But we're 20% slower here, here, and here."

    Again, if you're happy with what you have, more power to you. I prefer to wait. Since I prefer not to get cut up on the bleeding edge.

    Chas - The one, the only.
    THANK GOD!!!

  • OK, I'm biased. Still, until very recently the processor market was as dominated by intel as M$ dominates the OS market, if not more so. So I've been cheering on AMD all the while, just 'cause I always feel the need to support the underdog. And of course, I've also scraped up every little technical detail about the K7 that I could. And the benchmarks are no surprise.

    Yessiree, the K7 rocks. Perhaps it will start some true competition in the processor market.
  • from the site:

    Interger Performance

    it's OK AMD, I still love you.
  • It's worse than that - the size of the partitions in some of those graphs are the same, but the labels sometimes show different interval sizes. See the 3D performance chart - the left hand interval is 10% (90% to 100%). The remaining intervals - the same physical size on the graph - apparently span 20%.

    This sort of blatant graphical manipulation of statistics only undermines their credibility in my book.

  • I think it may be the "more sophisticated motherboard" that saves AMDs butt in the end. Whatever part of the Alpha architecture they could bring in and still maintain the sacred cow of the wintel market (backward compatability), the better off they are. DEC did some amazing things with the Alpha, and the more of us that can reap the rewards the better. Anyway, it's a marginal speed gain, and people will continue to buy AMD processors for the same reason they always did. They're cheap. Besides, Linux already runs on the Alpha, let's just ditch CISC and be done with it.
  • Doesn't it make you mad when people do shit like this? If you were to look at the bars and not notice the numbers, you'd conclude that the Athlon @550 was about three times as fast as the P3 @550 in the top two categories and about twice as fast in the third. It is in fact a good deal faster, but not twice as fast. When presenting just a part of a graph like this, the least you can do is indicate a break in the axis. This is pure evil.
  • All these benchmarks and stuff proclaiming that it's fatser then then PIII etc, but what speed is the L2 cache running at. I'm assuming that it's 1/2 core speed like the PII/PIII. I can't find any info to say otherwise.

    I seem to remember AMD saying that you could eventually buy chips that had a full core speed L2 cache. Link that with the fact that upto 8M of L2 cache can be supported on the EV6 async bus and these babies are gonna be pretty kick ass as AMD roll out the server version later next year.

    I think i'll wait just a little while so that i can go SMP K7 :))


  • Now really, modest 10% look like whooping 10%? What a nice trick. But in fact moving the origin away from zero is common in such diagrams, and it is useful, because more space of the diragram is used for the differences instead of the 0..100% range that bears no information. Of course in this case, just showing one processor is faster than another, the whole diagram itself is eye-candy and not really useful.
  • by Anonymous Coward
    The benchmarks on the K7 compared to the PIII seem very similar to the comparisons of Linux and NT. I can't wait until some TRUE third party tests (none of this "sponsored" crap) tests come out for the K7 vs. PIII and Linux vs. NT...
  • Posted by TRF:

    From AMD's Athlon FAQ: []


    Will the AMD Athlon[tm] processor support multiprocessing?


    Yes. The AMD Athlon[tm] processor bus architecture is designed to support scalable multiprocessing. As the AMD Athlon evolves into a family of processors, multiprocessor systems (workstations and servers) based on forthcoming AMD Athlon platforms are planned to become available. The number of AMD Athlon processors in a multiprocessor system is a function of chipset implementation, and not the AMD Athlon design. Forthcoming optimized chipsets are planned to enable multiprocessor system designs based on 2,4,8 or more AMD Athlon processors.
    Every 45 seconds, another arrest for Linux. 695000 last year. It's time for a change.

  • The alpha is increadable simple.. by simple I don't mean sucky, I mean very elegent.

    The alpha represents (for cpu design) a brilliant analysis of what is fast and what is not.

    The alpha has a limited set of VERY simple instructions. Branches are predicted and delt with in the first or second pipe stage.....

    Basically the Alpha ROX!

    But it does much less per cycle then a IA-32 processor....But that is why the clock speed of an alpha is so high!

    "There is no spoon" - Neo, The Matrix
    "SPOOOOOOOOON!" - The Tick, The Tick
  • SPECint_95 = 25.7
    SPECfp_95 = 22.5

    Probably the fastest (affordable) an Compaq AlphaServer DS20 (21264@500MHZ) machine has:

    SPECint_95 = 23.6
    SPECfp_95 = 48.4

    and the fastest intel (PIII xeon @550 MHz):
    SPECint_95 = 23.6
    SPECfp_95 = 15.1

    So unless Compaq starts making $3000 machines I will use K7's soon
  • Wasn't the K6 supposed to blow away the Pentium chips? The AMD graphs said it would. Sorry, didn't happen. Plus the little beasts were hot and temp-sensitive (15 degree operational range).

    Wasn't the K6-2 supposed to blow away the P2? The AMD graphs said it would. Sorry, didn't happen. Even the Celerons whupped up on them (and then the Celeron A-series just made it look even worse for AMD).

    Wasn't the K6-3 supposed to blow away the P2/P3? The AMD graphs said it would. Sorry, didn't happen YET AGAIN. The only areas where the chip really performed better was low-end Winstone. As if I really cared that a window would render itself on-screen in .001 nanoseconds instead of .02 nanoseconds.

    The only machine based off AMD's technology that had any excessive power was the HYPE MACHINE.

    I'll wait for some independent tests from the various hardware sites and my Q&A department before I go dropping cash for them. All these morons (Dubbed Athalonic Supporters on various forums) who pre-ordered theirs are dancing on the razor's edge.

    In addition, the commodity market isn't going to see K7 for a while yet, as their production is going toward OEM supply FIRST. And the prices they're quoting right now ($324 for the 500mhz I think was one), are BULK pricing (in lots of 1000). Figure a fairly splefty markup to take place.

    Sure, they COULD get lucky and get a chip that absoloutely tears up. Then again, they may get stuck with what the other AMD chip-buyers got. Buyer's Remorse.

    I hope that AMD finally comes down with a winner on it's side this time. But I'll play it cautious and wait a while.

    Chas - The one, the only.
    THANK GOD!!!

  • They also don't need a "break in the axis." Stop assuming graphs always start at 0. They don't. Your fallacy, not AMD's
    Really? When they tought me about graphs in high school they said the break symbol was required.
    You admit you judge a book initially by its cover
    I'm having a hard time seeing how "judging a book by its cover" describes what I did. Perhaps you think I think the Athlon sucks because AMD's marketing is misleading. I didn't say that.
  • Um.. well you're right if you're a gamer. However, in the real world, not everybody use their machines as frag stations. K6 interger performance is still strong.

    Besides, it's only because of the K6 that we're seeing sub $80 Celeron prices. Think Intel would throw out money like that without competition? No way..
  • You are correct at first when you say that the K7's double-precision floating-point is quite fast. Actually, the K7 on double-precision should be (relative to the P3's performance levels) much faster than in single-precision. This is relatively speaking -- meaning that the K7 should be *much* faster than the P3 on double-precision, and only a good bit faster on single precision (depending on the code mix, see below).

    Also, the K7 is (or should be, depending on your level of optimism) quite a bit faster in single precision, too. The main difference is that the K7 has two true multipliers: the PPro/2/3 only has one multiplier, so to get full utilization of the pipelined floating-point unit, you have to intersperse your FADD and FMUL instructions. This is the "Intel optimization" that makes such a difference for floating point on the P6-series, and seems to hurt the non-Intel CPUs. The K7 should run fast even on code that is Intel-optimized, in addition to being able to handle the cases that cannot be optimized to run quickly on a P6 processor. Furthermore, the K7 has a halfway-pipelined division unit (something that is not pipelined on the P6 series), so that it can do a division and a multiplication sorta in parallel -- the multiply doesn't have to wait for the divide to completely finish, they sorta share the multiply unit. (Maybe the best way of saying it is that the K7 has 3 and a half floating-point pipelines. Or maybe that's an oversimplification.) Last, the K7 has a dedicated load/store/housekeeping pipeline in the floating point unit (2 pipes for add/multiply, 1 for load/store/housekeeping), and that eliminates a lot of the penalties associated with the x86 stack-based floating point architecture. This is one of those hard-to-quantify (in terms of absolute cycles) things that should make the K7 blaze on floating point.

    As for Intel-specific optimizations, the K7 and P3 are more alike than they are different: the K7 is just bigger (more instructions/clock) and easier to manufacture at higher clocks. (The deep buffering and scheduling depth doesn't hurt, nor does the 128K L1 cache ... :) So really, anything that's optimized to run fast on a P3 will also run fast on a K7, unless you're referring to SSE instructions. In the case of SIMD instructions, a P3+SSE and a K7+3DNow have the same theoretical throughput -- it's always up to the compiler/programmer to get the most out of that.

    There have been a number of different "third-party" (unnamed third parties, that is) confirmations of these numbers. Generally, they all seem to put the K7 and P3 about on par (with the K7 slightly edging ahead) in integer performance, and the K7 dusting the P3 by 40-50% in raw floating-point power. Furthermore, remember that the K7 has a number of design features (deep read/write buffers, a deep scheduler) that take better advantage of high clock rates -- so it should scale to higher clock rates better than a P3. How this all translates to real-world speed and real-world yield levels is yet to be seen, but the initial results would say the K7 will be the king of the x86 hill for the next year or so.
  • This would of help in the old P5 days of 16k or 32k cache memory. But it doesn't help too much in these celeron 128k cache days.
    As always when performance is involved, that should have a large "YMMV" caption. Whether or not it helps depends on whether you run apps which shove lots of data around or run more than one app at a time.
  • The short answer is: sorta, but not like you're thinking. Anything purely integer based, probably about the same speed as an equivalently-clocked K6. (Remember that the K6 really doesn't get much above 500Mhz on the .25mu process, so ... it's hard to compare them clock-for-clock.)

    All the things you mention, though -- playing mp3s and mpeg2 movies -- are floating-point intensive. So, in that case the K7 should completely leave a K6 in the dust -- on the order of 100% (or more) faster at the same clock. (Pipelining is a wonderful thing.) More like "AMD K7 can play a 44100 16 bit mp3 using 5% *of* the cpu time (95% less) that a K6 uses", and the same for the movie. That's a substantial difference, if you ask me. Granted, maybe you don't need that much power -- the K7 is targeted at the high end engineering workstation market and the "enthusiast consumer" (people who live for Quake3, they mean :) ... if the K6 does all you need (and you wrote the above as if you already own one), then don't upgrade.
  • Heh. I'm on a 192kbps DSL line. I just think it's dumb when sites are not optimized. That page could've been an order of magnitude smaller and still contained the same information in a pleasing format.
  • ...but liars use statistics. I can stand them being faster than the PIII. The PIII ain't all that. But the scale bothers me. Whether its a 9% gain, or a 30% gain, it looks the same...suspicious. Gimme integer values and I'll be happy.
  • by sellout ( 4894 )
    My guess is that it's a typo. Here's why:

    1. If you replaced the 90% with 80%, the graph would look identical.

    2. If someone were to look at the graph and make the assumtion that the lines are at 10%, it would look _worse_ for AMD.
  • I think John Carmack once mentioned that Q3 performance seemed to fall roughly between SPECint and SPECfp. I think it was in a .plan.
  • Possibly not, but the information is right there on AMD's website (admittedly for batches of 1000, but that's the best you ever get out of chip manufacturers....)

  • The K7 doesn't have a huge edge despite requiring a more sophisicated motherboard. Even if they make their volumes and already had the same reputation as Intel, it would still be a tight run for the money.

    But AMD and Intel aren't on equal footing. Intel has enough cash in their pocket to be very competitive with their chips. For some reason, they also have a better reputation.

    I'll buy one of these guys if I see one, but unfortunately, I think I'll see AMD go bankrupt before they get enough volume out the door to put a K7 in front of me at prices near the MSRP. I'm still waiting to see a K6-III for less than the price of a Celeron with motherboard.

    I'm beginning to think it is a good thing that I didn't buy stock in the company.

  • As someone has pointed out below, SPEC benchmarks are designed to measure high-end performance, so they may not translate *directly* to Q3 frames-per-second. But they are proportional, to a large extent (at least on the x86 platform). And saying that CPU X beats CPU Y in Unreal FPS is a far cry from CPU X beating CPU Y in SPEC. The rules are a lot tighter for SPEC -- on the SPECbase numbers cited for the K7, there really isn't much room for compiler optimization or any other forms of cheating.

    SPEC is a collection of thing like gcc, some heavy-duty simulations, other text processing, etc. designed to measure the integer (no extra "r", AMD!) and floating-point performance of a CPU-motherboard-memory-compiler system (a CS class in architecture will convince you that these are largely inseparable components, as performance goes). So while SPEC is "only a benchmark", it does determine that measurement using things we care about (e.g. gcc).
  • The Intel site was not set up for Joe Q. Slashdot, but rather for promotion agencies, local hardware shops, and other websites/print media.

    I sometimes wish more companies would put up full quality promo shots that I can work with. It makes it a lot easier than begging.
  • Most people with a functioning brainstem would look at the little figures known as numbers next to the graphs and conclude that, in fact, the K7 still does beat the holy hell out of the Pentium.

    - A.P.

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

  • i could deal with the fact that he was benchmarking with broken code, as long as he labeled the benchmarks as such, but when brian hook said in his .plan to take the benchmarks with a grain of salt, because the code was broken, and tom's rebuttal was "no, brian hook, you are wrong, my benchmarks are valid" i lost pretty much any respect i had for the guy.
  • We might not expect anything else from a manufacturer.

    This, however, does not mean that we should accept this kind of deliberate marketeering placidly.

    ("Oh, I won't bother to mute the commercials, I don't really take them in." "What? I can't hear you over the jingle!")


  • In 1995, we payed close to $3000 for a Pentium 100 with 16 megs of ram and 540 megs of disk space. What else could we have gotten? Nobody was *making* PCs with AMD or Cyrix chips in them, because AMD and Cyrix had nothing comparable to the P100 at the time! The spares market (think "pricewatch") was nearly nonexistant, so forget about buying the pieces and making your own. I didn't know how to build a PC back then anyway; most people still don't. The fact is, you can buy a top-of-the-line PC nowadays for half what we paid for the top-of-the-line back then. For this we must thank AMD and Cyrix et al.

    - A.P.

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

  • This practice always annoyed me as well. 3d video card makers are particularly bad about this. When one card runs Quake at 29.5 fps, and their competitor runs it at 28.7 fps, they set the x-axis at around 25 or so, so that 0.8 fps difference looks like a lot. If they set the x-axis at 0, where it's supposed to be, you wouldn't be able to see that there was much of any difference at all.
  • There's some people complaining that they can't see the actual #'s, so here they are:

    AMD K7 @ 550 MHz /w 1/3 speed (183 Mhz) L2 cache:

    SPECint_95 = 25.7
    SPECfp_95 = 22.5

    I haven't found a good list of x86 chips vs. #'s to compare, but if someone has them, feel free to post.
  • The K7 is supposed to be a full generation ahead of the PIII.

    The bus speed is doubled, the cache is better, and yet it only gains 9% improvement in Integer performance.

    It's dissapointing.

    This is a comparison of an ASUS motherboard over a year old, and a chip which you can get now to an AMD "reference" board, and a chip which I know you couldn't buy unless you tried really hard or paid a heck of a lot of money for. According to AMD "Consumer systems based on the AMD Athlon processor are planned to be available in Q3 1999." -- I don't think Intel has any big plans for September, but they're not going to stand idly by.

    I want AMD to kick some butt as much as the next guy, but abyssmal FP didn't sink AMD in the past, and stellar FP isn't going to make them big in the future.

    But you're right, the FP performance is definately strong. If this chip is produced in decent volume and cheap enough, it will be serious competition for Intel. I just don't think AMD's production track record is all that great, and I can't think of any reason why a 200MHz SlotA motherboard made in lower volumes than a Slot1 board is going to be cheap.

  • ivity/specint95.htm []

    PII 450 18.5
    PIII 450 18.7
    PIII 500 20.6
    PIII 550 22.3 ty/specint95.htm []

    300A 12.0
    333 13.1
    366 14.1
    400 15.1
    433 16.1
    466 17.0

  • Something else to point out, IMHO, is that integer performance is not that big of a deal anymore. It's kinda like 2d acceleration; pretty much everyone has it figured out. Sure, I'd like a cpu that runs 45% faster on integer, but that's not going to happen on x86. The big optimizations are still fp.
  • This is probably a trend for the future. I can imagine that most of the integer parrellelism has been extracted from x86 code. I can only imagine x86 integer parrellelism would increase only via better compilers and optimizers. However, it is hard to expect significantly better performance from any new integer core for x86 code. 9% on unoptimized code is good, and I can imagine that the difference will increase slightly as optimization also increases.

    Expect future x86 chips to focus on high MHZ, wide buses, and floating point performance.

    Oh, and stellar FP will make AMD big if they can pull it off. Look at intel and it's celeron line. The only reason the celeron and PII/PIII was big (last time I checked) was that it had a good FP unit (compared to the K6) and did well on 32bit code.

  • Currently the intel 4.0 compiler generate slower code than Microsoft's.

    I'll bet it doesn't for SPEC. Take a look at the real reported SPEC data on Everyone uses the Intel compiler.

    What does the MS compiler beat the Intel compiler on? Is there even an Intel Fortran compiler?

  • "It's an i386 chip; that's supported. It has 3DNow; that's supported (not by AMD, but hey, it's a compiler problem). What else do you want?"

    How about they help work on gcc to make k6, k6-2, k6-3, and k7 optimizations?

    Intel releases their own compiler which beats the crap out of gcc for optimization of Intel chips, I would like to see AMD produce optimizing code for gcc.

    If AMD did produce an optimized version of gcc that and you could compile a Linux kernel or entire system that ran 20-33% faster on AMD that the equvalent Intel chip they would dominate not only the linux home user market, but also the linux server market which is starting to develop.

    High end corporations such as Dell and Compaq would still put Xeon's in their DB servers, but if VA and Penguin started using AMD because it was optimized, AMD would start to penetrate a market that is still owned by Intel.

  • I would say 50% faster at the same clock speed for FPU operations is a pretty damned *enormous* edge. What do you consider huge?

    - A.P.

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

  • But in fact moving the origin away from zero is common in such diagrams, and it is useful, because more space of the diragram is used for the differences instead of the 0..100% range that bears no information.

    The use of a chart is to compare two or more items. It should give enough information at a glance to not only decide which item is (better|larger|faster) but also by how much. The part of the chart which "bears no information" is used so that you can interpret how much (better|larger|faster) one item is from another -- if one piece is twice as large as another it should be twice as (good|large|fast) as the other. When companies move the origin away from zero, the graph no longer serves that purpose. Sure, you can look at the numbers and do the math, but why not just post numbers instead along with big letters spelling out We Are Faster since that's the only information a non-zero-origin graph gives, it would even save the time spent downloading the graph. It may be common, but it is certainly intentionally deceitful.
  • Just an observation about the end price and the published prices... The last couple of years anyway, with intel chips, you have been able to buy them quite a bit cheaper on the net than what intel said they were going for, usually in the neighborhood of 20-40% less. I'm not saying that this will happen with the K7's but who knows. I'd be absolutely thrilled for a $250 (or whatever) K7 500.

  • Their stock roughly triples whenever they get a good chip out which seems like it is going to beat Intel's. When I say better, I mean like the time when the fastest chip you could buy was AMD's, and it was cheaper than Intel's. The only thing Intel had going for it then was reputation... then the math bug became a hot issue...

    Personally, I think the pattern is broken. The K7 with motherboard isn't going to significantly outperform or undercut the Intel chips. If I thought it would, I would look into the precise correlation of the stocks to the particular chips.

    Yahoo has some good charts. Check the long range here []:

    I think some of the peaks are the AMD386DX40, the AMD486DX4-120, and the tallest one is the K6... those sad little peaks on the end are probably speculation and hype from the II and III.

    Whomever is watching AMD closer, feel free to tear this analysis to bits. I don't know squat about stocks. I do know a lot of people lost money quickly when their stock went from $27 to $17 earlier this year. All the while AMD just pretended that their production issues were solved. How did that silly class-action lawsuit go anyways?

    Only put the money into them if you can afford to lose it. I do agree with the "supporting a good competitor" argument. I just don't think AMD has much of a chance anymore. Corel could be fun though.

  • The "good" thing is that I recall reading somewhere how AMD got up to making 5 million CPUS a month, but was only getting 3.2 million out the door or some such. Dont quote me on the numbers, but the point remains: with the K6 family, selling them is now the issue.

    Matt will still probably build a dual celeron box.
  • As a K6 owener I am quite pleased. Not only has the K6 performed, but for me it was worth the price I paid. When I bought my K6 it was half the cost of a same priced Pentium... for what, a gain of at most a second in opening a window... I'm going to use your statement As if I really cared that a window would render itself on-screen in .001 nanoseconds instead of .02 nanoseconds. I really wouldn't care if it took as much as half a second longer. Keep in mind, I only have a Pentium killer, and when finding the best deal for my money, the K6 was The Deal®. They have only gotten better with each release since. I look forward to getting my K7, but I will be waiting for a while before I do. Not because I don't want it, but because I can't afford it myself. You won't find me buying a PIII either, but for the same reason. When it comes right down to it, the K6 serries has always performed as well as I expected it to, and often better.

    I will admit that the K6 I have is somewhat temperature sensitive. I tried to overclock it, and raised it from it's rated 233 to 225, by raising the FSB and lowering the clockrate. While a slower clockrate doesn't seem like an improvement, the faster FSB increased my system performance by close to 10%. More than enough to make it worth it, and the CPU runs cool to the touch. I could have raised my core voltage and bumped the clockrate up slightly I think, but it is already faster, and again, an excellent deal.

    Am I slightly biased now. Yes!!. But only because I've had excellent performance from the K6. Only because I found that the processor exceeded my expectations. When I finally collect enough money to consider a K7 or PIII, I can assure you that the K7 is going to get my first consideration.

    Time flies like an arrow;
  • If I was shopping for a new system.. I'd either get a really cheap and reasonably fast cel or.. really cheap and kickass fast dual cel. The word is making dual cel systems is painless now, with some adapters or something.
    Yes, I know, dual systems only work with only a few OSen/programs. But hey, these are the programs that require that much horsepower.
    It seems like AMD is having huge losses because of celeron and they can't afford to sell athlon cheaper..
    Does this all mean AMD will go bankrupt? I hope not. If they do, intel will start overcharging for fast cpu's again. I wonder if DOJ going to attack intel for cel pricing..

    ah well
  • I know with the G3 processor (and I admit I am assuming the same with CISC processors) that the size of the backside cache made more of a difference that the main memory bus speed. Newer Tech posted a paper demonstrating that moving from a 30Mhz to 66Mhz bus (this was when the G3 first came out) with a G3, produced a mere 5 - 8% speed change, but moving from the 512K backside cache to the 1MB produced an increase of aprox 20 - 30% and you could get another 10-15% (if I remember correctly) moving from the 2:1 cache ratio to the 1:1 cache ratio, this is why Apple's old 300 beige (which had the 66mhz bus but a full 1MB cache), could beat the new 300 blue (with a 100mhz bus but only 512K cache) in processor tests, of course the new ATI 128 vs the Rage Pro made sure the new 300 beat the old in every other important test (Quake frame rates!!!)
  • The Spec benchmarks were down with 1/3 speed L2, but the first shipping models will ship with 1/2 speed. Support exists for 1/5 -> full speed L2, it's just a matter of whether AMD will use the whole range.
  • When I bought my K6 it was half the cost of a same priced Pentium.

    half the cost of the same price??.. what? hehe sorry for raggin on your grammer.. just couldn't resist :(
    (you know. I really think this comment doesn't deserve 2 points.. but its automatic for me.. umm as a suggestion would it be possible for people to purposfully lower the scroes of their own post. This would be helpful in things that we don't think deserve a high score or things that we have to resubmit and would like to moderate the old one out of existance..hmm maby for having this nice suggestion this does deserve the 2 hehe -laugh- )

  • Besides, I don't want to support a company that sticks an ID on my processor... Call me paranoid.

    Then you don't like Sun or any ethernet NIC manufacturer either. But it sure blows, because an awful lot of the world runs one (or likely both) of those. When you throw Pentium III's into the mix, you're going to be hard pressed to do anything digital in the modern world without at least passively supporting the concept of serialized computer hardware.

    And when you go out and actually buy that 3com NIC, you're really feeling the pain. Because you just told 3com that you have no trouble whatsoever with a unique MAC address. Like McNealy said: "You have zero privacy anyway. Get over it."

    Don't get me wrong: I agree with you. I think think PSNs suck (no, I'm not talking about packet switching). And I've been a long-time AMD user (I'm even an AMD stockholder). But you can't possibly try to use the argument that you'll buy AMD because you're in some way against serialized hardware. It just won't fly, man.


    Oh yeah, you can take a look at pictures of the K7 [] if you're curious. There's no mistaking what hardware you're running, that's for sure. I might have to actually put the case on...

  • PC100 memory on the systems. Imagine the difference when you're running that K7 on a 200MHz bus with the appropriate memory. What's Intel's next bus speed bump? Only to 133MHz?
  • Marketing, evil, whats the difference? =)
    After all, our favorite evil empire was built mostly on marketing...
  • IIRC The K7 does double-precision fp ops in a single cycle while the P-III requires 2 cycles. In which case, before interpreting these results we need to know how many of the SPEC tests use dp. The only test which mentions dp is 146.wave5. So I guess the margin for the K7 will only be slight lower for pure sp, but significantly higher for pure dp.

    Also the results on the website are all 'BASE' values (i.e. little compiler optimisation). Presumably the P-III will beat the K7 under optimisation until compilers are modified for K7 optimisation.

    Still, these figures are uncouraging.

  • I just want to know if this new AMD chip will have SMP, I'd lvoe to have a dual 600Mhz AMD system, although I do not know what I'd do with it yet.. maybe some Distributed net stuff ;-)..
  • We still need to support AMD. Think of a world where AMD did not exist, do you have any idea how much a pentium would cost? Intel only lowers its prices because it has to. Intel has publicly expressed that the only reason it is selling Celerons so cheap is because they need to compete on the low end. As soon as AMD is gone we are gonna get *ucked up the yin-yang by Intel. Besides, I don't want to support a company that sticks an ID on my processor... Call me paranoid.
  • Yes, one of the great features of the K7 is supposed to be its SMP performance. It should scale better than the Intel chips because of the setup on the backplane.

    I agree... a Dual K7-600 would rock :-)
  • From

    CPU's at
    500MHz $298
    550MHz $459
    600MHz $686

    I guess that with several mobo vendors, long term mobo prices shouldn't be more than a current dual P-II, but the first few are bound to be pricey.
  • Not the same thing Sparky, the processor doesn't need a distinct ID. Of course, I'm sure the original poster would take a SPARC though, and they've had processor ID's long time

    /. acct# 449
  • Ok.. I'm getting sick of you people whining about how they show their bench marks. Just take 100 off of it and stop your bitching.

    As opposed to whining about people whining about the graphs?

    ...remember the test systems are even fully optumized yet for this chip.

    I presume you mean "aren't"...

    But I feel the 2nd batch should be alot better and the support will be alot better in the hardware and software areas.

    And I feel that you would look better with a telephone pole stuffed up your ass. Doesn't make it come true, though (more's the pity).

    Also note that the 600's will overclockable to 700 mhz.

    And where did you get that information? Too much crack can do nasty things to your brainstem.

    Note this too, you will be able to do 8 meg of cache, better marks with that much cache, and you can do 16 way processing.

    When they actually have the chipsets taht can handle these features available, you mean. Even AMD won't give a firm date for SMP, and the 8M cache version is gonna take a lot more moola than what you steal^H^H^H^H^Hearn.

    Now if AMD wanted to kick some ass in the field, the would give 100% linux support.

    And in what way do they not support Linux, moron? It's an i386 chip; that's supported. It has 3DNow; that's supported (not by AMD, but hey, it's a compiler problem). What else do you want?
  • Why does everyone expect AMD to magically get cash and resources put into their pocket? AMD has a single fabrication facility, in which they make several lines of CPU's. Intel has many more in which they make many lines of CPU's. How the hell do you think AMD is going to get their chip prices lower? And why is it that the K6-III is seen as a competitor for the Celeron. It's not! The Celeron is a budget chip, it has a small onboard cache and uses the PII's processor core, therefore it doesnt cost nearly as much as a PII (SRAM is very expensive). The K6-III is not a budget chip and hasn't been marketed it as one, it's supposed to compete with the PII not the Celeron. I can't stand how people expect AMD to just come up from behind and hit Intel with a shovel. Only recently have OEM's started using AMD's chips in their machines. A year ago how many new computers could you find with an AMD chip in them? Not very many. Seeing as how AMD has just begun to really give Intel some competition why not calm yourselves and stop talking about bankruptcy and get off AMD's case for a minute. When their new fab facility is finished we're probably going to see a marginal drop in their chip prices because then they will be able to produce more chips and let the price drop on them.
  • Per 1000 unit quantity, dont forget the markup by the distrubutor, otherwise they would make no profit.
  • I'm not too sure, but could someone clue me as to what kind of SMP support (if any, this chip would have?). As it is, I dont really see myself buying a single processor based system anytime in the future.

    The prices for 500mhz seems to be reasonable and if they are good enough, and does SMP, hell, i'd even dump my PII's and get some of these babies.

    On SMP, saw the cool abit dual celeron board? very nice (now i wonder if that would compete with a single K7).
  • SMP will be available, but the first MB's probobably won't hit the shelves until the end of the year. With the EV6 bus, the SMP systems should be nice due to a dedicated 200 MHz+ bus between processors. We should see 2, 4, 8, and even 16-way K7's in 2000 (Poseidon Tech makes will be making the 16-way).

    Also, Intel's about to disable the line on the Celeron that allows it to be used in SMP configs. Apprently the dual Socket 370 board ticked them off a little too much.
  • No significant increase can be gained by feeding the CPU data twice as fast? What are you smoking, and where can I get some of it?
  • Actually, more bandwidth shouldn't change the SPECint or SPECfloat scores shown.

    Now, when Toms Hardware, Ars Technica and others start posting the FPS of Quake 3 THEN we should see quite a speed up between a 100MHz and 200MHz motherboard.

    (Assuming HD access is non-exisistent of course ;)
  • Yep it supports SMP. The cool thing about it is, it uses memory bank interleaving. The memory still runs at 100Mhz, but the CPU's skip between banks, essentially reducing the memory bottleneck of SMP machines.

    Some of AMD's documents about the K7 went over this in detail, but it seems AMD took most of those docs down. Anyone have them?

  • The AMD K7 can play a 44100 16 bit mp3 using 5% less CPU time than the K6. The K7 can play a 24 fps movie using 5% less CPU time than the K6. The K7 boots Linux 5% faster than the K6. Is the K7 going to do anything new instead of running the same applications 5% faster?
  • How is this BS? They used a standard benchmark (SPECint, SPECfp) designed to test _different_ CPUs with the same benchmark.

    If you run ZiffDavis benchmarks on a Sparc, it's gonna say that an Intel chip is faster, because the ZiffDavis benchmarks are written for Intel chips.

    I seem to also remember that a month or so ago when QuakeIII came out, somehow TomsHardware had benchmarks for it, even though the 'timedemo' and 'timerefresh' commands were _broken_.

    Here's [] the story on Blue's.

    You sure you still want those benchmarks from Tom?
  • you won't be running tasks that fit only in the CPU's onboard cache, you'll be needing data from main memory all the time. Right now, the bus between main memory and the CPU is a bottleneck on both Intel and AMD. AMD'll just be making it less of a bottleneck by opening it up to 200MHz. SO we should see those worthless ZDNet benchmark scores go up quite a bit. Along with just about all your other apps.
  • Up to 16 CPUS, IIRC ;o)
  • There is however a significant difference between tv commercials and amd's graphs. You "watch" their graphs by choice, however you are forced to either change the cannel or be brainwashed when commercials are waving in on you.

    People just don't advertise in an "objective" way, it's the way it is. Of course it's crooked, but whining about it is rather pointless. I think this should never have been posted actually, to avoid discussions about information that is deemed to be of inferior quality.
  • > Really? When they tought me about graphs in high school they said the break symbol was required.

    Well, when I took math in college, the prof said that all graphs are arbitrary anyway. Starting at 0 is a convention, and a good one at that, but not absolutely required. As a matter of fact, the prof also mentioned that in the topology class, they also throw out the convention of straight graph axises and learned how to do "normal" math on weird graphs. I suspect that your high school teacher probably made that rule to instill in you a respect for convention as well as to lessen the time he/she spent on grading homework.
  • The K6-III is the fastest chip at compiling the linux kernel (it beats everything except an overclocked Celeron), so it is useful for something - software development.

    AMD's chips have also tended to be cheaper than the equivalent offerings from Intel.
  • Misleading? Read!

    See the little numbers at the end of the bars? They give the actual percentage. Do you really expect a CPU manufacturer to create graphs that are flattering to the competition?

    Grow up, and realize that the job of marketing is to SELL products. It's up to the consumer be able to critically think.

  • There's some people complaining that they can't see the actual #'s, so here they are

    • Where did you get the figures
    • Why aren't the figures on the official SPEC site [] like they should be with all relevant data, like compiler version, hardware availability dates etc. etc.
    • Everyone else in the industry is obliged to write 'estimated' disclaimers all over unsubmitted results. Why isn't AMD? Are they afraid to admit they used the Intel compilers (what else could they have used?).
  • First of all, as for 3DNow under Linux -- there is absolutely *no* reason you can't use 3DNow under Linux. In fact, the new version of the Mesa OpenGL-ish libraries can use 3DNow if you have it, and people have reported 10-15% speedups in 3D geometry setup (which is what it is designed to accelerate). There's also a patch to mpg123 to use 3DNow to knock down the CPU load playing mp3's, and I think one of the mp3 encoders also has a 3DNow patch. 3DNow is simply some special instructions for using the floating-point registers in a special, funky, parallel way -- there's no operating system support required (unlike SSE, which requires rewritten context switching code). In fact, the K6-3 is probably one of the best available Linux CPUs, because it's cheap, blazes on integer, and the on-chip 256K L2 cache helps it out a lot in multitasking performance.

    As for it being a headache to use AMD -- this is sometimes true. There were some Super7 motherboard compatibility issues (esp. with the TNT cards), and in general video card makers haven't made that big an effort to support AMD, thinking that most of their market was based on Intel processors. This is changing -- and if the K7 is as high-performing as it initially seems, then I would bet most video card makers will optimize for it, if for no other reason than to show off the performance of their high-end 3D accelerators.

    Your last question is about the K7 in 3D games -- if anything, the place the K7 should shine is floating-point performance. The K7 should wipe the floor with P3s in Quake and other 3D applications. I kid you not. And this is just straight, raw floating point power -- no 3DNow or SSE optimizations required (although they could make it go still faster, theoretically). If you want to go really fast come October, you're going to buy a K7-650 and a Matrox G400Max, and that will be faster than anything you have ever seen ... under Windows *and* Linux.

    If your thing is fast floating point, if gaming performance is what drives you, then the K7 is your chip.
  • You can never really trust benchmarks by these hardware developers against other hardware, there is no way they are going to make themselves look bad. Look at the bar graphs, the way they made them look is 9% looks like 100% more than the PII, because they start at like 90%, there actually pretty simular, when they make them look drastically different. I am just going to wait and find out what people really think, in the real world. Benchmarks aren't all there cracked up to bed, in many ways. So I wouldnt go around bragging about this or that, just wait.
  • by JoeBuck ( 7947 ) on Saturday June 26, 1999 @01:06PM (#1831052) Homepage

    While the distorted way the graphs are drawn is a form of cheating, the SPEC benchmarks are industry standard and very hard to cheat, because the rules are very tight, so the underlying numbers can be counted on: the new processor is a hair better than the PIII for integer, with a more significant gain for floating point. Of course this doesn't mean that your floating point program will run 50% faster, but as benchmarks go SPEC is about the best available.

    The real question, though, is whether AMD will be able to manufacture this processor with sufficient yield to be able to meet demand or make money. In the past, the fact that they haven't been able to do this is the primary reason that AMD has lost money in recent years. If they can't convince the PC manufacturers that they have their past problems beat, no one will do a deal with them and the speed of the processor will not matter.

  • No significant increase can be gained by feeding the CPU data twice as fast? What are you smoking, and where can I get some of it?

    Only if it's hungry!

    This would of help in the old P5 days of 16k or 32k cache memory. But it doesn't help too much in these celeron 128k cache days.
  • Compare the results for 550mhz and 600mhz....

    If memory bandwidth is not an issue, then the percentage for 600 would be exactly 600/550 times the percentage for 550.

    For integer performance this is the case: 118 is very near to 109*(600/550). So this test fits in the K7 cache and increasing the bus speed won't gain you much.

    For the other two tests, the result for 600 is significantly lower than the expected value. So bandwidth is an issue here, and running the bus at 200mhz will boost performance.

    About the difference with a P3: since the K7 at 200mhz won't do better at the integer test, and, for all I can tell (I don't have enough information here), the P3 might benefit from a busspeed of 133mhz, the difference in integer performance would actually be less with the K7 at 200mhz and the P3 at 133mhz. (If you don't get this and suck at math, then just forget what I said.)
  • by jpgrimes ( 15330 )
    A better comparison would probably be with the Pentium III Xeon not just the PIII. But, several groups have gotten ahold of late model K7's and there benchmarks are similar (try looking at for more info).

    SMP is coming but your going to have to wait a while. There currently just trying to get to market. (supposedly K7's are shiiping but you won't be able to buy them for at least a month)

    This processor should scale to higher speeds more easily than the pIII which is being increasingly delayed. however, intel has such huge manufacturing capabilities compared to AMD.

    AMD financially is in really bad shape but this is a kick ass processor. I hope they pull it off. I tink my new machine is going to be a K7 :)

    its hard to beat the giant

  • Can you tell more about this? Would it need support from the kernel to get the most out of this?
  • by Samhailt ( 4891 )
    Oh give me a break. The whole let's attack Tom thing got way out of hand. Tom himself has even said the benchmarks were not perfect but they did say some interesting things. I just think alot of people finally found the chink in the armor they were so desperatly looking for. How many sites covered Tom's so called _broken_ benchmarks? Did you ever ask why they were covering it? Maybe stop to think and read Tom's side?

    The guy deserves a lot of respect. He's always been honest about a product and never falling for Name brand zealotisim. Unlike some of the reviewers out there. Never took mony from the company's or signed up for any "Review Promotions" (ahem, Hello 3dfx). Next time find out a little more about something before just taking a news byte off of some site about it.

    (I don't mean to sound upset here. it's just very few bothered to look at Tom's side and just attacked him with the rest of the crowd.)

"Mr. Spock succumbs to a powerful mating urge and nearly kills Captain Kirk." -- TV Guide, describing the Star Trek episode _Amok_Time_