Follow Slashdot blog updates by subscribing to our blog RSS feed


Forgot your password?

PIII - dead end technology? 151

S. Casey writes "It looks like Intel is beating a dead horse, and then some, at least that's my impression from this review of Intel's "PIII" and its Slot-1 competitors. The best part? "The PIII at 560 has a 7.57% increase in 3DMarks over the 300A at 450MHz. Less than an 8% increase in speed despite a 24.4% increase in clock speed... from a processor that costs twelve and a half times as much as the 300a." The review compares 4 of Intel's "best" CPUs and pretty much demonstrates that the PIII is a waste of time. Bring on .18 micron. "
This discussion has been archived. No new comments can be posted.

PIII - dead end technology?

Comments Filter:
  • Intel is turning into as much or more of a marketing company than MS.

    Can you say Pentium98?

  • Go get yourself a Sparc or MIPS or Alpha, even.
  • by Anonymous Female ( 17974 ) on Tuesday February 16, 1999 @09:04AM (#2013689)
    Anyone have any more information on the K7, other than it's going to rock?
  • at least the commercials are kinda nifty
  • I'd prefer and AMD, they've been doing some cool stuff lately. Those alphas are pretty sweet too.
  • That we should buy Intel only for the low end, as there is where the x86 architecture's better price/performance ratio resides. If you want real kick-ass speed that's worthy of the price, look somewhere else.
  • I like my ppro, it does what I need, and Linux likes it. I agree that building a faster processor does not always make a better processor. I think that what should be done by Intel is a complete rebuilding of the core in the PII, and actually having a new chip. Reworking the core dosn't need to add instructions to it, just make the logic faster, and require fewer clock cycles per instruction. Also possibly make it so there are effectively parrallel processors in the chip, like Cyrix and the 6x86. That would be an improvement.
    If Intel continues to build faster, but not better processors they will stagnate and become a has been. Creativeness and ingenuity make a company great, and a product welcome. Reliability makes it perfered. If a company dose not offer both, it's going to decline.
  • Silly Rabbit MIPS are for kids.

    uh, sorry. I couldn't resist. RISC, however, is my current favorite processor. I wonder if I could use a couple of the P3's to replace my oven though.
  • excuse me, but 3D-Now! has been out far longer..
    and all the game developers are happy to support it.. also, yes, the CelA has on chip cache, but the K6-3 has twice as much.. and both these chips were starting on the drawing board at least several years ago, and neither one could have likely known of the other..
    in fact, 3D-Now! should be, from what i have gathered so far, a better performer..
    KNI uses 50+ instructions to do what 3D-Now! does in 21.. Intel used a CISC approach, AMD a RISC..
    AMD didn't waste as much valuable opcode space..
    sure, game developers will support KNI - but it'll take a year for that support to reach the levels of the games that are out now or about to be that support 3D-Now!
  • by drwiii ( 434 )
    I'm quite happy with AMD, thank you very much..

    Can you say "Abandon Ship []"?

  • I just did a small writeup at work exposing the bullshit Intel is selling its customers. They recently started shipping 450 MHz Xeon chips with the 2 MB cache. In comparing 4-way servers from IBM and HP, I found that the price increase from a 4-way 400/1MB server to a 4-way 450 MHz/2 MB was roughly $14,000 - a 35% price increase. The improvement in performance? Well, that was only 14%. Take that number with a grain of salt because these systems (used in TPC-C) are super-tweaked/tuned for bragging rights in the server market. I bet you the P3 will show similar, marginal gains over the P2. Yet, Intel will charge a huge premium. Meanwhile, you can grab a Celeron box for half the price (not a 4-way, but you understand the point). Notice how the P3, from what I know, will only be available for 2-way systems intially....
    My question is, how much horsepower do you need for a departmental file-print server?? Run Linux on an old Pentium, and you should be fine, right? Anyway, my point is that Intel scams its customers by creating a lot of confusion in the marketplace (Celery, PII, Xeon, etc.....) and trying to impress the stupid public with new clock speeds, even though the gains in performance are marginal at best......
  • you know.. K7 has 200MHz FSB..
    and on module cache...
    and possible L3 on the mobo..
    and i doubt intel will do DDR unless forced because Rambus ain't workin, but K7 north bridges will handle it..
    and i did hear that sometime end of year, the K7 will be available in socket variety..

    hrmm.. :)
  • This kind of crap makes me sick. In case you didn't realize it, ars-technica are the people who did the article. if you think its crap, tell them. don't bitch about it to the /. people. They don't write the articles. I read the ars article before it was posted here, and although its not the best review, it is pretty good. The PIII is worthless as far as i'm concerned right now. I would much rather go out and buy a C300a and save ~$700 .. for god sake, I can practicaly build a whole new 300a machine for the price of a pIII... If you want to go out and buy one, great! but you're wasting your money.
  • I'd like to check this out but the /. effect has struck again.
  • Of course you aren't seeing much of an improvement with current apps. Guess what? CURRENT apps aren't written to make use of the new P3 instruction set, so of course you won't see a major improvement!

    Once hardware vendors start releasing P3 optimized OpenGL and display drivers, we will see a teriffic performance boost from P3's. And no, I don't work for Intel, but I happen to be writing P3 optimized stuff right now, and believe me it's really something :)
  • It's time some EE firm hired a molecular biologist to incorporate macromolecules into chips.
  • Ask them when they will be updating the PPro core to reflect the more competative CPU industry, ala the newer AMD K7 and Alpha 21264 CPUs. Do they believe in innovation and creative design? While their PIIIs make very nice space heaters(P2 actually, but close enough, under my desk... Mmmm, toasty), will they offer the capability to toast bread to varying degrees and perhaps keep my coffee hot and fresh?

    AS =)
  • Yeah, except for putting a media coprocessor on the chip it brings nothing new to the table.

    Like it or not, this chip will define Intel's high end desktop offerings. If you buy intel, this is where you are going to find your highest clockspeeds.
  • Your 4 yold P90 had a fraction of the bandwidth and much higher latency out of cache than a Celeron or a P III. There is more to cache performance than size.

    Also, the Alpha's are a different animal. They have an on chip L2 cache. The 2 & 4 MB caches are on the motherboard (or chip carrier) and have higher latency and lower bandwidth than the cache on the celeron or the P3.
  • What AMD has is late, but it deserves a bit more respect than you give it.

    1. They have Socket 7 chips that keep up with PIIs on cheaper motherboards, and have for some time.

    2. They have a 3d api that actually has shipping code written for it. KNI may be better, but it has come later and it has no support.

    3. The K6-3 is an incemental improvement, but given what they have done with an older bus-protocol and off-chip cache, I think it will be a competitive chip, when it finally comes out.

    4. The K7 looks very cool and it may well be the fastest x86 chip around for a while after it ships (provided it doesn't slip much further)
  • the point is that one buys a cpu today to use against today's software. in two to three years the cpu market will be very different than today and the merced III and amd k9 will make the lil-ol' p-III look like a toy.

    besides, in three years, the 3d accel boards will offload everything anyhow ;)
  • If Intel continues to build faster, but not better processors they will stagnate and become a has been.

    Can you say "IA 64"? I'd bet my annual salary that Intel's best minds are hard at work on their 64-bit architecture. The x86 architecture is definitely minor league at this point.
  • Some AC writes:
    Aren't the Celeron 300A and 366 on similar dies? Why does the 300A do so much better than the 366 o/c'd? I mean, everything else makes sense because the dies are different, but no that.
    Yes, the 300A and 366 are essentially identical. The only difference is the (fixed) bus speed multiplier used in the cache and core. For the 300A, it is 4.5; for the 366, 5.5.

    The reason this difference strongly affects the overclockability is that the fundamental speed limit of the P2 line is around 450MHz. Most P2s and Celerons can do that, or around it -- some higher: up to 500 or so fairly common (20%?); 550, possible but almost none. Lower is occasionally the case as well, with some of the 300As unable to get up to 450 (maybe, 20%). I have a 300A that does 450 stably (2.1v required), but it just will not do 464. A fine demonstration of the limit.

    Given that limit, the 300A is usually the champ. You raise the bus speed to 100Mhz, and it goes 450 -- 80% of them do, anyway. With the 366, the 100Mhz bus would yield 550Mhz core -- but almost none of them can actually run stable (or at all, really) at that speed. So, you can try using intermediate speed, 75Mhz or 83Mhz bus. These will most likely work, at least as far as the CPU is concerned. The problem here, especially with the 83Mhz bus, is that it yields a PCI bus at 41MHz -- too fast for many peripherals; 33Mhz is the spec they are built for. (At 100, the PCI bus gets a 1/3 multiplier, keeping it perfectly in spec.)
  • The 128Kb of cache that the celeron has is on-chip and runs at the same speed as the chip, thereby ensuring that it's *very* fast. The L2 cache for your P90 will be on the motherboard and run much slower than the chip.

    The thing with cache is that in most cases it follows the law of diminishing returns. You get much more performance increase going from 64Kb to 128Kb than you do from 128 to 256, and from 256 to 512 etc...

    The alphas are obviously aimed at server applications, where large amounts of cache are more useful than in most consumer applications, due to the huge amount of data they are required to handle. The improvement in speed that you'd get from doubling the cache on a PIII is minimal for most consumer applications. Especially when you compare it to the increased cost.

  • The first one took like 5 years to finish. Something tells me the third one won't be out in my lifetime.

    - A.P.

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

  • For those of you familiar with the SAT notation for analogies. Big, bloated, inefficient, and loaded down with dubious 'features'.
  • I am not saying that the added $14K is a burden for a large company. The point is that you are paying 35% MORE for a measly 14% performance gain, which is barely noticeable in any computer. You still pay the same premium for a 1-way box, btw. You bring up a good argument surrounding low price/performance for the low end. However, even for 1-way boxes using these new chips, the gains are minimal. If you need a decent 1-2 way server, screw the high-end chips. Buy something cheaper and PRACTICAL. And that's what I am getting at. There is no need to buy into this endless cycle of upgrades. Sure, the Xeon was a big improvement over PPro, but the incremental upgrades, in most cases, are a waste of money, for high AND low-end boxes.
  • ...but the k6 outsold the p2 last year. It is certainly already ahead of the p3. Those 50/5 figures might be backwards.
  • I was likening it to Windows98. Small benefit, full upgrade price == Pentium 98! Get it? Get it?

    Aw... nevermind.

  • This is the pentium processor that fits in the PC that runs the CDROM that enables you to play Monopoly with some dude in Italy.

    Here at intel, trained dancers are busy putting FUN into the processor! Shake it!

  • ...despite the label "workstation" now have PCI slots and use many commodity pieces - until you get into the "real workstations" and servers. I think Sun and hell, even SGI to a point are taking too many cues from the Wintel world. I don't care about commodity RAM. That's fine. Even IDE has it's place - but not in a Sun workstation. Then there's SGI with their new VW... That's a whole different league though, and I won't disparage SGI (they're going after the low-end NT graphics market, so their box makes total sense).
  • PC mag I think it was? Anyway , the article talked about Linus' super secret company , Transmeta, there are rumors that they are making a low-cost risc chip.
  • These situations aren't parallel, 'cause Intel doesn't
    do the OS.

    It's interesting to see how AMD and Intel have made
    a big deal of these "new" SIMD instructions, which
    appeared on MIPS and UltraSparc years ago.

  • They wanted to call it the "Bill", but decided it was too obvious. A little whiteout and an exacto-knife, an voila! The "P!!!" !

  • It is just like talk shows - you don't have to watch them. You don't have to buy an Intel P3. Is anybody making you? Is anybody rewriting old software so that the new version is locking out non-P3 owners? No, no. Does anybody care? No. Intel can do what they like, and I couldn't care less, but at the end of the day if your so-called "stupid public" want to buy them they will. No probs here, I know what is good for me, and what works for me. Stop complaining, it isn't a compulsory purchase (unlike windoze....).

    The whole issue here, I think, is an underlying bitterness that Intel, the proven market leader, is not supplying the minority demands - that is, it isn't catering specifically to the more informed section of computer users. It knows where its income is coming from, and it isn't going to do a U-turn purely because a few people don't like the way it is going.

    Ya got options, bub.

    -seizer, of the AMD.
  • best price?
    prices from Pricewatch's PC proccessor price listings []

    PIII 500Mhz = $747
    PII 450 = $469
    PII 400 = $303
    AMD K6-2 400Mhz = $139
    AMD K6-2 450Mhz (preorder, due out this month) = $279

    disregarding the fact that AMD uses super 7 (allowing for cheaper prices at equal speeds to intel brand PII motherboards),

    • 400/139 = 2.877 Mhz per $1 for AMD []
    • 400/303 = 1.320 Mhz per $1 for intel []

    So AMDs are less than half the cost and only slightly less powerful at equal Mhz. (not to mention the fact they're smaller and not run as hot. better design, imho).
  • Little bit of confusion here, I think.

    3dNow is AMD's MMX.
    The K7 is supposed to contain SIMD features similar to KNI, or Altivec.

    Correct me if I'm wrong.
  • Rambus, to offer an analogy, is like a big-rig truck going 90mph; faster and carries more in a day than a mini-van at 55mph; but it's acceleration/deceleration is crap.

    PC133 offers some solutions, especially with the ddr; read/write on rise and fall of clocks, so not only is latency reduced/minimized, throughput is simultaneously increased by 2.6 times as well.

    And it will be cheaper too.
    Gotta love Intel's marketing machine. Go read some articles on new tech and what competitors offer; Intel may be more reliable yes, but innovation from companies willing to take risks to outperform the incumbent is vital.

  • The really revolutionary thing about P-III's (the reactionary thing being the hardware ID) is that they have hardware NURBS instructions on the chip. NURBS are non-uniform rational B-splines. Intel's betting that someone will come up with a NURBS-based graphics format for the web--Adobe Illustrator, and to an extant PostScript, already use B-splines, which could be rendered with the same instructions. Then, of course, only people with P-IIIs will be able to view those graphics reasonably quickly.

    As soon as I have the money I'll cough up for one, though. A lot of buck for the bang, sure, but hardware NURBS support will be a huge help for high-quality 3D imaging, and the CGI landscape will look a lot different once 3DNow! and NURBS have been incorporated into all Intel clones. I'd love to write a program using the NURBS instructions for a senior project!

    P.S. Don't let Intel make you think they invented NURBS--the paper was published in the 80s.
  • Intel's new commercial states that the Pentium III is going to make the Internet MUCH better. Anyone have any guesses how this could be true? Surely not just a marketing tactic eh? ;-)
  • Posted by The Mongolian Barbecue:

    And why would these be better? Could it be that you simply thought that biological technology is en vogue, so it will be faster and more effecient? This kind of thinking is so prevalent... grrr
  • >OK, DOS 6.22 won't quite run, but who in their >right mind is going to put that onto a system >that is brandy spanky new?!
    I would! I wanna see how insanely fast One Must Fall would run!
  • It's pretty obvious that Intel is simply milking the IA-32 architecture for as long as possible while simultaneously working on IA-64. Who wouldn't extract the maximum amount of revenue possible from a development? If you know better, simply don't buy it.

    -- IG, a happy Cyrix customer

  • but Intel still sucks. :)
  • Errr... MIPS *are* RISC chips, ain't they? Unless there is an American TV commercial reference I'm missing somewhere.
  • Didn't you know? If you hook up your PIII to that cute little 110 acoustic coupler back there in the corner, you'd be suprised at how fast the Internet(tm) would be! :-)
  • We could always ./ the P3 site...

  • We should allow the technology-illiterate public to be duped by the Intel marketing machine into believing that the PIII will speed up their internet connections, and that they are the best price-performance chips on the market.

    Pricing based on actual value? Get outa here! Companies should be able to slide by on market share, advertising, and backward-compatibility.

    Whats the point in getting our collective nickers in a twist? Let's all relax, take some soma, and enjoy reruns of Friends! If you haven't seen it 2-3 times already its almost sorta new to you!!!

  • You must have a 3D solution before you have a fast 3D solution. When Intel or AMD accelerates all current OpenGL processes, then I'll take notice.

    Actually, MMX and 3D-now have been supported for a while, and people have been working on KNI drivers for a little while also (the figures I've heard state a performance boost of about 25%).

    Of course, as was pointed out, the graphics card will usually be better at accelerating graphics than the processor. KNI will be coming out just in time to meet graphics cards with geometry processing, which renders it useless.

    MMX is very nice for accelerating 2D graphics operations. If I wanted to do, say, alpha blending of a 2D sprite in software, I'd love to use MMX for it. However, we've had graphics cards that do 2D acceleration for quite a while now. So, MMX is just used as a way to move memory around twice as quickly (it has 64-bit registers which can be loaded or unloaded in a single operation, as opposed to 32 bits per operation for EX).

    3DNow is a nice solution for packing SIMD floating-point instructions into the old Intel register model. Unfortunately, this means that you can't operate on very many floating-point numbers at a given time, which makes real performance gains marginal (you have extra overhead for massaging the data into a form that can be readily fed into the new registers).

    KNI is looking like a somewhat nicer solution, as Intel has the clout to introduce new registers and make software vendors support them quickly. However, it's still a bit cramped, and will shortly become useless when on-card geometry acceleration is introduced into the consumer market some time this summer.

    So, while I agree that in the long term KNI won't amount to much, I think that you are incorrect about it and similar additions not being used.

  • Last I heard, wasn't VCR also doable on PC133?
    The other thought was that Rambus had latencies slightly lower than the fastest SDRAM, with the point of the comparison that you would only use the fastest SDRAM against RDRAM or whatever Rambus call's its memory.

    Rambus2 I may very well believe to be a useful technology, just as EPICs second generation of the newest IA-64 architecture will be better than current generation IA-32, PowerPC, or Dec Alpha. In the meanwhile, it is cheaper, easier and more effective to go from PC100-PC133, double the data rate by using both clock edges, using a slightly faster speed to increase the data rate to 2.667 times, and then use VCR and cacheing algorithms to further reduce latency times...

    Its really an argument of Intel pushing an expensive, proprietary, royalty ridden architecture slightly before it's cost is feasible, vs a slight evolution of the current design. It's the same argument that a 1GHz Dec Alpha or a 600MHz PowerPC would more than likely outperform the newly released 700MHz Merceds, next year.

    We'll see if industry standard PC133 or Intel endorsed Rambus succeeds.
    BTW, there are 2 Rambus standards; 600MHz and 800MHz. The 600MHz is for manufactureres and motherboard designers that can't cut the mustard, and can only go halfway =)

  • Once hardware vendors start releasing P3 optimized OpenGL and display drivers, we will see a teriffic performance boost from P3's. And no, I don't work for Intel, but I happen to be writing P3 optimized stuff right now, and believe me it's really something :)

    About a 25% speed increase under real-world conditions, from what I've heard elsewhere.

    This is definitely worthwhile, but it unfortunately won't mean much when graphics cards with geometry acceleration come out on the consumer market (this summer IIRC).

  • None of this will be particularly helpful to alternate OSes such as Be or Linux. But not many people will be using those on a PPC machine anyway

    Why wouldn't it be useful? Granted, it'll take a while for EGCS to support the processor (it's just now getting "official" support for the G3, and even that code's still in CVS). But once that's taken care of, Linux and BeOS will be able to use it (and if only Be would get off their butts and get support for the G4 or at least the G3, than man is that OS gonna fly...)

    I might add that the specs for AltiVec are already out there; Apple even has an AltiVec emulator for developers. Theoretically it's possible to start adding in the AltiVec support now.
  • Posted by Nr9:

    it is wrong to use SPEC- compiler dependent
  • by Sven182 ( 10469 )
    I think I read somewhere that Cyrix were planning to put two FPUs on their next chip to get decent performance. Anyone else heard this?
  • I've never owned any Intel _chip_, let alone a processor... looks like I never will ;)
  • figure if we're going this far off topic I might as well bring up a line only a true loser would remember from having to suffer through summers of boy scout camp.

    Hmm... the topic was the PIII. What a waste.
    It's all about AMD this year... I see their stock at least doubling by the end of 1999. Nasdaq is getting nailed now, but it will rally in another month, AMD will ride that wave up, and the popular press will cry "the King is dead, long like the King!"
  • People claim this vast amount of stability with the celeron 300A at 450. On the one hand, it handles great under linux (so of course I'm happy) but I find when playing games under windows, its a must that I run at the normal multiplier. I'm using an abit BH6 motherboard at a 2.05 core voltage if that makes any difference. I find it hard to chose sides as I own intel equipment, but have stock in AMD :)
  • http://www.tomsha []

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

  • Since basically all the processors that have come since the Pentium Pro (Pentium II, Celeron, Pentium III) are based on the same Pentium Pro design.

    The biggest thing holding back my PPro right now is the 66MHz bus. That's the only thing that would make me want to upgrade to a K7... faster memory access.
  • Maybe Intel should have named their processors after elements.

    They did that! Xenon. No, wait.. it's Xeon.. never mind ;-)
  • A fair comment considering that the processors at the heart of the Playstation and N64 are MIPS designs.

  • Apple started with a gigantic lead. Then they refused open source technology, wanting to build the machines, the OS, and everything inbetween. Along call Microsoft. They open sourced with Intel, IBM, Compaq and software developers. They all make money on Windows 1,2,3,4-1,000.
    While the government looks at Microsoft one can only wonder how the other members of the gang get to slink away with their tales between their legs.
    I'm not a programmer or hacker, just an average computer consumer and let me tell all open source, gift ecomony advocates reading this post...the consumer is sick of the Wintel, software ripoff!!!
    With the digital invention of Internet, UPgrades could be delived over the net, right? Not a $100 UPgrade at that. Digitally delivered software, UPgrades and who knows what all, at an affordable price.
    Now is the time for action. The computer is in enough homes that a truely new and innovative way to deliver the goods of high quality has a ready and willing market.
    I hope i'm seeing that need met with Linux, etc.
    The demand from the consumer will only increase.
    I'm 47. Maybe it's time to learn new tricks?
  • Posted by johnny the homicidal maniac:

    the 300a, clocked at 450 or even not over clocked, is a damn nice bang for your buck. feel free to add more L2 cache to processors, and you will experience the nicely curved Diminishing Returns graph. at some point, it just doesn't matter you have more. your p90 only has 2 ALUs, the Celeron has the pPro architecture, with 3 independant ALUs - meaning if one stalls the other continue - unlike your p90. the L2 cache is fullspeed at 300 or higher. if you haven't built one and run it, then shut the hell up. it is like people moaning about fords when they own chevys - but have never owned one. i have built two 300a boxes that just kicked ass and were cheap. i don't use them, though, i have my two pPro boxes, so i'm happy
  • Tom's right... it is going to be costly... but what did you expect from a chip that effectively merges x86 and Alpha architechtures? (granted it is only a first try) The entire reason for the 200 mHz front-side bus speed is the integrated Alpha technology. The cost comes not just from the chip, but the new motherboards and memory chip required to make it effective, not to mention plan to integrate anywhere from 512K to 8MB of back-side L2 cache. You will pay for features, and the K7 is merely AMD's stab at trying to grab a secure hold at the top of the performance heap.

    And while the price might seem outragous for those of us with little money to spend on such costly items, keep in mind that, as far as I have heard, AMD will continue to provide the K6-x generation of chips on the socket 7 boards (I believe that the K6-3 is in fact waiting for release, with all testing done).

    So, in my eye, I look at the the K6-x/K7 relationship like the PIII/PIII-Xeon relationship, except, of course, the chips are better... =)


    PS - I also have used nothing but AMD chips since my last Intel 386
  • I still remembers 199"Intel Outside" logo (Post Script format) created by Warren Toomey (24 March 1994).
  • Already done. What do you think the SGI VW basically does? Memory, video, and CPUs are all one the same switched memory bus. 1.6Gbps (may be 3.2Gbps) directly from memory to video. Same for Sun machines and SGI Unix workstations.

Q: How many IBM CPU's does it take to execute a job? A: Four; three to hold it down, and one to rip its head off.