Become a fan of Slashdot on Facebook

 



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

Willamette and Other IDF Highlights 54

Hoodoo Extreme writes, "There's a new issue out of the Private Eye which takes an interesting look at Willamette from IDF as well as some new info on RAMBUS. Later on in the piece there are some new findings on various 3D chips such as the S3 GX4-C which will follow the Savage 2000. Has anyone heard about this one? " IDF == Intel Developers Forum.
This discussion has been archived. No new comments can be posted.

Willamette and Other IDF Highlights

Comments Filter:
  • might pass the athlon. But then again maybe not. I think that AMD has been listening to people and giving them what they want while Intel has been riding their laurels for the past couple years. Intel had better be afraid.
  • I thought the big push from Intel was to switch to copper interconnects. Isn't copper supposed to increase speed at lower power? Or am I just talking out of my ass? Oh well, it is pretty late. Also, I'm glad to see that the Willamette is using a 400 MHz bus (Or a quad pumped data bus running at 100 MHz for the nit pickers)... wow.. that's pretty damn fast. Seem's like quite a jump from the PIII 133MHz. All in all, seems like another new crop of goodies that I won't be able to afford for a year or two... boohoo.
  • I'm reading it and I got to say that it sounds like a bunch of marketing hype for Intel (well it is their show). I hear nine MPEG-1 movies and think 'wow', but there's no explanation of how much to attribute to Willamette and how much to attribute to the video card (and the T&L problems). Try some MPEG-4 movies to impress and let us know what the chip's doing (real numbers).

    So looks like Will'll be out in Q3 of 2000 and probably called Pentium IV. And then there's Rambus:

    Not only was RAMBUS a 'Gold Sponsor' (that means they're in bed with Intel) at this year's IDF but did anyone notice their stock soar earlier this week (up over 50%!)? We met with RAMBUS' Steven Woo, who explained the sharp rise in stock being due to Intel's announcements that the Willamette, Tehama and Timna platforms will only run with RDRAM.

    Do we really want to pay 3x the price for RAM?

    Fortunately, there's AMD, who was "showing off a very impressive 1.1Ghz Athlon with the new "Thunderbird" core (on-die L2 cache)."

    Talk about power of 10! I think I'll be going with AMD from now on. //end rant [tired,sleep]
  • I'm not a big Intel fan (I laughed my ass off when I first saw Athlon benchmarks) but at least Intel seems to be doing something very new. That fast a bus will be a big deal.

    All Athlon contributed was a supposed 200MHz bus, but using ordinary (100MHz) SDRAM. Even Sledgehammer sounds like more of the same.

    -sig-

  • No mention of the Intel Itanium. I wonder whether this will ever see its light of day.

    With AMD hitting at its low end range of processors, and Crusoe bound to fight for its higher end range, Intel has to do much more than it is doing now to stay in contention in the chip market.

  • With AMD hitting at its low end range of processors, and Crusoe bound to fight for its higher end range, Intel has to do much more than it is doing now to stay in contention in the chip market.

    You mean that the other way around, right?

  • I don't see how anything you've described will lead to a lawsuit. Can you cite your examples? I'm not on the inside track of IP and trade restriction issues with respect to microprocessors, but the size of the chip seems doubtful as being restricted by EU import restrictions. Why would it be? Second, why would the EPA care about extra pipelines in a chip design? Is this an energy issue? How? Last, the issue of barratry, which as some of us probably don't know means "The offense of persistently instigating lawsuits, typically groundless ones." It sounds as though your argument is that Willamette will cause lawsuits because it causes lawsuits. Seems a bit circular, no? So, again, give us some hard evidence of Intel's non-compliance issues (the two you give do not seem very convincing). Then let us go about in peace (ie, stop with the profanity already). And if I'm wrong, I'm sorry. Your post just seems like a psychology experiment.
  • It does have a 200MHz bus, between the CPU and the Chipset. To my knowledge, they never said it would use 200MHz RAM. Althons still kick ass though! BTW: If someone offers you a Savage 2000, its a trick, RUN!
  • I've seen plenty of reporst from intel about how the willamtette is going to be 1.5 ghz, however, that's all, nothing along the lines of having fixed stability problems in that were in the p3 core. Much less this processor isn't for another 8-10 months... I don't remember hearing about the athlon like this 8 months before release. Maybe i'm wrong, I don't remember the exact release date that is set for the willamette, but i'm pretty sure it's no time soon. Where's any real benchmarking, where's the things that actually mean anything about the chip. Also, in general intel tries to kill you with chip prices. on a second note, this is a question. How many fpu's did were on the chip?
  • I know this is off-topic, but I dunno where to post it, and I wanted to put it somewhere where most ./ers would see because it's kinda funny.

    In this picture [xoom.com], if you take off the moustache, I swear that mahir "I kiss you!" guy looks a lot like CmdrTaco. It's scary. Go ahead and moderate it down if you want, but I can't think of anywhere else to post it.

  • Well, I quite liked the fact that Heresy was thrown in there.

    This is essentially the same thing as me explaining the average non-techy person how computers have pressure sensors in the keyboard that increase the processor speed when you hit the keys impatiently. It's amazing how many people believe me.
  • The other 25% are fiddles to make the numbers look bigger.

    2560 just seems a suspicious number unless there's a 256 bit pipeline with 10 stages and they're adding the stages together. They would after all be moving 2560 bits at the same time. This is of course speculation.
  • The Memory bandwidth you are quoting is for the graphics subsystem. Not the main memory. Main memory has a bandwidth of 3.2 GB per second, just like the Willamette.

    The 48 gigabytes of bandwidth does sound impressive but it is listed as "DRAM bus bandwidth" for the graphics. It isn't clear (on the list of specs) if this is part of the 32 megs of rdram or dedicated graphics memory.

    From 3dfx's website the Voodoo3, the Voodoo3 3500 has a peak bandwidth of just under 3 GB per second.
  • which is equal to the Willamet

    Which is damn impressive for a whole machine that costs less then the Willamet
  • I know, I probably shouldn't be writing this, but this is some of the best trolling I've ever seen. I mean, where does he come up with this stuff? The EPA restricting the number of pipelines on a processor?? That's classic!
    The only thing that could possibly have made it better is if it was a little more readable :)

    Anyways, congratulations on elevating trolling to an art form.

    (Who needs Karma anyway?)
  • I'm sorry, but I seem to remember a TV reference a while ago taking the mickey out of Demi Moore for how she wants her name pronounced - "Remember, it's De-mi, not Demmy..." Sheesh. Will-am-ette, not Willa-met?? I bet everyone's feelings will be really hurt if I mispronounce it.

    I propose calling the Willamette by an alternative name, maybe the Death AMD Killer 10000. At least that has style.

    And don't get me started on Itanium....

  • There have been a number of articles on Slashdot recently about the plight of RAMBUS, specifically how it isn't really delivering, poor yields, problems with chipsets etc. This article seemed to reference RAMBUS almost in passing; does anyone out there in the trenches have the real deal on what's going on? It seems Intel have really thrown their weight behind this, but is it a case of "If we don't follow through on this then our credibility in the marketplace is screwed?"

    It just seems that every time I here about RAMBUS the ante has been upped, and sooner or later something's going to hit the fan....
  • If you are referring to me, I was refuting the AC's arguments so that others would not have to waste their time on it. He did have a good start, but there was so much more potential if he had laid off the profanity and used available facts, like Intel confirms problem with chip sets [zdnet.com] and Taiwan VIA hits back against Intel's patent claims [zdnet.com]. I didn't even have to leave /. to demonstrate the weaknesses in what AC wrote. Not exactly a bravo performance.
  • Well, couldn't sleep, so here's some links I found over at ZDNet on the IDF:

    It's all about the Pentiums, baby! [zdnet.com] that is, the Willamette and Timna at IDF.

    AMD crashes Intel's desert chip-fest [zdnet.com] with its 1.1 GHz (1.116 MHz to be exact) Athlon demo.
  • I do not have modarator points right now, but if I did, I may have considered moderating this one up a little.

    Its language may offend some of us, and I don't want to encorage too many simillar posts, but OTOH, this one is just too hillariously funny to be ignored, so we might forgive the streetlawyer.

  • by Greyfox ( 87712 ) on Monday February 21, 2000 @05:30AM (#1257135) Homepage Journal
    It seems to me that Intel pulled all this stuff out of their corporate ass overnight. Up until just recently all we were hearing was Itanium, Itanium, Itanium. Nevermind that they're joining the ranks of 64 bit processing about a decade after everyone else (That's OK -- DEC released a 16 bit processor in 1970.)

    Anyway, someone must have looked up and realized that AMD was coming from one side and Transmeta from another and the two of them were getting ready to divvy up Intel's entire market between them. And while they're out there showing blue sky, here in the real world AMD is still leading them in terms of processor and memory speed.

    Isn't competition wonderful?

  • It takes years to design a processor. Rest assured that Intel has been working on Willamette, Coppermine, etc. for a long, long time. The x86 isn't dead yet. Intel has tried to kill it plenty of times before.

    --

  • by Kit Cosper ( 7007 ) on Monday February 21, 2000 @06:04AM (#1257137) Homepage Journal
    Once again there is someone trolling around with nothing better to do than stir up trouble.

    There isn't a Slashdot Giveaway

    Things like this would be posted on the VA website [valinux.com], not on Slashdot.

    My apologies to anyone who has been misled by this poster.

    --Kit

  • Intel is already handing out prototype Merced chips to selected developers and vendors. Even if it does not meet Intel's performance expectations, I expect that Intel will release it and market it for high-end servers that need a 64-bit CPU.
  • I'm always surprised at the amount of heat Intel takes on Slashdot. While in many ways Intel is just Another Big Scary Corporate Entity, most of the things that make the internet and computing so accessable can be attributed to Intel's ability to crank out successive families of CPUs, each new one providing enough of a speed increase to drive the prices of the previous generation down to silly levels. Without this cycle could anyone really have a half-dozen Unix-capable CPUs sitting around the house?

    The other thing I don't understand is what makes Intel so bad but AMD or Digital/Compaq the good guys. AMD I can almost understand -- they make a CPU swap-in compatible with Intel CPUs, but its not like they're giving them away for free, either. And Alpha systems are outrageously priced, I don't care how good the performance is, for $4-6k, entry-level (prices from Pricewatch) doesn't do much for the average hobbyist.

    And while I'm on the subject, when are we going to see SMP chipsets for Athlon?
  • Copper interconnects reduce power consumption and allow higher frequencies. Intel is probably wanting to hold off on using copper for the initial production run while they work out any other issues with the chip, especially since the Willy pushes into some new territory technology-wise (whether for better or worse). With their recent track record of "errata", I am sure they will want to keep any potential problem sources to a minimum as long as they can keep performance competitive.

    As for the bus, 400MHz sounds pretty impressive, but until memory technology catches up, that extra bandwidth will remain wasted (at least on single processor systems).
  • It's SPELLED Wilamette, but it's pronounced "MUD"!
  • Intel has already publically stated that the Willy will not be released at this speed, but no word of what speed it will actually be available at. Release is 2nd half of 2000. IMHO, that means you won't be able to buy one until this time next year.

    The core of the processor is pretty much new, so I don't know that the problems with the P3 will carry over or not. Probably will just have its own, all-new set of "errata". :)

    Intel is really screwing with people on this one. They have lost their raw MHz & performance leads to AMD. To cover this fact up, they are:

    1. Playing games with processor speed. "Your processor runs at 1GHz...well ours runs at 3GHz!" Of course this is only one small portion of the chip. Also forgetting to mention the numerous trade-offs they had to make, like 20 stage pipelines that will take huge hits on missed branch predictions.

    2. I am sure in certain circumstances the Willy will blow the doors off the Athlon, but these most likely will be with some synthetic, hand-tuned benchmarks. The PR group will latch onto to these as proof that your Internet browsing experience will be so much better.

    3. FP performance is going out the door, they are not trying to improve it anymore. For example, the 'FXCH' instruction which is commonly used in FP code is now taking a big performance hit. Intel is betting the farm on SSE, hoping they can get software developers on their side and ignoring the existing software base out there (forget good Quake framerates with this chip).
  • Actually, according to PriceWatch [pricewatch.com], SDRAM is DAMN close to 50c/meg, at $67 for 128megs, so if RDRAM is $8/meg, it's actually more like 16x more expensive...

    Also, when i read that Timna will have native RAMBUS only forced through a MTH so it can only take SDRAM, I was confused, as that cripples performance, until I realized, "Intel's doing that so Rambus will get a cut of the price due to royalties," which is probably due to exclusivity contracts with Rambus.

    Finally, it's interesting that Intel is pushing DDRRAM for servers, which are notorious for needing high performance, which seems to say Rambus can't cut it at that level, yet at a far higher price. So, in all likelyhood they know Rambus is inferior, yet are still pushing it for consumer-level. Gotta love that company!

    -----------------------

  • Simply because their FP performance stinks!

    They have always been significantly behind their RISC competition (especially Alpha). This is mostly due to the original design of the x87 instruction set and its stack-based register structure.

    In the x86 arena, they have always been in the lead. Not so much because their design was so good, but FP is something companies are willing to skimp on to make a chip, especially in the low-end of the market. Now here comes Athlon which showed everyone that you can still remain compatible to the x87 instruction set, but have signifiantly higher performance. Unfortunately, Intel has decided with the Willy to not try to compete with AMD, but instead shift focus to SSE. Time will tell if this works or not...
  • If, as some self-important trolls claim, trolling is an "art form," then it's not art on a level with the Louvre, it's like annoying street mimes who won't leave you alone when you try to take walk. And we all know what happens when street mimes get too annoying...
    (They get beaten savagely with olive loaves.)

    -----------------------

  • I don't know what you've been reading, but I've known about Willamette since before I bought my PII300 back in '98. Its been on the design table since even before then.
  • Not really, if you consider they can have multiple busses going to memory. (IE. You read the post correctly) They have a 2 1K busses and a 512 bit texture bus. Some simple elementary (or middle or high, depending on what state you're in) school addition and you get, wow! 2560!!! Isn't math FUN!
  • Just remember that with the T&L engine going to the card on the GEforce (and, based on its performance, probably on others as well) the flops of the machine, while still important, are less so; Now a lot of it is going to be about being able to push textures to the card. The CPU is going to primarily be doing physics and tossing data around. (Yes, I know there's plenty of other stuff.) Outside of games, FP math tends to be less important. How much FP does M$ Word do?
  • by Anonymous Coward
    Thank-you for using your big German word. Now go away you fucking retard. Find a school. Take an economics course. Learn supply/demand. Then take a law class and learn about anti-competitive monopoly practices and their harm to society. Finally, take some history. Intel is just Another Big Corporate Entity, the only thing scary is how people allow themselves to be educated by the marketing deparments and print ads of multi-billion dollar corporations.
  • I bet everyone's feelings will be really hurt if I mispronounce it.

    only the 1.7 million people living in the portland metropolitan area, through which the will-AM-ette river runs.

    it's not that hard. as the locals say: "it's willamette, dammit."

  • I would agree with everything you say except for your conclusion on why Intel is pushing RDRAM for the server market. I would say that it has more to do with price than any performance issue. 512 MB of RDRAM is about $4000. I think the vast majority of users buying that much RAM (which is not very large for the server market...) would rather spend their money in other areas.
  • According to this fascinating analysis [deja.com], RDRAM is inherently expensive to manufacture and test. These aren't teething pains, they're Bad Engineering.

    I'm betting on "Anyone But Rambus".

    If Intel continues to tie their fortunes to Rambus, they're in for lots of serious pain.

  • I believe that intel owns a significant share of rambus, which is a -propriatary- memory technology... Unlike other forms of ram, like SDRAM, no rambus owns "key patents" on the RDRAM architecture, so no company will be able to manufacture and sell ram to work with a system that needs RDRAM except those who -pay money to rambus-... One reason SDRAM has gotten so cheap is because it is a commodity, and massive competition between different manufacturors has driven the price pleasantly low... I think that this is a contemptible move on intel's part... They are clearly trying to create a monopsony on memory that is compatible with their systems... It is an ugly attempt at a hidden memory tax, on every Mb bought, and I hope it backfires horribly, as open chip ram chip [and bus] designs surplant it, and RDRAM dies a horrible death. ...Although I do value performance, I would not be willing to support an attempt to introduce a proprieatary technology into all systems... Luckily I feel better because I am writing this on a wonderfull K7 system [which I waited for for what seemed like forever, even though I needed more horsepower, because I didn't want to support the P3]... It was worth the wait... [And by the way, I do not think that any arguements here are based upon any implicit assertion that "all big cooperation's are evil"... most of us are technologists and love a free market... However, -some- companies, and the techniques they use to try to victimize their consumers, are evil, and I as a member of of the demand side of that market am proud to make ethical choices about what products to support [which really come down to self interest anyway, just with a longer view...]]
  • I believe that intel owns a significant share of rambus, which is a -propriatary- memory technology... Unlike other forms of ram, like SDRAM, no rambus owns "key patents" on the RDRAM architecture, so no company will be able to manufacture and sell ram to work with a system that needs RDRAM except those who -pay money to rambus-... One reason SDRAM has gotten so cheap is because it is a commodity, and massive competition between different manufacturors has driven the price pleasantly low... I think that this is a contemptible move on intel's part... They are clearly trying to create a monopsony on memory that is compatible with their systems... It is an ugly attempt at a hidden memory tax, on every Mb bought, and I hope it backfires horribly, as open chip ram chip [and bus] designs surplant it, and RDRAM dies a horrible death. ...Although I do value performance, I would not be willing to support an attempt to introduce a proprieatary technology into all systems... Luckily I feel better because I am writing this on a wonderfull K7 system [which I waited for for what seemed like forever, even though I needed more horsepower, because I didn't want to support the P3]... It was worth the wait... [And by the way, I do not think that any arguements here are based upon any implicit assertion that "all big cooperation's are evil"... most of us are technologists and love a free market... However, -some- companies, and the techniques they use to try to victimize their consumers, are evil, and I as a member of of the demand side of that market am proud to make ethical choices about what products to support [which really come down to self interest anyway, just with a longer view...]]
  • Also just announced at IDF was the world's smallest PIII motherboard [cellcomputing.com]. Yes, it's the same size as the world's smallest PII motherboard [slashdot.org] (3 X 5 X .7") and it's from the same company, Cell Computing.

    400 MHz in your pocket.
  • Probably more than you think... General key to seeing if something uses floating point is to look for a decimal point. Not a sure-fire way, but works well enough in practice.

    While M$ Word won't care if your CPU does only 1 FLOP/s or 10 billion FLOP/s (well, at 1 FLOP/s...) or even use that much, oftentimes programs use floating point.

    As for other than games, if you use a spreadsheet, that's a nice use of floating point there. Or perhaps sound editing (usually integer, but sines and cosines have an awful habit of being in [0,1], and FLOP/s are important when you want to do accurate fourier transforms). Or perhaps GIMP/Photoshop. Or perhaps apps like CAD/FEM(finite element modeling)/etc.

    Just my pointless bool bit[2]; .
  • You must be REALLY new to the computer industry. Willamette has been in design since late 1995, well over four years ago, and all new flagship processors take approximately that long to produce. It was not just thought up in response to AMD or whatever, as all of the clueless journalists will have you believe. Intel does a pretty good job of keeping things secret (no info on Willamette until last week, while AMD was hyping Athlon at least a year before first silicon).

The confusion of a staff member is measured by the length of his memos. -- New York Times, Jan. 20, 1981

Working...