Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
Hardware

Chip News To Crunch On 90

Posted by timothy from the why-thank-you-mr.-moore dept.
kupolu writes: "According to this article at Techweb, AMD announced last Friday that it is dropping its plans for the Mustang processor in favor of the new AMD-760 DDR-Enabled chipset. The Mustang was going to be AMD's entry into the server market, with it's amazing up-to 2MB L2 Cache." (Actually, from this article it's hard to tell if even AMD knows what's going on; tweezing apart the code names from the capabilities of particular products to be offered is complicated.) But on the coming-out instead of dropping-off front, proxima writes: "This story on Yahoo describes that Intel is releasing two new Celeron chips on Monday. One, a 733 Mhz model, will cost $112 per chip in bulk. A 766 Mhz model will cost $170 per chip."
This discussion has been archived. No new comments can be posted.

Chip News To Crunch On More

Chip News To Crunch On

Comments Filter:
  • Hell, I just bought a Thunderbird 750 for a C-Note.
  • Clock rate is like a heart beat.

    If I'm in a race with someone and they're heartbeat is faster than mine it doesn't mean s/he's quicker.

    I have a feeling that you are preaching to the converted here, not wishing to sound like a flame (but probably will anyway) Most of the readers on slashdot understand the differences that BUS speeds, the different implimentations of the x86 that intel and AMD have chosen (as far as transistors, &tc, not the actual set.) But what is needed is careful education of the public. mmm tough jobbie that one :(


    How every version of MICROS~1 Windows(TM) comes to exist.
  • Hey! I resent that :)
  • The P4. The first chip designed by MARKETERS!! Seriously, this chip is only being released to keep Intel in the AMD vs. Intel game. The P4 is not going to be much to be impressed with. It is sad however to see the Mustang go. In a server environment, a Dual, or even Quad's of these running at the 200MHz FSB would have been awesome. We have a few Quad Xeon boxes here at work, and they are just great. We just got some Athlon 850 boxes in, and they run our code faster than Intel's at the same clock speed. (Our code involves speech rec, database insertion and extraction, etc). Our code is very CPU-intensive, and those Athlon's just keep it pumping. But oh well, there still is hope. Dual-Athlon's won't be anything to sneeze at. I can hardly wait.
  • Rambus is indeed junk - however the idea isn't bad, and address fundamental problems of the SDRAM interface. By putting chips in series, a much higher clock frequency can be obtained, as the signals do not go out of phase so easily (the higher the frequency, the shorter the wavelength, the more phase variation is introduced by the same distance variation...) For platforms likely to be very bandwidth limited and not latency - such as a PS-2 - Rambus is quite likely superior.

    Additionally, let's not forget that rambus is also dieing because of intel's laughable chipsets, which need some fundamental rewireing to support such a different memory philosophy.

    Don't get me wrong - Rambus is a bad buying idea, and the company is highly associal - but the idea is a good, and eventually necessary one. The memory of the future might not be RDRAM, but it certainly won't be any form of SDRAM.

    e.g. 600MHz DDR SDRAM has half a wavelength of 25 centimeters - So that's how far apart signals will be. Signals certainly don't normally arrive at the same time in a clock cycle (this depends on where the memory is getting it from, whether a bank switch occurs, etc.) that's already a very, very, small margin in _theory_. And in practice it's going to be very hard - I'ld think. The fastest DDR SDRAM around is about half as fast, at the moment.


  • I think Tet's point involves price as well as performance.

    Basically, if your lottery tickets were priced based on the number of tickets sold, that you would know exactly if you were getting your money's worth of probability.

    Like, for example, the multi-state Powerball jackpot with a 1:8e7 chance of winning, with maybe another factor of 3 or 4 thrown in for the annuity boondoggle effect of the prize that is defined as trickled over 20 years - It's not worth paying for a $1 chance unless the jackpot gets in the 2e8 territory.

    IIRC, there were actually investment clubs (in Oz?) that would wait for jackpots to get sufficiently high and then run out an purchase a copy of every number.

    With processors, I suspect your best performance ratio will lie in the low end. A friend of mine advocates buying the leading edge system for your new computer, but delayed 6 months from the time they were defined as leading edge.

    Next month I may replace my 200 MHz K6 with an 800 MHz Duron, but spend more of my upgrade budget on a new graphics card than the motherboard and CPU combined.

    So, after two decades have gone by, my earliest lessons that CPU's are the holiest parts of computers have become obsolete, I take your advice and walk from this piddly sandpile.

  • Transistors are so small these days that the primary limiting factor among these is interconnect length.

    Some friends of mine are doing an interesting senior project where they use signal reflectance to determine state of memory. Since virtually no current is used to read the state of the device, this memory could be very very fast to read...

  • Yes Carl Sagan and BHA were the code names for the 7100. Of course the other code names for the first Power Macs were Piltdown Man (6100) and Cold Fusion(8100), two famous hoaxes. I can't really blame Sagan for objecting. What I think was excessive was sueing Apple and forcing Apple to change the name. Then again BHA supposedly meant Butt Head Astronomer. Sagan sued again, but the judge ruled against him. It was probably a good thing that the next PowerMacs were code named Catalyst, TNT, Nitro and Tsunami.
  • How does that work? If I lost a little on each chip, I would most certainly not sell more chips to increase my losses.
    They sell in volume, huge volumes with a smaller margin on each chip. Those dollars and cents add up quickly when in the millions of chips.
  • > All i see this as is a Plot to make my new mobo obsolete asap.

    KT133 boards should be compatible with future Mustang-based Athlons. Only the Mustang server has been axed. The Mustang variant knows as "Palomino" will be called Athlon and the "Morgan" will be named Duron. They will work with current boards, though the Palomino's multiplier will usually be fixed for a 266/133MHz chipset connection, so you may have to change its multiplier to get it to work on your board.

    BTW, the Hammer series won't really be out until 2002, and there will in fact be a 130nm version of the K7 released for you then, too. :)

    -JC
  • Transistors are still a big factor. To manage power consumption (heat) lower voltages have to be used. This reduces the gate signal which reduces the on current which increases the rise time. There is lots of work still being done on transistor developement. Small gates are required to keep the gate capacitance low, which reduces the channel size which limits the drive current. The small size also makes uniformity of the critical deminsions very difficult. Lower operating voltages makes gate sensitivity critical and dopeing the silicone to work with low voltages increases the leakage current. It is a big balancing act to get it to work. Pushing the limits makes the path to functionality a very narrow tightrope.

    The senior project sounds interesting. It sounds similar to Intel's Strata Flash memory. The memory holds 3 charge levels per cell instead of just on/off for higher density. The cell is read by checking the e field of the stored charge. The Strata Flash memory is low power memory and is used in cell phones and such.

  • It looks like chips in the Sledgehammer family will be named after hammers.

    So keep an eye out for:

    the Jackhammer

    the Hammer Drill

    the Hammer Toe

    the Rodgers & Hammerstein

    the Hammerhead

    the M.C. Hammer

    the Jan Hammer

    the Hammered Dulcimer

    the Rubber Reflex Hammer

  • I want my SMP Athlons (Score:1)

    by Anonymous Coward
    I've been holding off on a hardware upgrade for several months now, in the hopes of going to a dual athlon configuration early next year.

    So, for a cunsumer, this sounds good -- more resources targeted at the SMP market helps me.

    But for higher end servers, this is not so good. They need the bigger cache and quad CPUs.

    Is the 760 chipset dual CPU only? I can't get to the AMD website -- does anyone have a short list of 760 features and a forecast of when the mobos will probably be out?

  • Um, I could be wrong, but I really think they're named after WWII-era fighter planes, not cars.
  • because they can concentrate on the 760 chipset with it's DDR memory and multiplocessor facilities.

    You know the saying : "A bird in the hand worth 2 in the bush"
  • They didn't keep it inexpensive, they charged you way too much in the past...and they'll keep charging what you're willing to pay. Somebody give me one good reason why there's no 4 GHz processor available at this moment...

    If I ever meet you, I'll Ctrl-Alt-Delete you.
  • 766 MHz is not such a big difference over 733 MHz.

    Indeed. I was wondering why anyone would buy the faster chip. 4.5% more clock speed for over 50% more price. They're both using a 0.18 micron process, and there was no mention of different cache sizes, so I can't see why anyone would spend the extra.

  • Actually, that's a paraphrase of a famous Winston Churchill quote (uttered after he was criticized for ending a sentence with a preposition). For details, check this [winstonchurchill.org].
  • But will they attract anyone beyond the 5cr1pt k1dd135?

    Not unless the "knmowledgeable people", meaning you and me, tell the businessmen & women to be attracted to it.

    If I ever meet you, I'll Ctrl-Alt-Delete you.
  • get the news directly from AMD:

    http://www.amd.com/news/prodpr/20161.html
  • I heard that they started out as classic cars, and then some carmaker got upset. So given the names they had, it was fairly easy to change over to horses.

    Anyone remember Apple's "Sagan" project codename that was changed to "BHA" after the namesake objected? Sounds like someone didn't learn from that lesson.
  • One major reason 4 GHZ processor's are not available at the moment is...(there is more than one)

    All wires have capacitance, inductance and resistance. To change the voltage on a wire (change the charge), current must be applied to it. When the current is applied, it takes time for the voltage to change. The amount of time it takes is determined by how much capacitance has to be charged and how much current is avaliable to charge it. It also takes time for voltage to get from one end of the wire to the other. The time for travel is figured by the capacitance of the wire, the length of the wire and the inductance of the wire. To reduce the time, (inverse of speed) the

    (1)transistors must be improved for more current

    (2)The wires must have lower resistance

    (3) the amount of capaticance of the wires must be reduced. (using low K dielectrics)and

    (4) wires must be shorter by using smaller transistors

    These are all items that research and developement do constantly. IBM is using copper instead of aluminum. Intel is working on the same. Low K dielectrics are being used. Lithography is pressed to use smaller features so the parts are closer and more of them fit. Also the clocking on the chip has to be uniform. If you want to see some of this stuff in action, Look up the wavelength of 4 Ghz. 1 wavelength away on a wire isn't very far. A wire with a signal caries a signal slower than the speed of light. A clock signal on this wire would be one full clock cycle late at the other end of a wire one wavelength long. At 4 Ghz, how far away can your cache memory be before a memory read request and clock get to it and the data return without being late? 4 Ghz signals tend to treat wire as an antenna (it is a radio frequency). Signal tends to leave one wire and show up on another causing errors called crosstalk. All of these issues are the things that have to be worked out to produce a faster chip. Oh by the way-add heat and power distribution to this dense mix of transistors. That is another couple of issues.

  • You're kinda right, but

    Linguistics is _descriptive_ not _proscriptive_ or _prescriptive_.

    The proof of this is the existance of the words proscriptive and prescriptive to mean the opposite of each other. (Both roots break down to "written in advance".)

    So there is no absolute "right" as such.

    FP
  • hehe! perhaps I read too many action catalogues and magazines past their read-by date;)
  • Given that Steve Balmer's horse is called "Pentium", this hi-tech horse fetish thing seems to be spreading!
  • From following the AMD news sites, I believe the real version of this confused story (replacing a processor with a chipset???) is that AMD are dropping the server (large cache) variant of the Mustang in favor of concentrating on the Clawhammer (first member of their 64-bit x86-64 Hammer family, indended for 2 or 4-way SMP) as their first server targetted processor. However, they will still launch the Mustang based Athlon desktop replacement (Palomino) and the Duron version sometime fairly soon (Q1 2001?).
  • That sounds highly ineficient

    ---------
    Defraggle
    Keeper of the monkeys
  • That's about the best use for a P4 without a board. Of course, I would wait for the first DDR chipset which would uphold a reliability reputation similar to that of the 82443BX (no significant driver revisions, can withstand 133MHz with proper cooling, et cetera). Of course, that probably won't be for another year, at which time Intel will finally violate their Rambus contract, or NVidia will release their chipset.

    Other obsolete hardware uses: hard disk platters make great coasters, and the magnets inside make really strong refrigerator magnets.

  • No, You see the more money you loose the company is doing! With a small loss margin, d a very high volume the p/e ratio of the stock will be imposible to figure due to a divide by zero error and therefore amplify the stock price by 100 fold.

    ---------
    Defraggle
    Keeper of the monkeys

  • AMD dropping Mustang now is Great (Score:1)

    by Anonymous Coward
    For AMD to drop Mustang now shows that AMD is constantly adjusting itself to the shifting marketplace terrain.

    At this point in development, AMD knows that the processing power of the Athlon core is heavily dependant on the latency of the L1 cache. Much more so than the PIII. The instruction pipeline cannot feed information into the core fast enough.

    My understanding is that the Mustang/Palamino group was supposed to increase the accuracy of the branch prediction AND increase the number of instructions available to the core by enlarging the cache.

    The downside of a large L1 cache is that a larger L1 cache has a larger latency (to demultiplex the addresses) than a smaller cache.

    After running several tests in simulation and possible early silicon, AMD probably realized that a very large L1 cache does not add value to the processor, especially when coupled with DDR RAM. Instead, the money and research should probably go toward the other, more promising research.

    The extra cache is bound to increase the price quite a bit when you consider the increased probability of lower yields.

    Think about it cost-wise, would you prefer a 1.2 GHz with 1 MB cache, or two 1.0 GHz with 256 KB for the same price?

    There are OTHER x86 companies out there that stick with technology for too long regardless of the performance, and wait to change only after the market has incurred the cost of their mistakes.

    Score one for AMD.
  • Hum, hello? A few months ago, I read some articles on Cnet/ZDNet about a lot of people abandoning Xeons for regular PIII because because the performance/price gap was rarely worth it. There *is* a strong market for x86 CPUs with lots of cache but prices ahve to go down because there usually are less costly alternatives.
  • Thank God someone out there with half a brain. I believe that you are absolutely correct. Rather than have a "high end" 32 bit processor infringe on server territory they intend to fill that area with the 64 bit hammer stuffs.

    I'm a little disappointed with this course but I can understand where they're coming from. In early 2001 I will be building a new computer and was drooling over the idea of dual Mustang-core Athlons with 2 MB L2 caches. (The new computer is going to be an all-out beast.) I don't know if I can wait for x86-64 but it sounds so good. Anyone know what quarter they want it to ship in? Maybe I could hold out 'till Q2 2001 but beyond that it gets rather painful.

    Some trouble with such large amounts of cache on-die are: larger dies, higher power requirements, and much more heat. Larger dies cut manufacturers' yeilds two ways:

    1. Less cores/wafer
    2. Higher failure rate.

    Prices would likely be sky high for a processor with 2 MB of L2 cache on-die. I doubt many people/companies would pay the premiums with x86-64 around the corner.

  • There's a lot more to it. AMD's planned "server" chip would have contained 1-2(!)MB of level-2 cache. That's an big feat - very difficult to achieve.

    On the other side, AMD has very fast processors, and a multi-processor capable chipset(using DDR memory) is on its way.

    So, as far as I can see, AMD decided to a) not risk everything on a chip that would be difficult to design/manufacture, and b) rely on their current, fast, processors and the soon to be released multi-processor, DDR capable chipset.

    Add to the mix that we might be seeing chipsets with ~8 megs of integrated cache within the next year or so, and you see why they might not want to "waste" time on a "server"-oriented processor.

    Dave
    'Round the firewall,
    Out the modem,
    Through the router,
    Down the wire,
  • I believe Clawhammer (x86-64) isn't due until the end of next year at earliest. Still, we're getting to the point now where processor speeds arn't the bottleneck for too many tasks...
  • I realize the point involved price/performance and you are correct with the above. It really just reminded me of the lottery paradox and pile of sand thing =)

    Woz
  • Its not a terrible way of doing things. Translation to risc is quite trivial compared with even something like a 64 bit adder. It makes the rest of the chip much simpler though. The other benefit is that you have a RISC core that can be sold separately for embedded systems and whatnot. I have no idea if AMD do this, but the option was probably considered in the design.
  • No, because "up" is also a preposition.

    So to be colliquially correct in the passive voice, you would say "This is a situation I will not put up with", while to be 'strictly' correct in the passive voice, you would say "This is a situation up with which I will not put."

    The point of the "up with which I will not put" statement is to point out the artificiality of "strict" English grammar, which is modeled after that of classical Latin.
  • I wonder what kind of overclocking blocks these celeron's have.

    bum bum bum bum
    Clocklocking inside

    ---------
    Defraggle
    Keeper of the monkeys
  • What you do it loose a little on each chip and make up for it in volume.

    ---------
    Defraggle
    Keeper of the monkeys
  • Shouldn't the title read, "Chip News on Which to Crunch"?

    Connah
  • I believe that because of AMD and Intel's increased frequency of releases, the companies loose a lot of money because of they sometimes have to cancel public releases like this. In the end, the end-user must pay for this, thus increasing CPU prices.

    This is not healthy. Another view is that this race might lower the quality of chips, just like Intel's drawback recently.

    So listen to my words, Intel and AMD! Sit back in your chairs and relax. Have a cup of coffee.. :-)

  • Once again we have another new chip. Once again, its just a really really fast 386. One of the least future proof chips ever. Only capable of handling 1 multiplier, so more clock speed is required. Faster clock means more pipeline stages. More pipeline stages mean less efficient branches. When will AMD, Intel and the rest produce a decent new architecture?
  • I'm hardly a hardware expert ... actually, I was an English major. So pardon my ignorance.

    As I read over the aforementioned article, one question lodged in my brain:
    Where do the names for these chips come from?

    The following monikers were mentioned in the article:

    Morgan
    Mustang
    Athlon
    Palomino
    Clawhammer

    I suppose the typical focus groups were convened, and chose names that connote speed, agility, and -- in the case of "Clawhammer" -- driving, forceful impact.

    'Else someone at AMD has a severe horse fetish.

    Sincerely,
    Vergil

  • Mustang, Palomino, Clawhammer, Morgan, Athlon

    And What The Guy About To Buy The Computer At CompUSA Says:
    "So, this one is 733 and this one is 766.., I want the bigger number one. Well the 766 has got to make AOL faster than that crappy 733"

  • It is called RISC

    ---------
    Defraggle
    Keeper of the monkeys
  • This Moore's law has to hit a limit someplace, doesn't it? The next generation of even faster stuff is already on the map. Maybe 2 Gig by this time next year. Wow. How do they do it and keep it innexpensive?
  • This is a situation up with which I will not put.
  • Please keep in mind that AMD or Intel waste their effort in supplying us with such processors. 766 MHz is not such a big difference over 733 MHz. Even without taking into account the other components of a system, 733 is 4.5% increase. I think that these companies should direct their efforts towards improving bus speeds and memory. If we had a processor with 500 MHz speed but a memory bandwidth of (let's say) 10 GB/sec, then that would be speedy!!!
  • Athlon is meaningless, according to the story Slashdot carried when they announced the name.

    Instead of naming things in the Athlon line after rivers (like Intel does with its chips) they name them after breeds of horse.

    It looks like chips in the Sledgehammer family will be named after hammers.
  • then why did you take the time to make a message saying that you didn't want the moderators to kill you?
    Fear my low SlashID! (bidding starts at $500)
  • Because, with current processes and materials it would be a very large ineffecient overheated piece of crap. They have to increase the effiency of the chip before they can start to push that speed... eg smaller transistors, better heat sinks, possibly even better materials... galium arsenide anyone?
  • > You know the saying : "A bird in the hand worth > 2 in the bush"

    You must be a better shot than I am... I need about 20 birds in the bush to equal one in the hand...

  • On Topic!!! AMD "not knowing what they're doing" (Score:3)

    by fatphil ( 181876 ) on Monday November 13, 2000 @03:55AM (#628302) Homepage
    The way I have interpreted several of the articles regarding AMD "dropping" musthang (just made that up!) is as follows.

    "We have decided that we don't want it in the _marketplace_, but if we need it, we could probably roll it out without too much delay".

    It's probably because they don't want to have a smooth transition to their next generation chips (the 64 bit ones, my what a novel concept*). If they try to flood the high end market with Xeon-bashers, then they'll not have any market for the 64 bit chips. Better to wait and then ship Xeon-anihilators, so there's no confusion about what to by and when to upgrade.

    Rant rant rant, all I want from them is SMP, rant rant rant!

    FP
    (* Alpha owner for _years_)
  • Well, how about the IA64? And is it really AMD's fault that Microsoft refuses to support consumer Windows on anything but x86? And, as others have pointed out, the h/w is whatever Intel and AMD feel will run well -- the point of instruction set to abstract the h/w behind it to allow for improvements. Finally, 3dNow!, MMX, and SSE all allow multiple multiplies per clock-cycle.

    -_Quinn
  • This Moore's law has to hit a limit someplace, doesn't it? The next generation of even faster stuff is already on the map. Maybe 2 Gig by this time next year. Wow. How do they do it and keep it innexpensive?

    Moore's law doesn't have anything to do with speed.
    It has to do with doubling the # of transistors that can be fit onto a chip.

  • Re:766 not a big difference over 733 (Score:3)

    by Nagash ( 6945 ) on Monday November 13, 2000 @04:07AM (#628305)
    This is a classic statement/argument that reminds me of the Lottery Paradox.

    Take n people who have bought tickets for a lottery. Look at the first person. The chances are so remote that they will win that you can almost say with certainty they won't win. Same for the second person, third and so forth up to n. Well, you've covered all the people saying that they will not win, but you know for a fact that someone will win.

    Another thing this reminds me of is the pile of sand problem. You have a pile of sand up to your head and take away a single grain. It's still a pile. Keep doing that. When is it not a pile?

    Same goes for processor speeds. 766 is not a big difference over 733. 733 is not a big difference over 700, etc. However, 766 is a big difference over 333 and that is why they keep pushing the envelope. Not to make recent developments seem worthy of an upgrade but rather, older developments.

    So while it seems every small processor upgrade is not a big difference, there is, at some point a big difference and that pile of sand is eventually not a pile.

    Woz
  • Don't quite understand the reasoning for it, but the bulk pricing on the processors is always quite a bit higher than the "street" price. My guess would be that because chip cycles are so fast, companies need to dump their supplies in order to make room for the new chips that are always coming out. Having lots of yesterdays technology in stock that you can't get rid of, is not a good thing for companies, but it's a good deal for consumers.
  • This is probably an error. A slightly older story [techweb.com] on the same site quotes final quarter of 2001 for first samples with production in the first quarter 2002. The AMD 64 bit [amd.com] FAQ agrees.
  • Rambus doesn't support higher frequencies due to serialisation, but because they use transmission lines to transfer the bits.

    While clock synch is an issue, this isn't what constrains the speed of traditional memory interfaces.

    sdrams modules (both traditional and ddr) use a simple node driver on it's output and don't do any tailoring of the waveform or impedance matching. This is fine as long as you keep most of your signal energy at longer wavelenghts than the length of the wire (or more precicely pcb trace) so that you can fully charge the entire trace to the voltage you want. Speedwise this technique got pretty hard limits on ow fast we can go, which cannot be resolved by just improving processing or circuit technique.

    However as we increase the datarate pr. pin it become infeasable to shorten the trace enough to do the transmission this way.

    Enter the transmission line. On this kind of interface we can have multiple bit 'on' the wire at the same time. This does however tight control of the wire's analogue properties and the waveform issued by the driver. When we do this we must limit the number and severity of bends and corners on the wires, and both sender and reciever must be designed to spesific impedances.

    Since an TL interface is expensive both on die and pcb resources rambus opted to serialize the architecture to reduce costs and and increase capasity.

    Eventually we will have to go to TL interfaces, not only to ram but just about all interfaces on our computers. However, right now it's apparent that TL tech isn't quite there yet. When they can push the speed up to say 2GHz we might want to reevaluate, though hopefully without rambus' disruptive influences.
  • I always thought Athlon and Duron were Klingon starship classes.

    Okay maybe I didn't but it almost makes sense.

  • You're rteplyiong to someone with a who's had a 64bit processor with 4MB of cache for several years. 1-2MB just doesn't impress me any more.

    FP
  • Well, my guess is that dropping the Mustang is mainly a response to Intel's strategy shifts in the server market.

    Traditionally, you could get a cheap Intel 2-way SMP board like the BX or (bleh) 820, which was great for the server and 'workstation' OEMs like Compaq and IBM, as well as better mb shops like Asus. But no longer - now one has to get a 840/Rambus model. And worse, the Pentium 4 will be introduced with no SMP support, which implies that they will be forcing all SMP users to buy (emasculated) Xeons (for big $$$).

    The net effect is that Intel is jacking the prices on SMP systems. (Consipriacy theory - perhaps this is to make the Itanium look more affordable.) So, AMD has a huge market opportunity here to take the 'low-end' SMP market away from Intel with the 760MP and standard cheap Athlons. Now, they just need to win some OEM bids for "departmental servers" and "workstations", and if they are technically successful, they'll be in much tighter with the big OEMs (who don't totally trust them yet.)

    Plus, there's a bunch of production logistics with producing a high-margin, low-volume Xeon-like chip. Furthermore, you pretty much have to have a top-tier server OEM to move those systems, which AMD doesn't have right now.
  • Just buy a 733 and change your bus speed to 70MHz from 66. Voila! A 770 MHz processor for the same price! If Intel can squeeze 766 out of the same process that they're using to produce 733s, there's no reason 733s cant run at 766.
  • Maybe AMD should label their products differently: 650 Duron (Performs like 766 Celeron)
  • Now, I don't mean to offend, but what exactly, other than teaching, is an English degree gould for?
  • Hell, I just bought a Thunderbird 750 for a C-Note.

    I wonder what it was remarked from.

  • My explanation is always stride length vs. stride frequency, but that's just 'cause I did xc : )

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

  • Nope, clean as a whistle. According to Tom's Hardware that's about the going price for a legit 750 T-Bird (slot A)

  • When will AMD, Intel and the rest produce a decent new architecture?

    Why should they? That's not how they got where they are today.

    Imagine you're on the Intel board of directors. The president meets with you, and say, "Some guy on Slashdot says we should make modern chips." There's some nodding and hmming, and then someone asks, "So, what kind of profit can we expect from these two different types of products?"

    The president answers, "Well, we can make about $1.98 by selling modern fast chips that will run rings around our 386s. But by selling a slightly faster version of the 386, we can make oh .. approximately ... 1.666 kajillion dollars."

    "Ok. Let's go with that 386 idea, and leave the modern processor ideas to some other sucker."


    ---
  • Continue reading the series.. Another Leto II is born.

    Fear my low SlashID! (bidding starts at $500)
  • Hmm no thanks, i think i'll buy a duron 600 for $44 and O/C it to 900 MhZ and it'll STILL whip the pants off of any celeron.
  • This is a good decision by AMD to drop the Mustang. The high end market does not yet trust AMD, and with good reason, considering certain fiascos in the past.

    And the high end market is wedded to Intel, these people don't care about k00ln355, all they want is rock solid stability and trustworthyness. AMD would find it next to impossible to break into such a market, even if they had superior technology. The only way AMD can do it is by creeping gradually there, step by step. But will they attract anyone beyond the 5cr1pt k1dd135? Businessmen & women? I don't think they can at the moment.

    KTB:Lover, Poet, Artiste, Aesthete, Programmer.

  • I didn't say anything about modding me down. I said "Don't kill me". I was appologizing for making a breach of etiquette.

    ------
  • You're selling it? Mine even lower so bidding on this one opens at $5000 ;]

  • Err... shouldn't that be "This is a situation with which I will not put up." You can improve it further by switching to the active voice and saying, "I will not put up with this situation."

    The original poster was right, it should be "Chip news on which to crunch." In spite of that, most people would agree with timothy and say "Chip news to crunch on." But just because most midwestern americans (myself included) use poor grammar like that, doesn't mean it's right.

    However, one has to ask, does anybody really care?

  • That's not what they said. (Score:4)

    by Anonymous Coward on Monday November 13, 2000 @01:51AM (#628325)

    Yes, AMD said they were dropping the Mustang from their roadmap...

    ...But not in favor of the AMD 760 chipset. The 760 DDR chipset has already been announced and will
    probably ship sometime late this month/early next month.

    What they said is the 760 MP (their SMP version of the DDR chipset) will fill that niche.

    The reason that probably prompted this is that AMD is short on fab space, and since they're selling out all the processors they can make, why divert capacity to an expensive low-volume chip.

    What it says is that the SMP 760MP is on track.

    This was a good move by AMD.
  • No, I keep the monkeys.
    The fraggle rock was lost long ago and has yet to be found. Only remants remain.

    ---------
    Defraggle
    Keeper of the monkeys

  • What is Intel Thinking? (Score:5)

    by dgb2n ( 85206 ) <.dgb2n. .at. .yahoo.com.> on Monday November 13, 2000 @04:32AM (#628327)
    $170 for a 766 Celeron in bulk? Even if they plan to release the chip on a 100 MHz bus, the price is way out of bounds. Take a look at the current lowest pricewatch numbers on processors:

    AMD Duron - 750 $74, 800 $97
    AMD Thunderbird 850 $142, 900 $165, 950 $222
    Intel Celeron 700 $76
    Intel - Pentium III 667 $149, 733 $173,750 $181, 800 $181

    I'm not trying to start an AMD vs. Intel war but Intel's current release and price point doesn't even fit into their own chip lines let alone compete with AMD. Get real. You can get a 900 MHz Thunderbird for less money.
  • IIRC the AMD code names are based on classic cars. The current Athlon chips were called Thunderbird during their development cycle and Durons were identified as Spitfire.

    There was a bit of a fuss not long ago when AMD had to change the name of their future mobile Athlon/ Duron processors to Morgan (a hand-built british car with wooden chassis) and Palomino because Chevrolet moaned about the use of Corvette and Camaro.

    The names marketing uses to sell these chips (Athlon, Duron) are products of pure fantasy so that they can be trademarked.
  • Don't forget.. the motherboard will cost more ;)
  • The reason for the varying speeds is both technical and historical. Originally, you had a chip design with a theoretical best clock speed. However, it was unlikely you would be able to produce all your chips to meet the spec, so any chips that wouldn't clock up to the max (say 100 MHz) would be tested at 90, then 75, etc. Nowadays, production seems to be more consistent, so I imagine the reason for this is simply marketing now. That would explain why it was imperative that Intel lock their clock multipliers- because all the chips are the same anyway.
  • The issue of a better architechture is not as simple as you make it out to be. It is not only that there are large investments in x86 and great market opportunities due to the large customer base, but also the fact that so much research has gone into them that indeeed the x86 is the fastest architecture on the block for many applications.

    Specifically, in integer math the x86 chip is no doubt by far the best - even more so as its rather cheap compared to competitors.

    Many of the big x86 problems - small number of registers, bad bus architecture, not a load/store memory system - are greatly alleviated by good caching, in which case the x86 needs to compete on the microarchitecture level not the ISA.

    I agree that x86 is a bad thing, but it is certainly not a killer.

    Furthermore, I think its important to face the fact that software developement cost is probably the biggest expense around, and it is simply better to have a compatible chip for old software than it is to get a new thing - even if the new thing is a bit faster - because speed is just one factor in the whole equation.

    Change should come - but the x86 shouldn't be abandoned, rather a transmeta-like approach of emulation is called for.
  • I've noticed something:-
    Retailers here in the UK don't seem to be selling many Duron machines (I mean the all-inclusive computer deals with "this amazing machine comes with a printer, webcam, obligatory scanner and pants speakers, not to mention tons of free software"(heh))
    Either retailers have signed up contracts with Intel saying "our special deal machines will only contain your cpus for the next X years" or.. They're scared of Amd's minor incompatabilities with various bits of hardware - or..
    On the other hand, they might just have a ton of Intel stock which they haven't sold yet, we've all seen how pc sales have slowed.
    HA HA!! Look who's laughing now!
  • I've had enough of those pathetic 66MHz Celeron releases. When will Intel actually release their 400MHz FSB (in the same definition as the Athlon's 200MHz FSB [which is actually 2x100]) P4's? And this time, let's forgo the Rambus chipset. No one in their right mindset would even think of getting a proprietary RAM standard that has already been proven inferior to the de facto standard, SDRAM (among the morons who buy RAMBUSt: CEOs who blindly purchase workstations, and every single PS2 owner).
  • If I understand you correctly, you want AMD and Intel to relax the tempo a bit.

    No way!

    Speed increases may be more incremental at the moment, but on a per year basis, this competition has really sped up developement. Also, an unstable chip is not hell (though intel should really have known better, considering the 1133 failed consistently on a linux kernel compilation :-)) - You just don't have to buy the newest of the new.

    An interesting point you make is that companies lose money because of retracted releases. On the one hand, the P-1133 retraction did cost intel (not much) money, however AMD's retraction cost them nothing - the chips weren't even in production yet.

    On the way to faster chips there are some dead ends - so what?
  • Given Intel's recent history and AMD's, I question the reason you trust Intel MORE. AMD has been consistently offering faster and cheaper products - and has been less plagued by vapor-silicon, or retraction. In general, AMD has shown itself far more capable than Intel - recently.

    I really don't see your point
  • Robert Stack does dammit, haven't you ever seen Beavis and butt-Head Do America

  • AMD, just like Intel (Score:1)

    by Anonymous Coward
    Are the AMD executives all complete idiots? Have none of the engineers spoken up? I simply can't believe that AMD would drop a high level CPU in favour of an SMP consumer level configuration, and be serious about it!

    Seriously, AMD had a great opertunity there to overtake Intel in the server market. It's a well known fact that the Intel Xeons are seriously underpowered; a Xeon at the same speed performs no better than a Pentium 3 but costs twice as much. Intel just have a really good marketing department who have managed to dupe the buying public into thinking otherwise. The AMD Mustang could have blown it away.

    I'm sure that the Athlon DDR chips are great, for the consumer market. But they arn't going to be taken seriously in the server market, especially if AMD insists on using them in an SMP configuration. No enterprise deployment team would take that seriously.

    It's a shame in a way. I really like AMD.

  • Mhz (Score:1)

    by hoss10 ( 108367 )
    Clock rate is like a heart beat.

    If I'm in a race with someone and they're heartbeat is faster than mine it doesn't mean s/he's quicker.

    It's means s/he's a fat b*****d
  • In my opinion the most interesting part of the article was where it states What is known is that AMD will begin pushing into the 64-bit market during the fourth quarter of 2000, when the company readies Clawhammer for introduction in the first quarter of 2001. This interests me much, i had NO idea that they were that soon to announce that they would threaten to release the chip in Q1 2001 (which i read as some time 2001).

    The dissapointing thing about the release is that they did not touch on improved branch prediction &tc in the new cores, or a valid reason to drop mustang. All i see this as is a Plot to make my new mobo obsolete asap. Mmm kind of like other companies too :(.


    How every version of MICROS~1 Windows(TM) comes to exist.

Slashdot Top Deals

"Those who will be able to conquer software will be able to conquer the world." -- Tadahiro Sekimoto, president, NEC Corp.

Close