IDF: demos I2 and P4 dual-core!

Intel managed to demo <A HREF="http://www.anandtech.com/tradeshows/showdoc.aspx?i=2197&p=3" target="_new">Montecito</A>. It seems that bayshore, Montecito's supporting chipset, is to deliver >3x more bandwidth, which might be an indication of a move to a point-to-point bus... Let's wait and see. Montecito seems to be well on schedule, which is yet another indication that there's nothing wrong with Intel's 90nm tech, but rather with Prescott/Netburst.

Oh, and the inquirer reports that <A HREF="http://www.theinquirer.net/?article=18329" target="_new">Intel demos Dual-core Pentium-Ms</A>...

<b>Update: Intel also showed a dual-core netburst-based solution. <A HREF="http://zdnet.com.com/2100-1103_2-5356703.html" target="_new">Here's the link</A></b>!
<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 09/11/04 07:31 PM.</EM></FONT></P>
19 answers Last reply
More about demos dual core
  1. <A HREF="http://zdnet.com.com/2100-1103_2-5356703.html" target="_new">here</A> it says intel demos dual core.

    <i> :evil: <font color=blue>Futile is resistance,</font color=blue><font color=red> assimilate you we will.</font color=red> :evil: </i>
    <b>Hard work has a future payoff. Laziness pays off now.</b>
  2. Thanks for the info. The original post has been updated.

    Also, <A HREF="http://www.xbitlabs.com/news/mobile/display/20040908141343.html" target="_new">Xbitlabs</A> posted more news on an actual demonstration of Yonah.

    Will Intel actually manage to show off the complete picture? Yonah, Smithfield <i>and</i> Itanium? That would be great!!! Or is this just confusion on some dual-core demo of theirs?
  3. This should be somewhat a tribute to montecito's engineering:

    AMD: <b><font color=green>8 logical processors</font color=green> </b><A HREF="http://www.theinquirer.net/images/articles/amd_b.jpg" target="_new">(link)</A>
    Intel: <b><font color=blue>16 logical processors</font color=blue> </b><A HREF="http://images.anandtech.com/reviews/shows/2004/idf/fall/dscn4140.jpg" target="_new">(link)</A>

    People like to show off task managers with lotsa CPUs? Intel won the "wow" rating right now, at least.

    Montecito's clock will dictate a lot. If indeed Intel wants to pursue this other front against AMD, all they might need to do is manage >2.0Ghz speeds... In theory. They'd still need software support anyway, though, but with such a major Itanium overhaul, even 32-bit emulation tasks will probably get a big boost.

    Even so, Opteron has a lot more momentum than Itanium. In order to truly make a difference whatsoever, Montecito would have to be completely revolutionary... Which will in all likelyhood not be the case. I wonder how good Montecito's multithreading is?...

    Also, Madison 9M will probably be disappointing. It'll only be 1.6Ghz at most, and will probably only use current infrastructure. Montecito, on the other hand, will be accompanied by a major platform overhaul in 2005: PCIe and DDR2, which by that point will probably be worth something (unlike now!).
  4. Hmm, the last real news (not planning is going well) on the Itanic2, was that it was delayed, and would debute slower than planned.
    Intel says everything is fine, right to the failure, lately.
  5. Montecito's future is going to be in highly specialized applications. Unless, 32bit compatibility gets really impressive, and they drop the price of the things.

    The Opteron can't touch I2's floating point, that's really I2s strong point. That's why I see its future limited to specialized applications. Your typical server doesn't need blazing floating point, but your typical supercomputing cluster does. Render farms, massive physics simulations, etc... But you know, if IA64 delivered the goods in terms of price/performance, I'd be plenty happy to adopt it. Linux is linux is linux. :)

    I'm more interested in seeing the 2nd gen dual core Xeons. From what I've heard, 1st gen are going to be somewhat of a "me too" hack. That's all hearsay at this point, but I tend to trust my source.

    <i>Dual Opteron 248s, 5900nu, 4gig ram, dual 36gb raptor, 80gb hd, 550w Enermax, Suse64 9.1, and a bunch of other crap.</i>
  6. I think Intel has to drop itanium pricing now.

    Actually, they already did something unprecedented: they lowered the current madison lineup prices, without doing what they traditionally do: only shifting more technologically advanced products into the old price niches. This was baaad for Itanium acceptance... There was no way to get a reasonably-priced Itanium system...

    Therefore, I suspect Intel is quite serious about changing Itanium pricing policy... which is, of course, something that is looong overdue!

    A more than competitive Montecito should be high on Intel's priority list to justify Itanium's multi-billion research and development investments!

    <b>PS: Where the hell are the day 2 coverages, xbitlabs and anandtech?</b>
  7. Quote:
    I'm more interested in seeing the 2nd gen dual core Xeons. From what I've heard, 1st gen are going to be somewhat of a "me too" hack. That's all hearsay at this point, but I tend to trust my source.

    I was under the impression that Whitefield was the one which was much more than a me too hack. It will have platform compatibility with Itanium and four cores and is being designed from scratch to replace netburst... The next dual cores from Intel might be me too hacks indeed, though.

    Except maybe Montecito - that one seems to be doing more or less OK and well now, so I'd expect montecito to be a more polished processor than say, smithfield!
  8. Actually, the last news about Itanium that I heard was
    <A HREF="http://www.linuxworld.com/story/45989.htm" target="_new">this</A>

    Summary:

    "NASA is once again partnering with SGI and Intel, enlisting SGI's Linux-based Altix machines, along with the computing power 10,240 Itanium 2 chips can provide, in creating a supercomputer."

    That doesn't sound like a failure...

    I'm just your average habitual smiler =D
  9. Exactly, Montecito is probably going to be Itanium's breaking point - if Intel has its way...

    Oh, and there are some documents on the web about theoretical throughput of montecito-based systems that indicate that <b>Montecito's</b> top bin <b>core clock</b> will be of <b>2.5Ghz</b>. Even in 32-bit emulation, this beast of a processor should give xeon a run for its money! Itanium has to show performance now, because it doesn't have the "Hey, I'm 64-bit" advantage anymore at all. If they do manage to get a 2.5Ghz Montecito with Bayshore out the door by early 2H05, they might just beat Toledo, AMD's dual-core Opteron, for bragging rights in terms of raw performance. It depends a lot on toledo as well - and at which speed they can clock it without giving off too much heat.
  10. <A HREF="http://www.theinquirer.net/?article=18345" target="_new">The Inquirer's Guide</A> to montecito...
  11. ...however, unfortunately, <A HREF="http://www.theinquirer.net/?article=18348" target="_new">Montecito needs</A> a 150-200W cooling solution. TEC-based will probably what will happen - that's thermoelectric cooling based on the peltier effect. It's an active cooling, hardcore, but its production costs are probably high.

    In any case, if Montecito delivers like it's supposed to, the server guys probably won't mind having to pay for the cooling solutions...

    And TEC is highly interesting, anyway! Who knows, maybe we'll all end up with TEC-cooling...
  12. <i>Therefore, I suspect Intel is quite serious about changing Itanium pricing policy... which is, of course, something that is looong overdue!</i>

    You can say that again. When I first went dual Opteron, the cheapest single Ia64 cpu/mobo combo I could get cost more than the dual amd64 cpu/mobo I ended up getting. There were many other reasons I went amd64 as well back then, many of them still valid.

    <i>Except maybe Montecito - that one seems to be doing more or less OK and well now, so I'd expect montecito to be a more polished processor than say, smithfield! </i>

    I'm not calling montecito a hack at all. That's been in design for many years, and wasn't reactionary at all. I expect this to be a first rate first gen implementation that will smoke every other chip in existance. It's the high end x86 line that has them in reactionary mode, IMO. The dual core Xeons specifically I don't expect much from at first.

    <i>Dual Opteron 248s, 5900nu, 4gig ram, dual 36gb raptor, 80gb hd, 550w Enermax, Suse64 9.1, and a bunch of other crap.</i>
  13. Quote:
    If they do manage to get a 2.5Ghz Montecito with Bayshore out the door by early 2H05,

    Perhaps you were unaware that the shipping date of the new itanic has been moved to next year, and the targeted launch speed has been downgraded from 1.7 to 1.6.
    You seem the eternal Intel optomist, but maybe it's just that I am the eternal pesamist.
  14. I thought Madison 9M has been delayed to Q4 2004 from Q3 2004... Wasn't that the case? Here you have the whole <A HREF="http://www.theinquirer.net/?article=18226" target="_new">Theinq</A> article:
    Quote:
    TAIWANESE WIRE Digitimes claims that the next version of the Itanium, the Madison 9M, has been pushed back a quarter.

    It speculates that the delay may have been caused by its lightning shift to dual cores, and could have an effect on the dual core Montecito chip.

    It also claimed that the frequency of the Madison 9M is now likely to be 1.6GHz rather than 1.7GHz.

    Its launch date is now year end, the wire reported. µ

    It seems you <i>are</i> the eternal pessimist. :smile:

    I wouldn't expect much from Madison 9M, though, even with all my optmism. :wink: The real big gun is Montecito.
  15. Ah! I managed to find that <A HREF="http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=2588&Thread=1&entryID=36497&roomID=11" target="_new">link</A> from Paul DeMone (over from chip-architect) according to which Montecito might clock as high as 2.5Ghz. Here is the whole thing:
    Quote:
    There's an interesting slide in this presentation:
    <A HREF="http://www.t-systems-sfr.com/download/2rzk_juling.pdf" target="_new">http://www.t-systems-sfr.com/download/2rzk_juling.pdf</A>

    On slide 5 a number of system configurations are described. Phase 1 is a cluster of 116 dual CPU 1.7 GHz Madison 9M nodes. Stated performance is 1.58 TFLOPs/s. That's obviously a peak rating:

    116 nodes * 2 CPUs/node * 4 GFLOPs/cycle * 1.7 GHz = 1577.6 GFLOP/s

    No rocket science here.

    Now observe Phase 2. It is a 218 node system with 2 sockets/4 CPUs per node using "top bin" Montecito. Performance is stated as 8.72 TFLOP/s

    Freq = 8720 GFLOPs/s/(218 nodes * 4 CPUs/node * 4 GFLOP/cycle) = 2.500 GHz.

    A nice even number. Hmmm 20 GFLOP/s per socket, not bad at all!
  16. well i hope they can work on the heat issues. i dont care how powerful the thing is, cooling solutions will be a factor in corporate buying, you should realize that.

    look at it this way, lets say the itanium is twice or triple the performance of closest competitors, but costs twice to 3 times as much in infrustructure to maintain, while you could buy a larger cluster of the competitors, still be under that limit and be able to outperform the smaller itanium cluster.

    i hope 200w covers up to 2.5ghz...

    you know, i once had a more optimistic hope for itanium, but now after years of it sitting out there doing nothing, im not so eager to jump up and down and applaude intel. nothing has happend yet, they have laid out some blue prints to drool over and shown off some samples, but there are no tests and no numbers, and after recent delays, the rlease is not set in sotne as far as im concerned. im in a wait and see mode, i want intel to suprise me, ive been waiting for them to pull something out.

    amd is chugging away, its looking lieka horse race at this point, anyone's game to win or loose if both can keep up innovation wise. If intel can turn itanium around, that may be the first big step of coming even with amd all over again. but i dont think its going to flip flop like it has int he past, intle and amd trading leads. it jsut seems like things will begin to evne out at both camps, each offering new tech to stay fresh, but neither smashing the other. on hte market level side though, that owuld probably favor amd gaining share. the longer they stay even, the more comapnies look at amd as an respectible option. although itanium really does nothing to amd, it does alot fo intel. if they can get over that hump of profitability, that will trickle down to desktop and mobile, and thats what will take on amd.
  17. Apparently, from what I've read, montecito indeed sticks to the 100W thermal envelope. The big problem with what the inquirer reported is that they use TEC-coolers, which are pathetically inefficient. So while your typical fan consumes maybe 4-6W, each of the two TEC units on that heatsink is probably churning 40W of electricity, which all goes to heat. So while the CPU is responsible for 100W, the other say 40-80W may be due to the use of the horrendously inefficient peltier effect.

    Maybe that's why we don't have air conditioners and refrigerators based on peltier cooling...
  18. I agree with you. When I read that about the heat I just shook my head. I2 is destined for specialized high end markets.

    I mean seriously, you have to port all your apps, or run it in slower emulation mode. If emulation mode is made as fast as a Xeon(big if), you're still not realizing the full potential of the processor or speed gains over much cheaper hardware. Then, not all hardware is supported on the platform. Then you're looking at high operation costs of not just powering the systems, but of cooling them too. And on top of all this, unless Intel changes its past pricing schemes, you're going to pay through the arse for those systems.

    The typical corporate data center isn't going for that. Itanium's future is in the high end and specialized application space for the foreseeable future.

    <i>Dual Opteron 248s, 5900nu, 4gig ram, dual 36gb raptor, 80gb hd, 550w Enermax, Suse64 9.1, and a bunch of other crap.</i>
  19. Quote:
    for the foreseeable future.

    Hm, while you do have a point, I don't really think at this point that we can foresee what will happen by the end of 2005.

    I mean, Montecito is still a long way off, and is already topic for dicussion in several forums because Itanium's future depends a lot on it. It will only be available a year from now, and people are already talking about it...

    And the market reaction to dual core is also hard to guess........ I think it might be just safer to wait and see......
Ask a new question

Read More

CPUs Intel Dual Core