Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
AMD Hardware Technology

China to Crack Supercomputer Top Ten List 290

jsse writes "ComputerWorld (Hong Kong) has an article about Chinese Academy of Sciences building a supercomputer which has been shown in benchmark tests to process up to 10 trillion floating-point operations per second (TFLOPS) and is expected to take a spot on the list of the world's ten most powerful supercomputers for the first time. The computer is a cluster of 2,560 Opteron 800 series processors from Advanced Micro Devices Inc. (AMD) contained in 640 nodes of four processors each. AMD has announced the project last year when the cluster was building."
This discussion has been archived. No new comments can be posted.

China to Crack Supercomputer Top Ten List

Comments Filter:
  • Hrmm (Score:4, Interesting)

    by acehole ( 174372 ) on Saturday June 05, 2004 @10:47AM (#9344370) Homepage
    So what happened to that dragon cpu chip they were working on?

    • Re:Hrmm (Score:2, Funny)

      by Anonymous Coward
      they're saving that one to be #1
    • Re:Hrmm (Score:3, Interesting)

      by eean ( 177028 )
      From what I remember, they weren't planning those to be really powerful.

    • The dragon CPU was in the Internet Cafes, but alas, all of those got shut down.
  • Use? (Score:5, Interesting)

    by IANAL(BIAILS) ( 726712 ) on Saturday June 05, 2004 @10:50AM (#9344388) Homepage Journal
    I glanced through the article quickly and I see that it doesn't mention what use the computer will be put to. For some reason, I doubt it will be used to model climate change... bets they want to use the cycles to crunch some nuclear weapons simulations?
    • Use? (Score:3, Insightful)

      Well, why should the United States be the sole keeper of Weapons of Mass Destruction? History has shown that we aren't exactly the most trustworthy of these things. Just exactly what do you suppose they are developing at Fort Mead? Vaccines to save the world?
      • Re:Use? (Score:2, Insightful)

        by rayvd ( 155635 )
        Huh? How the heck has history shown we're an untrustworthy keeper of these things? The fact that we used two to avert an extremely costly and bloody land invasion of the Japanese home island?

        I don't seem to recall us using them on our own people or at the drop of a hat even when things turned for the worse in Vietnam.

        I *do* recall the simple existence of them preventing war with the USSR and in the end, being partly responsible for the fall of that country when it couldn't keep up...

        Sheesh...
        • Re:Use? (Score:5, Insightful)

          by jsebrech ( 525647 ) on Saturday June 05, 2004 @12:21PM (#9344878)
          Huh? How the heck has history shown we're an untrustworthy keeper of these things? The fact that we used two to avert an extremely costly and bloody land invasion of the Japanese home island?

          The use of nuclear weapons on hiroshima and nagasaki was tactically unnecessary, it did not decide the war, it only speeded up the ending of it. Especially the second bomb was unnecessary, since the japanese had gotten the message after the first one.

          They could have also dropped the bombs on low-populated areas, but instead they dropped it on civilian cities, knowing full well that the destruction and loss of human life would be massive. And they dropped them without warning, to make sure loss of civilian life would be maximized.

          This massive civilian massacre was a constant factor in WWII-era allied campaigns. Japan and Germany saw constant nighttime firebombing in the later stages of the war, designed to kill as much civilians as possible to destroy enemy morale. Ofcourse, since the allied forces won, the history books were written in such a way as to obfuscate this fact.

          I *do* recall the simple existence of them preventing war with the USSR and in the end, being partly responsible for the fall of that country when it couldn't keep up...

          Exactly. What kept the cold war from becoming hot was the fact that both sides had nuclear weapons. That's the theory of nuclear detente: if everyone has them, no one can use them.
          • Re:Use? (Score:3, Interesting)

            by mlyle ( 148697 )
            The use of nuclear weapons on hiroshima and nagasaki was tactically unnecessary, it did not decide the war, it only speeded up the ending of it. Especially the second bomb was unnecessary, since the japanese had gotten the message after the first one.

            Really? Allied projections for an island-by-island invasion of Japan involved literally millions of casualties of allied personnel and uninvolved civilians. And the Japanese military staged an unsuccessful coup [netcom.com] rather than allow Hirohito to surrender after
            • "Really? Allied projections for an island-by-island invasion of Japan involved literally millions of casualties of allied personnel and uninvolved civilians."

              It always comes down to two extremes, doesn't it? If you don't do what I want, you'll have to take the complete other extreme and look how evil that is...

              Why not drop the a-bomb in the sea. You're saying if we don't nuke a major city, the only other alternative is to hide the a-bomb and fight with sticks?
              • There was no guarantee that the bombs (and they only had a few at the time) would explode properly so demonstrating them would've been a bad idea.

                Why don't I hear the same arguments about FIREBOMBING major cities?
              • Well, given that the Japanese military attempted to prevent a surrender at all costs after both Hiroshima and Nagasaki, and that the surrender tapes almost weren't successfully broadcast due to actions by the Japanese military police forces against the Imperial Guard, I think it's reasonable to question whether dropping a nuke in the sea would have accomplished much. Indeed, there was little reaction at all to the Hiroshima device, nor to the Potsdam declaration before that.

                And there were only 3 atomic we
            • That estimate is complete bullshit. The Joint Chiefs of Staff estimated 38,000 causalities. The 'million' number was from revisionists wanting to justify Truman's decision.

              We demanded 'unconditional' surrender. The Japanese had one condition: Let us keep our Emperor. We said no to that condition. After the Japanese had refused to surrender, Russia joined the Pacific war. We also decided that we were willing to allow the Emperor to remain as a 'figurehead'. If we had told the Japanese these two fact

      • by elvum ( 9344 )
        Hello? Last time I checked, the UK, France, Russia, Pakistan, India and I think China itself all had nukes. Not to mention the Israelis (probably). There's no question of the US being the "sole keeper" of such weapons, even if it wanted to be.

      • Contrary to your "arm the world is fine" statement, let me counter your argument with a little known fact.

        China is not exactly run by the people of China, moreso than many other countries that do have nuclear capability.

        History has shown that when a country is controlled by one person or an elite few, the distance to the sword in any situation is shortened.

        Democracies generally promote peace with the "hell no you're not taking my son!" argument.

        Think about it.
        • My point is, we are hardly the people to be preaching about non-proliferation. Our own bio / nuke programs continue to expand while we tell the world that they should give up these evil tools and embrace "world peace". Shouldn't we be leading by example? To me, it seems like just another example of the US saying that the rules don't apply to us.
    • Re:Use? (Score:5, Insightful)

      by mslinux ( 570958 ) on Saturday June 05, 2004 @11:12AM (#9344500)
      It will be used to calculate, measure and to build mathematical models... like all super-computers. It's just a tool, what exactly they decide to calculate isn't important.

      Math is agnostic, it doesn't care if one is attempting to measure the fallout radius of a hydrogen bomb or what percentage of the earth's surface is water.

      "If you can measure what you speak of and express it by a number, you know something about your subject; but if you cannot, your knowledge is meager and unsatisfactory." --Kelvin
    • Re:Use? (Score:5, Informative)

      by tehanu ( 682528 ) on Saturday June 05, 2004 @11:35AM (#9344610)
      The US does a lot of nuclear weapon simulations as well. So does France and any other country with nuclear weapon capability. I imagine the US is doing even more now as it is developing new types of mini-nukes which the government is seriously talking about deploying in conventional warfare. So if you want to cast stones...

      Anyway, one idea is that the more computer simulations you do, the fewer real tests you have to do. So increased computer simulations may be beneficial for the minorities and rural Han Chinese living in the remote areas that they do the tests in. Of course it would be best to reduce development on nuclear weapons entirely but I don't see that happening in the present climate anytime soon. When even the leader of the free world is out there advocating the development of new nuclear weapons and uses loopholes in treaties to develop them, what exactly do you think the leaders of the paranoid and not-so-free world will do?
    • They are probably using it for web-filtering :)
  • I think maybe we should hold off on asking it "why 42?"
  • by ReeferCpe ( 613569 ) on Saturday June 05, 2004 @10:52AM (#9344392) Homepage
    This is probably the intended use :)
  • by caston ( 711568 ) on Saturday June 05, 2004 @10:53AM (#9344399)
    Would the Chinese need a computer capable of running Longhorn?

    Oh well, More power to them!
    • Herd of Longhorns?

      Sure! I've heard of Longhorns

      No, no, no! A Longhorn herd

      Don't be rediculous, Microsoft is writing Longhorn and Stallman is writing HURD, how could we ever have a Longhorn/Hurd? How would it be licensed? What would it run?

      Perhaps mono really is a Communist plot!

  • [Insert unecessary comparison between this and the G5 cluster at Virginia Tech].
  • i see... (Score:5, Informative)

    by abscondment ( 672321 ) on Saturday June 05, 2004 @11:01AM (#9344435) Homepage

    The computer's being created by Dawning Information Industry Co. (US Site) [dawningusa.com].

    According to their company profile [dawningusa.com],

    Dawning is the unique Chinese high-tech company which can manufacture not only low end PC server worth $1,200 but also high end MPP system worth millions USD.

    They seem to serve a lot of different customers [dawningusa.com], but I have a feeling the government will be making use of this baby.

  • by Anonymous Coward
    will happen from our politicians in the US. My understanding is that the new cluster being built at Oak Ridge is really intended to somewhat compete with the Earth Simulator cluster in Japan. Granted, I don't think it is intended to surpass it but still I just get the feeling that a lot of the supercomputer projects being supported by Congress are political in nature and not really scientific. Now that the Chinese will possibly enter the top 10, that should get the politicos all riled up.
  • w00t! (Score:3, Funny)

    by Anonymous Coward on Saturday June 05, 2004 @11:11AM (#9344494)
    Now the chinese can caluclate how many communists it really takes to screw in a lightbulb....

    (my estimate is the entire chinese population, because if only one person screws it in, then that's just not sharing with others, is it?)
  • by c++ ( 25427 ) on Saturday June 05, 2004 @11:13AM (#9344501)
    Man, I can't stand it when that list goes down from beeing hax0red...
  • by Anonymous Coward
    global warming has significantly increased
  • USA? (Score:3, Interesting)

    by John Seminal ( 698722 ) on Saturday June 05, 2004 @11:31AM (#9344592) Journal
    Does anyone know what the list of top 10 is? I wonder how many of the top 10 supercomputers are in the USA, and how the Virginia Tech G4 cluster ranks.

    Also, a side note. How much extra computing "power" is gained by adding an extra machine to a cluster? For example, I have about 7 or 8 pentiums (most are 166's, there is a 133 and a 200) sitting on the floor collecting dust. If I hooked them all up together, what would the usefulness be? Could I compile programs quicker? Would a cluster make a good web server, jps server? I know my PIII500 can drag with tomcat at times when crunching jsp.

    I am glad china is investing money in technology. With all the people they have living there, they could become a major technolohy hub. Look at what Japan did in the 1980's with manufacturing. Imagine all the cheap(er) products we could have on the market. It all starts with R&D. Even the old "star wars" spending from the 1980's has proven useful in new products in ways not imagined back then.

    • Re:USA? (Score:3, Informative)

      by imsabbel ( 611519 )
      Look at top500.org
      There is a new list comming the next few weeks.

      Your 7-8 pentiums alone could do some useful stuff as a cluster. Not much, but a little. A compile farm COULD be possible (never done it, so i dont know, but if they have enough ram why not, as long as you have a better machine to do the linking/ect)

      But adding them to an existing cluster of faster machines would slow the whole stuff down (more communication cost lowering the total efficiency more than outweights the computing power of the ne
    • Re:USA? (Score:5, Informative)

      by chamblah ( 774997 ) * on Saturday June 05, 2004 @11:44AM (#9344654)
      Does anyone know what the list of top 10 is?

      Here [top500.org] is the current list of 500, last updated in November of '03.

    • Re:USA? (Score:3, Informative)

      by foidulus ( 743482 ) *
      Several errors in your post I would like to point out:
      I wonder how many of the top 10 supercomputers are in the USA, and how the Virginia Tech G4 cluster ranks.
      It's a G5 cluster first of all, and it ranks third(IIRC the biology department wants to use it)
      Also, a side note. How much extra computing "power" is gained by adding an extra machine to a cluster? For example, I have about 7 or 8 pentiums (most are 166's, there is a 133 and a 200) sitting on the floor collecting dust. If I hooked them all up tog
    • Top 10 Summary (Score:4, Informative)

      by thakadu ( 776967 ) on Saturday June 05, 2004 @11:52AM (#9344702)
      The countries from top 10 are: 1. Japan 2. USA 3. USA 4. USA 5. USA 6. USA 7. USA 8. USA 9. USA 10. USA and the top 10 countries in the list with (highest rank) are: 1. Japan (1) 2. USA (2) 3. China (14) 4. France (15) 5. UK (16) 6. South Korea (22) 7. Canada (29) 8. Germany (31) 9. Netherlands (41) 10. New Zealand (44)
      • Top 10 *Known* (Score:2, Interesting)

        by brunes69 ( 86786 )
        It would not suprise me in the lest to find out that the NSA always maintains a supercomputer al least 2x as fast as the fastest "publically" listed supercomputer.

    • Re:USA? (Score:3, Informative)

      by Dun Malg ( 230075 )
      Also, a side note. How much extra computing "power" is gained by adding an extra machine to a cluster?

      Depends on how big the cluster already is. Are you adding one machine to a cluster of 6, or a cluster of 600?

      For example, I have about 7 or 8 pentiums (most are 166's, there is a 133 and a 200) sitting on the floor collecting dust. If I hooked them all up together, what would the usefulness be?

      Practically none. Buy a single 2.4GHZ Celeron box off eBay for $200 and you'll get nore out of it.

      Could I c

      • No. Clusters are only good for doing tasks that "parallelize". Compiling is too linear.

        Compiling parallelizes quite well. See distcc and XCode/Rendezvous.

        • No. Clusters are only good for doing tasks that "parallelize". Compiling is too linear.

          Compiling parallelizes quite well. See distcc and XCode/Rendezvous.

          I stand corrected. I was thinking along the lines of parallelizing the compilation of a single module. I hadn't really considered compiling multiple parts being compiled and then linked-- duh. Still, he'd probably be better off compiling on a single 500mhz machine than splitting it up across several old 133 or 166mhz boxes, considering the likely diffe

      • Re:USA? (Score:3, Informative)

        by sql*kitten ( 1359 ) *
        Clusters are only good for doing tasks that "parallelize". Compiling is too linear.

        Linking is linear; you could compile C in as many parallel tasks as you have source files. Java compiles can be parellelized quite a bit too, particularly if your code makes heavy use of Class.forName(). I do this a lot, tho' quite gratuitously; I've got 4 CPUs so I use 'em, but with the size of projects I work on and the speed of even a single processor and a modern javac, it doesn't really make a vast amount of differen
      • "No. Clusters are only good for doing tasks that "parallelize". Compiling is too linear. Besides, it's not like you hook up 8 machines to a 10/100 switch, start up a few "cluster daemons", and end up with a single virtual machine."

        That, would be AWESOME, though.

        "No. Clusters are only good for doing tasks that "parallelize". Compiling is too linear."

        Not that you are wrong, but Apple has a technology built into their development suite(which comes free with os X), that lets you use their rondevous networkin
      • No. Clusters are only good for doing tasks that "parallelize". Compiling is too linear.

        Compiling a single file is. But most software has more than one file, and you can compile different files on different machines. Check out distcc [samba.org].

        Of course, even using distcc, a single low-end Athlon will be 10x faster than all of those boxes combined, and use up a lot less electricity as well. So it's not exactly a useful excercise. Interesting, perhaps.
    • There are a lot of types of clusters. You could put those 8 boxes together in a shared something cluster and make a set of webservers that not faster than any single box but is a lot more reliable. The TCO is all wrong for it you would be better off getting a couple modern PC's and making an active active cluster out of them as it will be cheaper to power and have a lot more ram. A simple shared nothing cluster can make compiling something big like a kernel faster than a single machine but again a single
    • > Could I compile programs quicker? Would a cluster make a good web server, jps server? I know my PIII500 can drag with tomcat at times when crunching jsp.

      You could compile the linux kernel in two seconds if you used distcc, but your game of quake would have average frame rates, and wouldn't take much advantage of the parallelism at all. Any individual program still has latency bottlenecks that don't parallelize. And since it's busy crunching numbers at top speed, it doesn't have great interactive res
    • The publicly posted LIST [top500.org] would be a good place to start. The Virginia Tech computer ranks third.
  • by autopr0n ( 534291 ) on Saturday June 05, 2004 @11:42AM (#9344643) Homepage Journal
    I take it china's no longer on the list of countries we're not allowed to sell supercomputers too.
  • For those interested this article [businessweek.com] from BusinessWeek has an update on DARPA's goal of funding the building of a petaflop (quadillions of calcuations a second) computer. This and other BusinessWeek articles indicate that while Asia may have the lead on the US now, the US will likely take back the leadership position by the end of the decade.

I'd rather just believe that it's done by little elves running around.

Working...