Forgot your password?
typodupeerror
IBM Supercomputing Technology

IBM To Build 3-Petaflop Supercomputer 73

Posted by Soulskill
from the onward-and-upward dept.
angry tapir writes "The global race for supercomputing power continues unabated: Germany's Bavarian Academy of Science has announced that it has contracted IBM to build a supercomputer that, when completed in 2012, will be able to execute up to 3 petaflops, potentially making it the world's most powerful supercomputer. To be called SuperMUC, the computer, which will be run by the Academy's Leibniz Supercomputing Centre in Garching, Germany, will be available for European researchers to use to probe the frontiers of medicine, astrophysics and other scientific disciplines."
This discussion has been archived. No new comments can be posted.

IBM To Build 3-Petaflop Supercomputer

Comments Filter:
  • by Anonymous Coward

    Key take aways from the summary:

    1. IBM will be responsible for a large 'SuperMuck'.
    2. It will be used for probing by the Europeans.

  • This looks like a pretty awesome setup they have. I'm glad that the US has a few supercomputer projects planned for 2012 that will possibly bring the somewhat elusive #1 title back our way. We'll have to see, the competition as always is pushing the envelope and by that time who knows what else could be in the works from China, etc.

    Anyways, pre - gratz to the Germans for their new machine. Is anybody familiar with the hot water cooling tech developed by IBM as mentioned in the article?

    • They recently published this [discovery.com]. It's about cooling with relatively hot water (60 degrees Centigrade/140F). They probably researched this because the water is cheaper to cool by air when it's hotter.
      Also in 2008 they published this [ibm.com], a solution to cool inside stacked dies.
  • by pip-PPC (46392) on Tuesday December 14, 2010 @03:32AM (#34543776)

    From the article:

    "The system will use 14,000 Intel Xeon processors running in IBM System x iDataPlex servers."

    IBM has two in-house HPC platforms that could both reach 3 PFLOPS (BlueGene/Q and POWER7), but instead they're building a Xeon cluster. I'm surprised that they would want to put a machine near the top of the TOP500 that wasn't a full-on IBM benefit--maybe IBM Germany is the contractor, and they don't have the R&D expertise? Or the Xeon cluster is cheaper/easier to program and maintain?

    • by Anonymous Coward

      Didn't IBM help the Germans build such a machine back in the 1930's. As I remember it was to be used for counting people and general census taking....q

    • by Required Snark (1702878) on Tuesday December 14, 2010 @04:06AM (#34543880)
      Here is a look at the guts of the IBM next generation BlueGene/Q. http://www.theregister.co.uk/2010/11/22/ibm_blue_gene_q_super/page2.html [theregister.co.uk]

      The Sequoia super that Lawrence Livermore will be getting in 2012 — IBM said it'd be in late 2011 back when the deal was announced in February 2009, so there's been some apparent slippage — will consist of 96 racks and will be rated at 20.13 petaflops. Argonne National Laboratory said back in August that it wanted a BlueGene/Q box, too, and it will have 48 racks of compute drawers for a total of 10 petaflops of floating-point power.

      Both the Chinese machine and the German machine are not cutting edge designs. They represent what you can do with near commodity hardware and good but not fully custom packaging. They may look like top end machines today, but by 2012 they will not be in the top ten.

      • by mwvdlee (775178)

        Hah! 20 petaflops may look like top end machines in 2012, but by 2015 they will not be in the top ten.

      • by timeOday (582209)

        Both the Chinese machine and the German machine are not cutting edge designs. They represent what you can do with near commodity hardware and good but not fully custom packaging.

        The cooling system sounds genuinely innovative and beneficial, if successful:

        It will also use a new form of cooling that IBM developed, called Aquasar, that uses hot water to cool the processors, a design that should cut cooling electricity usage by 40 percent, the company claims.

        "SuperMUC will provide previously unattainable en

    • Don't be too proud of this technological monster that's about to be constructed. 3 petaflops is insignificant next to the size of the kickback somebody will receive in the process of purchasing 14,000 CPUs.
      • by Anonymous Coward
        Don't try to frighten us with your sorcerer's ways, Lord Dimethylxanthine. Your sad devotion to that ancient religion has not helped you conjure up the stolen wikileaks cables, or given you clairvoyance enough to find the Rebels' hidden data cent..gghggghgg
    • by Anonymous Coward
      A big benefit of Xeons is that you can supplement them with off the shelf GPUs. Currently there aren't really a lot of high performance GPUs available for the POWER platform right now. It would be possible to port them, but not sure if it would be worthwhile for either IBM or NVidia to do so.
      • by Junta (36770)

        However, this config makes no mention of GPUs, so it's probably moot. If you are saying they may upgrade these later, I would be surprised if they are using systems with enough space to accommodate GPUs if not doing it up front. Most of these configurations, regardless of vendor, go to half the CPU density to make room (space, power, cooling wise) for gpus. When dealing with a scale of 14k cpus, you generally pick the config up front and don't bother going back for piece-wise upgrades.

        The only way to tel

    • by Junta (36770)

      Many don't care about the architecture because all their work is not hard to redo per-architecture. Those will jump on Itanium, POWER, Sparc, or whatever architecture the vendor has that hits the sweet spot. You can tell those as their top500 entries from year to year frequently jump architectures. These are also customers most amenable to jumping on the GPU bandwagon, despite the fact they are more painful to program for and require particular care and feeding to avoid becoming memory bottlenecked.

      Many

  • SuperMUC is not cool on any level. Kind of makes my spine tingle with grossness actually.
  • by alvinrod (889928) on Tuesday December 14, 2010 @03:37AM (#34543796)
    I absolutely love Moore's law. Think that this is an insanely awesome amount of computational power? Just wait around for 10-15 years and we'll likely have that same order of magnitude in our personal computers. Just look back at the supercomputer list from a decade ago [top500.org] and notice that right now we have hardware capable of getting similar performance. The best Intel processors can put out over 100 GFLOPS. Graphics cards are closer to 1TFLOPS.

    Another way of looking at it is that we'll have a similar amount of power in our phones, tablets, etc. that we have in our desktops right now. Super computers are going to get even more super and the types of problems that are expensive to solve today continue to get cheaper. I'm still a young man, but given how far things have come since I was born, I can't help but wonder what the world will be like when I'm many years further along the road. If for no other reason than the vast amount of computational power that's available to us.
    • by NewtonsLaw (409638) on Tuesday December 14, 2010 @03:48AM (#34543824)

      So our desktop computers will wait thousands of times faster than they do today... for the next keystroke or mouse button-press. :-)

    • Moore's law is ok.

      I prefer that law, forget the name right now, that says that as computational power increases, windows will require ALL of it to run, greatly increasing demand for CPU and RAM, and lowering the cost of hardware just behind the curve for the rest of us.

      • by FeepingCreature (1132265) on Tuesday December 14, 2010 @03:59AM (#34543862)

        Moore's law is ok. I prefer that law, forget the name right now, that says that as computational power increases, windows will require ALL of it to run, greatly increasing demand for CPU and RAM, and lowering the cost of hardware just behind the curve for the rest of us.

        As Intel giveth, Microsoft taketh away.

        • by Memnos (937795)
          I have a cat. I routinely exceed ten petaflops daily, but it screws up her neural net processing big time.
      • Re: (Score:3, Informative)

        by Esvandiary (1302095)
        I believe May's Law is the one you're referring to; a corollary to Moore's Law, stating that software efficiency halves every 18 months (or two years).
        • I don't know if it's the efficiency that falls, or just that all the extra power gets used on 3D-shaded semi-transparent smooth-scrolling menus.

          Why yes Vista, I did glance at you!

    • by alienzed (732782)
      Computational power is definitely growing, but so are the requirements of the software that runs on them. Microsoft Word isn't any faster today than it was ten years ago, yet our computers are many times more powerful and more capable. Isn't it amazing to think that ten years ago, it was the year 2000. Time really does fly.
    • Overall, it's not necessarily the amount of computational power available that makes things happen; the extra power simply helps things happen faster. Having Petaflop systems available certainly helps out with the insanely complex tasks (like those in the medical and physics fields); the trickle-down of those developments also make for the nice shiny toys that we call phones. End of the day, though, to the casual user it means nearly nothing, except for when it affects the price tag and/or the performance
    • Back in 1983, I had an Atari 800 and a Kaypro. They prolly had more power than the computers used to land on the moon (and I remember that too). In 1969, my Dad was doing his PhD in fluid dynamics on an IBM with 64k of core memory. My calculator blows the old mainframe away (though the mainframe did useful work for 25 yrs).

    • by Memnos (937795)
      True, but right now there are many orders of magnitude more processing of information occurring in just a few hundred of the cells in most parts of your own body.
    • what's sad is the ability of bloat in the OS and applications to reduce our evermore powerful machines to the same sluggishness.
    • Moore's law is over with - transistor density isn't jumping like it used to and we're running into thermal limits with current designs. That's why you see 4 core and more on the desktop.
  • by Entropius (188861) on Tuesday December 14, 2010 @04:09AM (#34543888)

    Once upon a time, supercomputers were bunches of general-purpose cpu's, and you made them faster by connecting up more of them.

    Now people have realized that massively parallel special purpose chips (like Cell and, even more so, GPU's) can be used to do general-purpose computing, and have started to add those to clusters. But those chips have a lower bandwidth:flops ratio than the x86 etc. CPU's that have been historically used; the gap between a computer's "peak" FLOPS (on an ideal job with no communication requirements to either other nodes or to memory) and the performance it actually achieves is wider using something like CUDA than on a standard supercomputer. CUDA machines are so bandwidth-limited that people use rather hairbrained data compression schemes to move data from place to place, just because all the nodes have extra compute power lying around anyway, and the bottleneck is in communication. (The example that comes to mind is sending the coefficients of the eight generators of an SU(3) matrix rather than just sending the eighteen floats that make up the damn matrix. It's a lot of work to reassemble, relatively speaking, but it's worth it to avoid sending a few bits down the wire.)

    CUDA is wonderful, and my field at least (lattice QCD) is falling over itself trying to port stuff to it. Even though it falls far short of its theoretical FLOPS, it's still a hell of a lot faster than a supercomputer made of Opterons. But we shouldn't fool ourselves into thinking that you can accurately measure computer speed now by looking at peak FLOPS. It makes the CUDA/Cell machines look better than they really are.

    • by afidel (530433) on Tuesday December 14, 2010 @04:54AM (#34544022)
      No, the computers that the term supercomputer was coined for were all special purpose vector machines that couldn't even run an OS, they had to be fronted by a management processor. Only much later were clusters of commodity machines (often with specialized interconnects for high bandwidth and low latency) accepted as contenders for the name. Now with Cell and GPU's we are getting back to fast vector machine with a management computer in the front but now the front end computer is capable of computations (at least in the case of the GPGPU machines) and each machine is a few rack units instead of a couple racks.

      Oh, and the measure you are looking for are Rmax to Rpeak which will tell you how efficient the machine is (at least for LINPACK which may or may not track with your own code depending on how chatty it is in comparison to the benchmark).
      • by Memnos (937795)
        Perhaps, just maybe, we could adopt a very fuzzy metric instead of precision without accuracy. Say for example, what important questions can they help us answer? As amorphous as that criterion will be, it will likely stimulate smart engineers to do whatever necessary to get tangible, valuable, benefits.
    • by Anubis350 (772791)
      I was at the top500 talk at SC10 a few weeks ago, and that's their biggest issue right now. Hell, even core counts are contentious (how many cores does a Tesla C2050 have? Nvidia would say 448, the Top500 guys would say 56 ). The HPL benchmark does work, with some porting, on GPUs (I know firsthand, I used Nvidia's CUDA build of it for benchmarking for the SC10 cluster challenge) but there's a lot of devision I've seen in both vendors and researchers as to whether it's the best benchmark going forward.....
    • Re: (Score:3, Interesting)

      I had to really think about measuring the efficiency of a simulation and I came up with a single answer: money. I was at a lecture about gyrokinetic simulations, and when I heard about the amount of resources being used for some simulations, I asked "how much does one of these simulations cost, in euros?". Luckily for me, the guy knew (large simulations cost in order of thousands), and he also knew how much an experiment on ITER will cost (order of a million); his argument was "it's obviously efficient to r

  • By 2012 IBM will have built at least two Blue Gene/Q systems capable of 20 and 10 Pflops each. The "Sequoia" at Lawrence Livermore National Laboratory and "Mira" at Argonne National Laboratory. There should be plenty of petascale supercomputers in a variety of configurations and architectures by 2012.
  • To be called SuperMUC, the computer, which will be run by the Academy's Leibniz Supercomputing Centre in Garching, Germany,

    No doubt named after the delicious Leibniz [wikipedia.org] cookies, mmmm, mmmm.

    • No it's after the guy who got in an epic row with Newton over the invention of calculus that led to English mathematicians rallying around Newton's method in an upswell of patriotic fervor.

      It took English mathematics a century to recover. So the next time you hear someone criticize something because "that's the way they do it in France" remember: Good artists copy, great artists steal (a bon mot I just came up with, pretty brilliant if I may say so)

  • Hopefully WE can can get the Germans to get the answer to this question.
  • This wouldn't even come close to being the most powerful supercomputer, what with Blue Waters coming out in 2011..
  • It's gonna run super-a-muc
  • this november's list already reached 2.5Pflops. A machine delivered at 3Pflops in 2 years from now will not even be in the top 10. Long term trend is to reach 5 to 10 Pflops by mid 2012. http://www.top500.org/list/2010/11/100 [top500.org]

Their idea of an offer you can't refuse is an offer... and you'd better not refuse.

Working...