back to article Australia teases binning x86 for Power CPUs in new supercomputer

Australia's Commonwealth Scientific and Industrial Research Organisation (CSIRO) has issued a (regwalled) tender for a new petaflop-class supercomputer that can run either x86 or Power systems silicon. But those keen to declare Intel's high-end reign is coming to an end should not get too excited: the tender may mention Power …

  1. John Savard

    Doubly Amazed

    I would have thought that since a scientific supercomputer is used for running Fortran code, not for running popular off-the-shelf applications, there would be no reason to insist on a system that could run Microsoft Windows.

    But, since Intel has the advantage of a huge market for its processors to ordinary computer users, who do have a need for Microsoft Windows, in terms of performance per dollar, I would have expected that by now Intel would have such an advantage that IBM would find it hard to be competitive even in HPC without obstacles like this.

    1. Anonymous Coward
      Anonymous Coward

      Re: Doubly Amazed

      The hints are in the previous build, referenced in the article.

      large shared NFS/Windows file systems

      The Accelerator cluster supports both Linux and Windows HPC environments.

      Use Microsoft Remote Desktop utility and connect to:

  2. Andy The Hat Silver badge

    Have I missed something?

    On the one hand it's got to be hardware compatible with Win Server 2012, on the other hand it's got to run SUSE Enterprise Server ... is this an upside-down Aussie spec for a machine actually running SUSE but *could* run Win Server if they wanted to (as a nod to the competition gods)?

    1. Anonymous Coward
      Anonymous Coward

      Re: Have I missed something?

      "as a nod to the competition gods"

      Probably more a nod to the performance gods. Windows Server tends to outperform Linux in many HPC scenarios due to native OS support for features like SMB Direct, Network Direct, kRDMA, NVDIMMs and hardware offloading...

      1. Paul Crawford Silver badge

        "Windows Server tends to outperform Linux in many HPC scenarios"

        Really? Facts please, like how many Windows machines are in the top 500 supercomputers?

        1. Paul Crawford Silver badge
          Trollface

          Re: "Windows Server tends to outperform Linux in many HPC scenarios"

          Ah, 8 hours passed with 2 down-votes so far, and yet not a contradictory fact in sight :)

      2. Anonymous Coward
        Anonymous Coward

        Re: Have I missed something?

        Most are of the camp that understands Microsoft is a "secret organisation" with a single purpose, eg to force consumers of all types to buy ever more powerful hardware to do the same or less. That sustains the industry. There is no better way to make a petaflop system perform like a megaflop system than running windows.

        A few years ago my Microsoft HPC rep described windows HPC as a great way to run parallel excel (and not much else) if that makes the point.

        CSIRO has come a long way from their early days of scientific computing but the budgets have not, and their more able scientists are increasingly retired because they have higher salaries than newer graduates. Enter the era of the LCD scientist.

  3. Mage Silver badge

    Old fogey mode

    I remember when Windows (NT) did support Power, MIPS, Alpha 32, Alpha 64 and x86.

    I think Windows went downhill due to migrating win9x /WinME features and programmers to the dev team and concentrating on Shiny (Ribbon in Office, Aero etc).

    They ought not to be specifying Windows at all.

    1. Chris King

      Re: Old fogey mode

      Alpha was never 32-bit, and NT ran like chilled treacle on them.

      Nothing like buying a new AlphaServer while MS and Compaq were BFF's, then having to explain to explain to management three weeks later that Win2K on Alpha had been cancelled :-(

      1. Paul Crawford Silver badge
        Trollface

        Re: Old fogey mode

        Alpha was 64-bit always, but NT ran very nicely indeed on them. Until it was cancelled and we had to explain to a major European space organisation just how trustworthy MS were in terms of portability and cross-platform support. So much for the promises they made when we went down that road.

        Oh well, not that MS matters that much to us now. Hey, MS uses, just how are those price-hikes going down? [hence troll icon]

  4. Tom Womack

    If they're trying to fit a petaflop in five racks for $3 million, the processors will just be sitting there driving the GPUs. My guess is that they're anticipating nVidia, who has quite tight links with POWER, producing a POWER-and-GPU box which might be more cost-effective than the current Xeon-and-GPU box.

    ($3 million is interesting close to what you get if you divide a petaflop by the current $129k 42.4TF nVidia DGX100 and expect a small discount for buying in bulk)

    1. TaxiDriver

      Better price/performance but

      I'd agree with that, OpenPower-nVidia boxes would be the better price performance fit. But then Team IBM didn't win the NCI tender a week back did they, will a small IBM partner out of Melbourne with no HPC tech experience do better? Maybe.

  5. Big-nosed Pengie
    Headmaster

    "Compute"

    It's a fecking verb.

  6. sms123

    Coming soon (courtesy google translate):

    Das Register

    Beißen die Hand, die IT speist

    There's no other reason for having "ist" in the article is there (I mean apart from a typo)?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like