back to article Half-petaflop IBM BlueGene supercomputer plan announced

IBM will provide extra processing power to the US Department of Energy (DOE)'s Argonne National Lab, quintupling Argonne's power to 556 trillion floating point operations per second (teraflops)by using Blue Gene/P supercomputer tech. “[Argonne Lab] has been a valuable contributor in the development of Blue Gene/P,” said Leo …

COMMENTS

This topic is closed for new posts.
  1. Anonymous Coward
    Anonymous Coward

    Units of measurement

    Since computers are so fast now that the measurements we use tend to be so long that they're just a meaninglessly long row of zeros, I reckon all computers should be measured in terms of mouse brains.

    That would have an additional benefit - it'll be a few years before the humble mousebrain gets outgrown, so the boffins have plenty of time to figure out how many mouse brains make a rabbit brain, how many rabbit brains make a dog brain and so on. With a bit of luck, they'll get the paris hilton brain in there too, although obviously they'd have to be very quick.

  2. Anonymous Coward
    Anonymous Coward

    Mouse brains...

    How many mouse brains for a cat brain?

    Might make for an entertaining robo-cat hook up then, mice always seem a bit.... Instinctive...

  3. Stephen Gray

    Don't the mice run everything anyway?

    Douglas Adams, a very funny man RIP

  4. Anonymous Coward
    Paris Hilton

    Mouse Nuts?

    ...I'll just fetch me hat.

  5. Karl Lattimer

    WOW

    zOMG someone wrote a neural network that can self organise and match the processing power of half of a mouse brain. This is unbelievable!

    Actually, if someone had actually written the software to do that, I think we'd hear more about it.

    The reality is, they managed to simulate the interaction of bio-chemistry for about half a living mouse brain, big computational task but largely irrelevant for the field of AI, and mostly irrelevant for investigating brain disorders like parkinsons, as you need to know what you're looking for to design it into the "cortical simulator".

    Moreover a mouse brain is fairly trivial to simulate I could in fact probably get an entire mouse brain crammed into an app for PS3, which would mimic a mouse behavior almost exactly, and thought pattern pretty well. Which I think is far more exciting than throwing enormous resources at building vast von neumann machines which have the ability to simulate the interactions involved in tissues.

    Remember a turing machine built in 1942 could still outshine von neumann machines running in the hundreds of megahertz which is millions of times faster than the clock speed of the turing machines in question at the time, running at only a few hundred, maybe a thousand cps.

    Classical computing is dead, more cores does mean faster but its also still following the von neumann design too closely, and that design is wasteful. Thats why Cell BE rocks! It isn't quite like these "other" computers :)

  6. Anonymous Coward
    Anonymous Coward

    Not a brainsim.

    They didn't really simulate a mouse brain... they simulated a neural network of roughly the same shape as a mouse brain, and a lot smaller. I wont admit anyone has actually simulated a mouse brain until one is frozen, sliced, scanned, hooked up to a robotic or simulated body so it can be verified to be performing mouse-like actions, and simulated. Hello, immortality. I give it 60 years, give or take thirty.

  7. Anonymous Coward
    Boffin

    Sure, it's fast...

    ...but it'll be outshone by whatever ridiculous video card comes out next generation anyway. I do, however, wholeheartedly endorse the mousebrain benchmark. Much better than 3dmarks.

    "d00d, i OCd mah 8800gtx, its liek 45.8 mousebrains now, woot!"

  8. Anonymous Coward
    Anonymous Coward

    MBr

    Its pretty obvious the new standard for expressing computing power must be the MBr (MouseBrain), equivalent to 22.8 Teraflops

    We demand the Reg Standards Soviet get on it right away!

  9. Ian

    Vista

    Reckon it'll run Vista and Leopard with any great speed? Can't wait for the next OS release to use up all that spare capacity.

  10. Randall Churchill

    turing machine versus von neumann machines

    I don't believe there was a turing machine built in 1942. In fact, turing machines/universal turing machines are merely artificial constructs, theories more or less. The work Alan Turing was involved with at Bletchley Park with the Enigma machine used equipment that was not general purpose and thus is not considered to be a programmable computer. It was not until 1948/49 than the Manchester Automatic Digital Machine (MADAM) was created and that was the first working example of the Universal Turing Machine. A standard instruction time of 1,800 microseconds, means it executed on average less than 600 instructions per second.

    The so called von neumann machine is a computer design model implementation of a universal turing machine so really, there is no difference in architecture. As for WOW's assertion that a turing machine outperforms a von neuman machine....it really doesn't make sense nor is there much historical accuracy. The only machine one could compare the brit MADAM machine is with the EDVAC (Electronic Discrete Variable Automatic Computer) since ENIAC was not a stored program machine and would not be classified as a universal turing machine architecture. EDVAC required 2.9 milliseconds to perform a multiplication but only .864 milliseconds for an addition. Assuming only additions were performed one could expect a bit more than 1000 ops per second. A von neuman machine running 100s of megahertz in 1942 as WOW asserted just never happened. And thats all I have to say about that :-)

    R

  11. John
    Boffin

    The first working computer

    http://en.wikipedia.org/wiki/CSIRAC

    http://www.csse.unimelb.edu.au/dept/about/csirac/

    This machine had to work for a living.

  12. amanfromMars Silver badge
    Gates Halo

    Testing Your Higher Understanding.

    When you invent a New Real Machine working with Artificial Intelligence Algorithms in Virtual Architectures of Increasing Simplified Webs of Communication, ...NeuReal Internetworking, comparing IT to anything else is a divisive Red Herring ploy to simply maintain a doggedly entrenched and dumb Status Quo position, invariably encouraged by the further idiotic action of throwing good more after bad ....... such as is the Present Status Quo market.

    Such perverse intransigence is easily subverted and rendered ineffective by simply Ignoring the Arrogant Ignorance and Publishing to Web and be Damned.

    The Honorable Members commenting here being invited to Ponder the Honourable Member's comments [parts 1 and 2] here...... http://www.guardian.co.uk/commentisfree/story/0,,2203988,00.html ...... for some XSS Action.....Social Intercourse........ aka Screwing with the System if your with them in DC mode or in AC mode, Fine Tuning IT for AI Beta Higher Performance.

    You realise, of course, that when a Machine does not engage and or reply, IT is broken.

    That usually means Replacement if repair is not possible or prohibitively expensive and not worth the bother/uneconomic. The scrap heap of the skull and bones yard/knacker's yard beckons for duds. The curiosity of the museum for others if considered of valuable note....... but irrelevant history nevertheless as Time marches on into New Real Fields which are not held hostage to the Past.

    IT is all about the Future and Registering IT, is it not, for Sharing? Creating Wide Open Vistas ...not supporting Bird Brain Plans which may be of the Non-Plan variety for the Intellectually Challenged/Deprived/Depraved/Pickled .

  13. Toru Nagisa

    If I may...

    I recommend people go check out Herr Zuse's machines - http://en.wikipedia.org/wiki/Z3

This topic is closed for new posts.