Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
Supercomputing Science

Half-Petaflop Supercomputer Deployed In Austin 130

Posted by kdawson
from the one-teraflop-per-researcher dept.
SethJohnson writes "Thanks to a $59 million National Science Foundation grant, there's likely to be a new king of the High Performance Computing Top 500 list. The contender is Ranger, a 15,744 Quad-Core AMD Opteron behemoth built by Sun and hosted at the University of Texas. Its peak processing power of 504 teraflops will be shared among over 500 researchers working across the even larger TeraGrid system. Although its expected lifespan is just four years, Ranger will provide 500 million processor hours to projects attempting to address societal grand challenges such as global climate change, water resource management, new energy sources, natural disasters, new materials and manufacturing processes, tissue and organ engineering, patient-specific medical therapies, and drug design."
This discussion has been archived. No new comments can be posted.

Half-Petaflop Supercomputer Deployed In Austin

Comments Filter:
  • Now We Know (Score:5, Funny)

    by Nom du Keyboard (633989) on Saturday February 23, 2008 @08:55PM (#22531320)
    So now we know why there is such a shortage of quad-core AMD Opterons otherwise.
    • Re: (Score:3, Insightful)

      by Chris Snook (872473)
      If I had mod points, I'd mod this insightful, not funny. There are a lot of HPC projects that were planning to use Barcelona, that were held back by the TLB bug. I'm sure anything approaching this magnitude already had a contract with AMD that includes guaranteed delivery dates and penalties, either directly or through the OEM. If you don't have a signed contract with AMD or with someone who has one with AMD, you're going to have to wait in line.
      • by edsousa (1201831)

        I'm sure anything approaching this magnitude already had a contract with AMD that includes guaranteed delivery dates and penalties
        They had and wasn't met! Ranger was scheduled to be ready for inclusion in Top500 list released on November 15th.
  • by Anonymous Coward
    that it completes an infinite loop in only 5 seconds!
  • Apostrophes (Score:3, Funny)

    by gardyloo (512791) on Saturday February 23, 2008 @09:00PM (#22531354)
    Perhaps it can run spell- and grammar-checks on Slashdot submissions!
    • How many supercomputers would be needed to spell check for a million monkeys?
      • Re: (Score:3, Funny)

        by somersault (912633)
        There's a lot of variables here. Sadly I've just spent the last 5-10 minutes of my life considering them all and was writing up a post about it. Then I realised that sometimes, I take jokes waaaaay too seriously.
        • by pentalive (449155)

          Sadly I've just spent the last 5-10 minutes of my life considering them
          I would have thought there were better things to think of in the last 5 minutes of your life...

    • It's peak processing power of 504 teraflops will be shared among over 500 researchers working across the even larger TeraGrid system. Although its expected lifespan is just four years ...
      What I don't understand is how the editor could it wrong and then in very next sentence get it right. Do the Slashdot editors just toss a coin to see if they should use an apostrophe?
    • Nah. They should donate the spare cycles into running slashdot. Can you imagine how fast the main page could be updated?
  • by djupedal (584558)
    Ranger will provide 500 million processor hours to projects attempting to address societal grand challenges such as:
    • global climate change - btdt
    • water resource management - nimby
    • new energy sources - boring
    • natural disasters - omg!
    • new materials and manufacturing processes - yesterday
    • tissue and organ engineering - day before yesterday
    • patient-specific medical therapies - yeah, right...when pigs fly
    • ...and
    • drug design - Well! Hello! In that case, count me in and please proceed!
  • by Jugalator (259273) on Saturday February 23, 2008 @09:07PM (#22531404) Journal
    A: The more flops, the more powerful it grows.
  • AMD (Score:5, Interesting)

    by milsoRgen (1016505) on Saturday February 23, 2008 @09:09PM (#22531420) Homepage
    I'm glad to see AMD based projects like this, as they have certainly took a hit in the HPC space [top500.org] as of late.
    • Couldn't see details, but this may use Sun's hypertransport switch as an interconnect. Until Intel's next generation of chips with QPI, you couldn't do that sort of interconnect with Intel processors. Admittedly though, I'm not convinced that it is significant enough a benefit over recent Infiniband solutions despite the penalty of going through an Infiniband chip and then a PCI express controller.

      Even with the L3 errata straightened out, it still looks to be a rough road for AMD, who hasn't demonstrated
      • Re: (Score:3, Informative)

        by Wesley Felter (138342)

        Couldn't see details, but this may use Sun's hypertransport switch as an interconnect.
        Sun doesn't make a Hypertransport switch and Ranger uses Infiniband just like other high-end x86 clusters.
      • by afidel (530433)
        Power and cooling penalties? Are you looking at the same spec sheets I am because I'm seeing better performance per Watt out of Barcelona systems than out of Intel quad core Xeon's based systems. Most of it has to do with the fact that Intel uses power sucking FB-Dimms, but that's a design tradeoff that Intel made.
        • I'm saying in order to get the same number of flops and using price-performance instead of straightforward performance, you must increase node count. If the processor performance/watt *was* better (I believe with the 45 nm process on Intel's side for the moment, a Xeon 2.33 quad core comes in a 50W TDP variant, for example, Barcelona comes in at 95W TDP, goes along way toward offsetting the FB-DIMM power), you still have to worry about more AC power supply inefficiencies, general power usage of extra moth
  • Four Years? (Score:1, Funny)

    by dreamchaser (49529)
    What, is it an android from Blade Runner? It's going to die in four years? Slashdot summaries are the Suxx0rs!
    • by aktzin (882293)

      What, is it an android from Blade Runner?

      I think you mean replicants. Androids, like Lt. Commander Data from Star Trek, have practically unlimited lifespans.

      • I think the word android can be used since the book (by Philip K. Dick, 1968) that the movie Blade Runner was based upon was titled Do Androids Dream of Electric Sheep?.
    • by Lally Singh (3427)
      Nope, but I'll be scanning ebay in 2012! Daddy needs some Distributed.Net action!
  • by SEWilco (27983) on Saturday February 23, 2008 @09:18PM (#22531484) Journal
    And the other half is deployed in Dallas?
  • It's nice to know we're that much closer to skynet, but... won't there always be a faster computer?

    I think more shocking news would be "No supercomputing records beat this year!" But I suppose that would just be fiction.
    • Re: (Score:1, Informative)

      by Anonymous Coward
      There already is a faster computer. Ranger's peak is over 90 TF slower than BG/L at LLNL so unless there are some amazing efficiency breakthroughs (doubtful), it isn't going to be number one in June. That, of course, doesn't mean BG/L will still be number one either.
  • (15,744 processors) * (4 cores/processor) * (24 hours/day) * (365 days/year) * (4 years) = 2,206,679,040 core hours

    Seems like the "processor hours" metric needs some adjustment to account for multi-core. Otherwise I could build one of these with 15,744 single-core processors and claim the same performance.

    • It is donating 500M hours to research. Article never says how many total hours are available.
    • Re: (Score:1, Interesting)

      by Anonymous Coward
      ...How is this different from a regular supercomputer? Supercomputers are already parallel - this just adds a second layer of parallelism. Competing supercomputers have always had different economy on a "processor hour", multicore CPUs is just one technique to increase the processing power of the machine.

      And you could *TRY* to build a 15,744 single core machine and claim the same performance, but it would all fall apart very very quickly when someone asks "how many FLOPS?" (which is what computing power
    • Re: (Score:2, Insightful)

      by EvanED (569694)
      Seems like the "processor hours" metric needs some adjustment to account for multi-core. Otherwise I could build one of these with 15,744 single-core processors and claim the same performance.

      Why are you associating processor-hours with performance anyway? You could hook up 15,744 286s and get the same number of processor-hours too. So why don't you complain about that?
  • Gee, this computer is the BIGGEST flop generator of them all!

    That's too bad.

    NO! That's GOOD!

    It is?

    Yeah, lots and lots of flops per second - the more the better!

    So the bigger the flops, the better?

    Right!

    Fewer flops is bad?

    You got it!

    And researchers want more money for more time with bigger flops?

    Now you get it!

    So they got $59 million for this humongous flop generator?

    Yep!

    Why don't they just burn the money if they want to generate a really big flop?

    That wouldn't work - that wouldn

  • 4 year lifespan (Score:5, Interesting)

    by DeftPunk79 (1232522) on Saturday February 23, 2008 @09:29PM (#22531542)
    The 4 year lifespan in the /. article refers to the amount of time the award money covers for operations costs. So if it finds some others mean s of operation funds it could live longer... of course those funds will probably be from a private organization and the ranger would no longer be open for research.
    • Re: (Score:3, Insightful)

      by pimpimpim (811140)
      Within four years the Performance/Watt ratio will have dropped compared to state of the art, so it would make very little sense to keep the thing taking valuable computer room space and working hours of the technical staff. It happens with all supercomputing machines, just Moore's law in practice. What I think is still a big problem is that there are still many problems getting the hardware work correctly in parallel. Often half a year or longer is lost debugging file system/network issues, which is a consi
  • freaking lame, 59 million dollars... i go to UT, and if this means no more free condoms then we're gonna have words.
    • Re: (Score:3, Funny)

      by chillax137 (612431)
      Just grab a few and you'll be set for the next four years...

      Seriously though, this money comes from an NSF grant earmarked specifically for this project. We get these kind of complaints from other departments and especially undergrad editorials in the student newspaper. Unfortunately, the budget from the football team won't be used to renovate the social work buildings.
    • by LilGuy (150110)
      Do what most other slashdotters do... use a paper towel or a sock.
  • by Tmack (593755) on Saturday February 23, 2008 @09:34PM (#22531594) Homepage Journal

    attempting to address societal grand challenges such as global climate change, water resource management, new energy sources..

    With that many cores, they will need to find new energy sources just to power it, and re-think water resource management as they redirect the river to cool the thing and to prevent it from causing global climate change itself!

    Tm

  • by fabu10u$ (839423)
    In their backyard, and Sun gets the job instead. (Maybe this is why Dell has started offering AMD?)
    • by tomhudson (43916) < ... <nosduh.arabrab>> on Saturday February 23, 2008 @10:10PM (#22531834) Journal

      "In their backyard, and Sun gets the job instead."

      Do you really want to have a $59M machine dependent upon Dell customer support?

      • All i could think of is, after 4 years, i am sure the seti, or folding projects would love to have that system. I wouldn't mind it either.
        • Odd to have a computer of this magnitude only in use for four years, though that might speak of its power inefficiency perhaps in light of new technology in 2012.

          BUT

          The year 2012 is the end of the Mayan calendar. This computer may have been actually assigned to find out what is going to happen in 2012. So it better have the answer in 4 years.

    • by Temkin (112574)

      In their backyard, and Sun gets the job instead. (Maybe this is why Dell has started offering AMD?)
      Sun has a reasonably large Austin presence as well. That's where they designed the Niagara processor.

    • by Hadlock (143607)
      AMD has been trying to build a plant in Austin for years now. This is most likely a deal deeply connected with finally breaking ground on a plant people have been protesting for YEARS now, due to it's location (near watersheds, etc)
  • a witness [wikipedia.org] to the vandalism is quoted saying..

    It's burst into flames! It burst into flames....It's crashing!....Oh the humanity!....This is the worst thing I've ever witnessed.


    Oh wait, you said deployed, not destroyed..

    carry on
  • Hey. Is a Googleplex ever counted in as one of these big a$$ multi-processor monsters? If not, why? Those suckers are ginormous. R
    • I don't think the googleplex shares cycles quite like this.

      I remember reading the infiniband switch main inventor's bio while working at Sun , interesting guy who had these types of things on his mind all the time.
  • actually... (Score:5, Interesting)

    by elite1789 (1245036) on Saturday February 23, 2008 @10:44PM (#22532032)
    I was at TACC a few weeks ago, and the peak performance was around 519 teraflops.... Sadly, they also said the word on the street is that IBM wont take too kindly to the new king in town, and since TOP500 is biannually, everyone is biting their nails about blue-gene getting a quick upgrade in time to stay on top. Turns out the blue-gene systems are so scalable its quite easy to strap a few thousand new processors for a nice performance boost.
    • Re: (Score:2, Insightful)

      by Anonymous Coward
      BlueGene/L at LLNL already peaks significantly faster than Ranger. The only question is whether Ranger can get a =sustained= number that passes BG/L and considering the difference in peak performance, it's unlikely. June is going to be an interesting list as there could be quite a bit of shuffle at the top.

      You are also correct about the scalability of BG. If you look at last June's list and last November's list you'll see a big difference in performance for BG/L. That's entirely due to simply adding mor
  • Chuck Norris doesn't need high performance supercomputers, he can threaten any old 486 into running twice as fast.
    • Wowee.. at under 10 MFLOPS that's got to be some beast of a machine! Or do you mean get the 486 running twice as fast as a supercomputer? I don't think we need another sun in this solar system :o
  • Imagine a Beowulf cluster of those
  • And the answer is apparently yes. According to techtarget.com [techtarget.com] It'll be running CentOS just like slashdot [slashdot.org] does.
  • Sure they will work on climate calculations.. but first on the agenda is being able to play Crysis in hi-res with all options on.
  • how many bogomips per petaflop?
  • by HockeyPuck (141947) on Sunday February 24, 2008 @12:28AM (#22532608)

    Ranger comprises 3,936 compute nodes in a Sun Blade(TM) 6048 Modular System with 15,744 Quad-Core AMD Opteron(TM) processors, and Sun Fire(TM) x4500 servers providing 1.7 petabytes of storage.
    Since TFA says this hardware will last only four years, what typically happens to these supercomputers built out of so called commodity hardware? Is sun going to donate/resell this gear? Or does it end up in the scrap heap? Surely, these Sun Blades are supposed to have a useful lifespan greater than four years. Sun could probably give these blades to every elementary school in all of Texas. Is the future of super computing aka disposable computing?
    • by Albert Sandberg (315235) on Sunday February 24, 2008 @04:36AM (#22533592) Homepage
      It's actually the best way to aquire the new AMD Opteron processor, just wait for 4 years and get in line.
    • Four years is not how long the hardware will last, it's how long the $59 million funding for covering operation costs will last. After that they will have to get money from somewhere else.
    • Probably this means that the system is funded for 4 years. About 2 years in, they will try to renew their grants. If they do, they will probably upgrade the system at about 3 years with a mid-life kicker. This would probably be new blades using 8-core opterons and 2-4 times as much memory. They might even get a second kicker if AMD comes out with a socket-compatible upgrade. After that it's probably time for a forklift upgrade.

      This cycle is fairly typical for HPTC systems, and a 6 year total life-span is pr
  • It sounds like they don't need faster super computers but instead to narrow the number of areas they are crunching away at. Why not pick the top 2 or 3 issues and crunch away at those instead of running 20 jobs, all of which will hardly get anywhere in the four years this supercomputer has to live?
    • They've put quite a bit of thought into how to efficiently allocate time on a 60 million dollar compute cluster, I can assure you. If some jobs get time, it's because non-trivial progress can be made on those things in the time they've been allocated.

  • UT students have been pushing against putting a polluting plant in Austin (very green-aware city) for years. This is just a move to placate the protestors, I'm sure. 4 years? What happens after that?
    • What does this polluting plant have to do with anything? I can't see how they would have decided to build a 59 million dollar supercomputer to placate protesters that were protesting against something completely irrelevant to supercomputers.
      (to answer your question, after 4 years the computer loses public funding, but doesn't cease operation.)
  • by Anonymous Coward
    Some fraction of this machine was originally supposed to be in production in May of last year (a requirement of the original request for proposals), but as far as I know it wasn't even accessible to friendly users until some time last fall. I don't understand how TACC, Sun, and/or AMD avoided getting hit with penalties from the NSF.
  • I wonder if this supercomputer has bandwith caps like most aussies.
  • Computing at Ludicrous Speed!

  • From "The Hitchhiker's Guide to the Galaxy" by Douglas Adams

    Chapter 25
    There are of course many problems connected with life, of which some of the most popular are Why are people born? Why do they die? Why do they want to spend so much of the intervening time wearing digital watches?

    Many many millions of years ago a race of hyperintelligent pan- dimensional beings (whose physical manifestation in their own pan-dimensional universe is not dissimilar to our own) got so fed up with the constant bickering about
    • OK, the end of the skit from Hitchhiker's Guide:

      Chapter 26
      "Yes, very salutary," said Arthur, after Slartibartfast had related the salient points of the story to him, "but I don't understand what all this has got to do with the Earth and mice and things."

      "That is but the first half of the story Earthman," said the old man. "If you would care to discover what happened seven and a half millions later, on the great day of the Answer, allow me to invite you to my study where you can experience the events yoursel

Physician: One upon whom we set our hopes when ill and our dogs when well. -- Ambrose Bierce

Working...