Slashdot Mirror

← Back to Stories (view on slashdot.org)

Chinese Tianhe-1A Supercomputer Starts Churning Out the Science

Posted by Soulskill on from the thousands-of-feynmans-per-hour dept.

gupg writes "When China built the world's fastest supercomputer based on NVIDIA GPUs last year, a lot of naysayers said this was just a stunt machine. Well, guess what — here comes the science! They are working on better material for solar panels and they ran the world's fastest simulation ever. NVIDIA (whose GPUs accelerate these applications as a co-processor) blogged on this a while ago, where they talk about how the US really needs to up its investment in high performance computing."

22 of 103 comments (clear)

  1. Re:When will it be used for its real purpose... by snookerhog · · Score: 5, Funny

    you mean it is not for mining bitcoins?

  2. No shit? by Lunix+Nutcase · · Score: 2

    NVIDIA (whose GPUs accelerate these applications as a co-processor) blogged on this a while ago, where they talk about how the US really needs to up its investment in high performance computing."

    This just in: Company who makes GPUs for supercomputers thinks that people should buy more of their GPUs.

    1. Re:No shit? by Anonymous Coward · · Score: 2, Insightful

      Debt is a distraction. Ask Dick Cheney, "Reagan proved that deficits don't matter."

    2. Re:No shit? by dbIII · · Score: 2

      As just about everything Cheney did proved, nothing he said mattered because it was unlikely to match any sort of truth or reality. Hopefully that episode of a VP acting like a medieval prince and pissing all over the ideas of a Republic and a Democracy will not be repeated for a while.
      Go ask the UK, Iceland, Greece, Spain etc if debt is merely a distraction. Going too far down that road would mean that the USA would learn about how the USSR fell apart in Reagan's time - only they'd be watching it on their own streets.

  3. Re:Ah, the science of cracking passwords. . . by Lunix+Nutcase · · Score: 2

    Rendering photo-realistic 3D tentacle porn for the Japanese?

  4. A super computer.doing science? by Anonymous Coward · · Score: 2, Funny

    Is there cake?

  5. Re:the US is broke by nigelo · · Score: 2

    > US really needs to up its investment in high performance computing

    More like: 'we've upped our investment, now up yours!"

    --
    *Still* negative function...
  6. Re:Not the fastest by the+linux+geek · · Score: 2

    Anyone that measures aggregate CPU performance on a cluster in "GHz" is an idiot.

  7. its not fucking science. by nimbius · · Score: 2

    its a half-page blog post from the systems vendor, nvidia, singing the praises of their customer. as a computer scientist i want an independent scientific review...so should you, slashdot 2.0

    "Andy Keane joined NVIDIA in 2006 as GM of the new GPU Computing business unit. Previously, he was a VP at start-ups Morphics and Ageia, which focused on the development of parallel computing technologies for the telecoms and consumer industries. "

    oh goodie...the blog isnt even from a scientist at nvidia, just one of their general managers.

    Ill save some time for ya, the other link is just third party shit from HPCwire singing the praises of nvidia, plugging their NASDAQ tag, and once again singing the praises of nvidia with sponsored ads.
    oh, and of course authors will insist [insert potential rival competitor country here] invest more in high performance computing.

    --
    Good people go to bed earlier.
  8. has HPC really done anything useful? by Anonymous Coward · · Score: 2, Interesting

    I only know of minor successes in materials engineering and genetics but even those weren't very commercially successful. A course I took at MIT about 6 years ago the professor was talking about saving 10 million for Ford using HPC materials engineering modelling but 10 million isn't very impressive. I always hear HPC "is working on" a bunch of exciting sounding stuff but I never have heard of any big successes. Has there been any big achievement of HPC other than rendering 3D movies quicker, breaking codes during wars, and helping governments spy on their own people?

    1. Re:has HPC really done anything useful? by stox · · Score: 2

      Go talk to NOAA about weather reports. A fine example of success and HPC.

      --
      "To those who are overly cautious, everything is impossible. "
  9. Re:I've Had It With Slashdot by Anonymous Coward · · Score: 2, Interesting

    but, where are ya gonna go?

  10. And talk about BS by Sycraft-fu · · Score: 3, Insightful

    "The US really needs to up its investment in high performance computing."

    Really? Well let's have a look at the top500 list shall we? Now while Linpack leaves some things to be desired (like the fact that clusters perform better on it than they can on some things like particle simulation) but it is the standard.

    So of the top 10 system 5, half, are in the US. #2, 5, 7, 8 and 10 are US computers. The next 10? 7 of them in the US. So 12 of the top 20 are US systems. Most of them are US government systems too, and the ones that aren't are university. So not like "Big companies that happen to be in the US," but research universities and government research agencies like the DOE and so on.

    It also isn't stopping. The Advanced Simulation & Computing Program (formerly ASCI) is going on and more computers are being built. LANL has chosen Cray to build the next one, Cielo. It is partially complete and online now, though still being added to. Currently #10 on the list, it will move up a bit when it is complete.

    So I don't see the US as not investing in this.

    What I see them not doing is buying nVidia GPU based systems, which is fine. GPUs are neat and all but they are not as general purpose as CPUs. They are stream processors. Now if your project is one that stream processors are good at then great. However not all projects are. Particle simulation, that I mentioned before, is one that stream processors aren't so good at. So just because something is higher on the Linpack based top500, doesn't mean it is faster at everything.

    Regardless, it isn't like the US needs to be on the number 1 spot to be good. The more telling fact is how many are there, that the US has over half of the systems in the top 20. Clearly supercomputers are something the US has plenty of. Keep going down the list and you keep seeing a lot of US systems.

    This is just nVidia trying to play on the whole patriotism thing to make sales.

    1. Re:And talk about BS by Jah-Wren+Ryel · · Score: 2

      So of the top 10 system 5, half, are in the US. #2, 5, 7, 8 and 10 are US computers

      On the other hand, 10 years ago the US had 7 of the top 10, including #1, #2, #3 and #4.

      --
      When information is power, privacy is freedom.
    2. Re:And talk about BS by wisty · · Score: 2

      On the other hand, 10 years ago most Chinese were riding bicycles. Now they have electric scooters, and are lusting after cars. And it's a much larger country than the US.

      It's a matter of GDP multiplied by government commitment to HPC. Both China and the US have a similar commitments to HPC, so I'd expect China will overtake the US in 10 years.

      As an Australian, this doesn't bother me. If we are number 1 per capita, we are happy. The US wants to be number 1, which will simply not happen when China is 3 times the size. In 10 years, China will be the arrogant giant that pats itself on the back for being number 1, even if it has low achievements per capita.

      The US will have to stop considering itself a giant when the rest of the world grows up.

  11. US innovation is dead by gblfxt · · Score: 2

    anything we put into new technology just ends up in the patent portfolio's pocket with their blanket 'innovations'

  12. Re:When will it be used for its real purpose... by sortius_nod · · Score: 3, Insightful

    I think The Chinese government has much better ways of spying on people than to hog processor time on such a public machine. He'll, they don't even need to break RSA keys anymore.

    I love how people jump to the conclusion that any powerful computer that we learn about in China is suddenly s security threat. It's the ones they don't tell you about that you should be concerned about.

    Funny how people marvel if it's the US but run in fear if it's China. Both have terrible track records when it comes to human rights.

  13. Re:the US is broke by AHuxley · · Score: 2

    Dont worry the CIA understands China, like it understood the Soviets.
    At anytime they can induce cults, fads, music, addictive sci fi, secessionist movements, hardware, software issues, payment issues, NGO "outrageous", intellectual property, middle class rights, property bubbles ect.

    --
    Domestic spying is now "Benign Information Gathering"
  14. Re:the US is broke by dbIII · · Score: 2

    Didn't you even read the summary? They didn't steal it, they bought it from Nvidia just like anyone else can.

  15. And? by Sycraft-fu · · Score: 2

    Is the US the only country that is allowed to be any good at this? To me it seems the story isn't the US investment in supercomputing is dropping, because it isn't. The story is that other countries are investing more in it. That is a good thing IMO. The US shouldn't be the be-all, end-all of research.

    Also please remember the nature of these things are that you get displaced. You are #1, someone builds a new #1. It isn't like you can just whack out a new one. Supercomptuers take a lot of time and money to build.

    1. Re:And? by Sycraft-fu · · Score: 2

      Again, you are under the impression that this is some kind of contest the US is "losing" that if the US doesn't have the #1 spot and most of the other high spots that is a "problem" for the US.

      It isn't. It is a win for the whole world. I think it is more false patriotism BS. Too many people in the US have this idea that if you aren't the top in everything, it is a complete and total failure. You see this with things like people bemoaning the US becoming #2 to China in manufacturing output (happened this year), as though the world will end if you aren't #1.

      Well guess what? I've traveled the world a bit and been to a number of countries not the US. They are not #1 in most things, often not #2, #3 or so on. Turns out they are quite nice places to live, with happy people and stable societies. Canada, Norway, England, all places I'd be happy to live and they are "#1" in almost nothing.

      My point is it doesn't matter that the US isn't #1 on this list, in part because Linpack isn't the worlds best benchmark, but also because it distorts the issue. If the US was spending less on supercomputing, if it was getting out of that kind of research, that might be problem. However it isn't, it is spending as much as ever. It is just other countries are getting in to it too.

      Life isn't a race where only one person, or one nation wins, and everyone else loses and it sucks for them. We can make things better for all of us and we can all come out better for it.

  16. Re:That's the best they can do? by ScrewMaster · · Score: 2

    You provide inexpensive plentiful electricity free of emissions and you've just put the world's most profitable companies (The Oil Industry) on notice.

    Not at all. The reality is that petroleum is far too useful and valuable a substance to be burned for power. That's a waste for which future generations will revile us. Fact is, there will still be plenty of uses for oil even after the last internal combustion engine is melted down for scrap.

    --
    The higher the technology, the sharper that two-edged sword.