Slashdot Mirror

← Back to Stories (view on slashdot.org)

Six Companies Awarded $258 Million From US Government To Build Exascale Supercomputers (digitaltrends.com)

Posted by BeauHD on from the next-gen-computing dept.

The U.S. Department of Energy will be investing $258 million to help six leading technology firms -- AMD, Cray Inc., Hewlett Packard Enterprise, IBM, Intel, and Nvidia -- research and build exascale supercomputers. Digital Trends reports: The funding will be allocated to them over the course of a three-year period, with each company providing 40 percent of the overall project cost, contributing to an overall investment of $430 million in the project. "Continued U.S. leadership in high performance computing is essential to our security, prosperity, and economic competitiveness as a nation," U.S. Secretary of Energy Rick Perry said. "These awards will enable leading U.S. technology firms to marshal their formidable skills, expertise, and resources in the global race for the next stage in supercomputing -- exascale-capable systems." The funding will finance research and development in three key areas; hardware technology, software technology, and application development. There are hopes that one of the companies involved in the initiative will be able to deliver an exascale-capable supercomputer by 2021.

40 comments

  1. AAAAAAAAnnnd by Anonymous Coward · · Score: 0

    Outsourced to India! For shit made in China!

  2. Socialism, right? by fustakrakich · · Score: 1

    These people don't make enough money? Please! It's just another handout.

    --
    “He’s not deformed, he’s just drunk!”
    1. Re:Socialism, right? by AHuxley · · Score: 1

      When France does this its gets spied on https://wikileaks.org/nsa-fran...
      When the USA does this its essential to "security, prosperity, and economic competitiveness"

      --
      Domestic spying is now "Benign Information Gathering"
    2. Re:Socialism, right? by iggymanz · · Score: 0

      but do we get something for that money compared to the return on say giving a bunch of inner city savages money?

    3. Re:Socialism, right? by Anonymous Coward · · Score: 0

      Yup, we won't get ANYTHING out of the US being a leader in developing the largest, most powerful computers in the world. Nope, nothing, no sir.

      For fuck's sake...

    4. Re:Socialism, right? by fustakrakich · · Score: 2

      No, that only applies to the universities. This stuff will be locked down, if not by copyright/patents, then by being classified. Anyway China is doing it, so why not us, right? Another fine example of state run capitalism, or is it capitalist run state?

      --
      “He’s not deformed, he’s just drunk!”
    5. Re:Socialism, right? by fustakrakich · · Score: 1

      we won't get ANYTHING out of the US being a leader in developing the largest, most powerful computers in the world.

      You become a leader by properly funding and staffing the educational system, not by giving handouts to your multi-billion dollar buddies in the business. Besides most of that money will end being spent on lobbying for more money.

      --
      “He’s not deformed, he’s just drunk!”
    6. Re:Socialism, right? by 110010001000 · · Score: 1

      No it doesn't. Who says government backed research will be public domain and not patented? Such a naive idiot.

    7. Re:Socialism, right? by AHuxley · · Score: 1

      AC Existing nuclear weapons can be better simulated?
      New nuclear weapons testing can be simulated?

      --
      Domestic spying is now "Benign Information Gathering"
    8. Re:Socialism, right? by PJ6 · · Score: 3, Funny

      These people don't make enough money? Please! It's just another handout.

      You're kidding, right?

      Defense spending is supposed to be a handout to tech industry.

    9. Re:Socialism, right? by Anonymous Coward · · Score: 0

      Classified things are not profitable, don't scale and don't get cheaper over time. Assuming there will be a framework on programming these systems, the performance is all applied for the use of science and industry. Drug development require all the power it can get and the 1000 petaflops is barely a start. The exaflop race also have that space race vibe to it, so the Chinese would do it for the national pride alone. Usable performance have actual economic benefits, which appears to be the US viewpoint on the subject.

    10. Re:Socialism, right? by Anonymous Coward · · Score: 1

      better than spending billions on weapons systems that the pentagon doesn't even want

    11. Re:Socialism, right? by thinkwaitfast · · Score: 1

      The educational system is nothing without cultural and community support. One of my engineering profs liked to say that you won't learn anything any better or more at MIT (where she obtained her phd and taught) than you can at a community college.

    12. Re:Socialism, right? by ranton · · Score: 1

      No it doesn't. Who says government backed research will be public domain and not patented? Such a naive idiot.

      Who says it has to be public domain to be a benefit to the public? I use many products whose IP are not in the public domain, and they still provide me more utility than the money I spend on them (or else I wouldn't buy them).

      --
      -- All that is necessary for the triumph of evil is that good men do nothing. -- Edmund Burke
    13. Re:Socialism, right? by Anonymous Coward · · Score: 0

      You know that pretty much every company in Silicon Valley, and the mini-SV's around the country like Charlotte, got their start with some government grant or funding or tax incentive, or was created from some tech developed in some university lab funded with a grant.

      The government actually does a lot of good things for the economy. Not that you'd know it from the media.

  3. US companies? by Anonymous Coward · · Score: 0

    I though IBM was an Indian company

  4. Imagine a Beowulf cluster of these ! by Anonymous Coward · · Score: 1

    obligatory Beowulf reference

  5. Ready for overruns by Reaperducer · · Score: 1

    > with each company providing 40 percent of the overall project cost

    So, the six companies are going to contribute 240% of the project's cost? I guess they're already expecting overruns.

    --
    -- I'm old enough to have lived through six different meanings of the word "hacker."
    1. Re:Ready for overruns by gl4ss · · Score: 2

      no they give the 60% of the project cost of each company and the company pays 40%. that's the individual projects.

      combined they are that 430 something million.

      so sounds like they have multiple projects and each company runs their own project of which they pay 40%.

      or rather.. come on.. it goes like this. the company gets the 60% from goverment and then uses that 60% to run the project as their own funds and pockets the 20% difference as profit.

      --
      world was created 5 seconds before this post as it is.
  6. Amused by Anonymous Coward · · Score: 0

    Just chucked to myself thinking about going ago office space on the these in about 30 years when they are obsolete.

  7. Nothing out of the ordinary here by Anonymous Coward · · Score: 0

    AMD/Intel, HP/IBM/Cray, and Nvidia. That's pretty much expected for any large computing project.

  8. In the era of Quantum computers ... by Mosquito+Bites · · Score: 1

    ... are exa-scale supercomputers still vtital?

  9. Exascale supercomputers by Clived · · Score: 1

    Just trying to keep up with the Chinese, I see ...

    --
    Clive DaSilva Email: clive.dasilva@gmail.com Ubuntu 18.10 Kernel 4.18
  10. Flash player by pablo_max · · Score: 1

    Does this mean that firefox will finally be able to run flash without coming to a complete standstill?

  11. Is it actually useful? by aberglas · · Score: 1

    Sure, there are problems that need some serious computation, but could they also be done with smaller (large) computers working longer? Super computers are generally a shared resource anyway.

    One wonders whether this is a real investment in progress, or just a keep up with the Chinese project. Or like the international space station, nothing to do with science.

    1. Re:Is it actually useful? by Areyoukiddingme · · Score: 1

      Sure, there are problems that need some serious computation, but could they also be done with smaller (large) computers working longer? Super computers are generally a shared resource anyway.

      The Department of Energy likes to use supercomputers to run exceedingly detailed simulations of nuclear weapons, in all sorts of states, from exploding to moldering away inside of ballistic missile submarines. For the exploding part, they do it because the US has signed treaties agreeing not to detonate real ones as tests anymore. For the moldering part, they're trying to verify that they'll still explode properly if needed, again without actually dragging one out of a submarine and trying it out. Being able to run such simulations much faster means being able to make much better simulations.

      There may be some non-military applications, but neither Rick Perry nor Congress gives a shit about those.

    2. Re:Is it actually useful? by Anonymous Coward · · Score: 0

      Dick measuring.

      Tiny hands.

    3. Re:Is it actually useful? by enjar · · Score: 1
      Yes. It's useful, in the same way that research into the extremes of anything is useful. People develop better things all the time and some of those innovations will become changes to mass market goods. Examples of that kind of thing might include carbon fiber (was exotic, now seen in many autos and planes), all the innovations associated with the microprocessor, solar cells, etc. Supercomputing is the far end of the scale of the computing spectrum where really hard problems are tackled, pushing the bounds of what we can do with computing to the limit. In some respects, it's the Formula 1 of computing.
      1. Medical - Drug discovery, proteins, brain modeling all require enormous amounts of computational power.
      2. Climate modeling - more compute power means climate models become more sophisticated. Not only is this valuable for climate change research but also better weather forecasting, which feeds into commerce, farming, air travel, tornado warnings, etc. Weather forecasting is time-dependent -- there's no use in a forecast that's hours or days late.
      3. Nuclear weapons simulation - I'd much rather have a nuke be popped off in the cyber
      4. Machine learning/intelligence/AI work
      5. Chemistry/biology - genetics, chemical reaction simulation, etc
      6. General computing algorithms - making things work at scale often provides insight into better algorithms, and paves the way for when the rest of computing catches up. Maybe in 20 years our smartphones will be capable of exascale computation, although at that point of course they will just be using our brain power since they will be part of our neural lace.
      7. General computing hardware - running at giant scale with networking, storage, etc uncovers new and novel ways to do things better that eventually make their way down to commodity computing hardware.
    4. Re:Is it actually useful? by thinkwaitfast · · Score: 1

      I recall when /. and the nerd community used to be pro-technology and applauded supercomputing developments, it's how I originally came across this site twenty or so years ago. Now it is filled, like the rest of the internet, with anti-science troglodytes. What I don't understand is why they bother coming here.

    5. Re:Is it actually useful? by enjar · · Score: 1

      I agree. As someone whose entire career has been involved with working on tools used by engineers and scientists, it's disheartening to see so many turning their backs on advances in science and technology, especially when the anti-science folks use the fruits of decades technological advancement (Internet, smartphones, high speed mobile communication networks, flash storage, antenna design, touchscreen interface, battery advances, integrated circuit improvements, etc) as the bullhorn for their ignorance.

    6. Re:Is it actually useful? by aberglas · · Score: 1

      You do not answer the question, probably do not know the answer.

      To restate the question, is this a good use of research funding? Is computation a serious limitation. Obviously we all want faster computers so that we can write slower software, but that does not actually advance anything. Is it like the atom smashers, were every extra bit of energy adds a bit more knowledge? Or is it polishing a round ball, with much better uses for the funding, like AI research for example.

      You would not be in a position to answer unless you actually work on the sorts of problems that needs this sort of grunt. E.g. AFAIK most molecular modelling is done with ordinary GPUs, although there are models which require more.

      Vague statements about progress do not address the question. The International Space Station is about progress, but it is also almost completely useless from a science perspective.

    7. Re:Is it actually useful? by enjar · · Score: 1

      Yes, I do believe this is a good use of funding. Some particular areas: 1) weather forecasting and modeling. Weather forecasting is time-bound, so computation speed matters. Places that better weather forecasting saves lives or property include tornadoes, hurricanes/typhoons or effects/severity of winter storms. Tornadoes the place where minutes matter. Given inputs from radars, being able to accurately predict when tornadoes will form and what path they will take can give people more time to get to shelter. For hurricanes and typhoons, similar conditions but do to the fact the storms are larger and move more slowly, you do have more lead time, but forecasters can still have trouble estimating intensity, rain/snowfall amounts, etc. 2) Tsunami detection and forecasting. Given a earthquake, will there be a tsunami and where will it hit? Sounding alarms allow people to get to higher ground. Given that a lot of humanity lives next to the ocean this use case can matter. More computing power means models can be more accurate and alerts more targeted. 3) brain research -- current supercomputers can only handle a bare fraction of even a simple brain, e.g a worm or part of a rat brain, but even with that amount of power there have still be advances made. Scientists are finding that it's not merely a linked node problem, but the 3-d geometry of the brain cells that matter, too. As we live longer and memory diseases such as Alzheimer's become more prevalent, not to mention costly for heath systems to pay for, understanding the brain and developing therapeutics also matter, and a computer simulation allows for better drug testing -- and we need vastly larger amounts of compute capability to model the human brain. 4) GPUs are used in many supercomputers, nVidia has entered the Top 500 Supercomputer list with a machine put together using their DGX machines.

  12. It's down to picojoules by Laxator2 · · Score: 3, Insightful

    I remember an article a couple of years back on this subject, and it explained that exascale computing is not feasible unless the energy cost of moving a single bit around goes down from the picojoules range into the femtojoules.

    The closest reference I can find now is:

    http://spectrum.ieee.org/tech-...

    The relevant part is:

    "Data needs to move on interconnects and they found that even using some really cool emerging technology it still cost 1-3 picojoules for a bit to go through just one interconnect level"

    1. Re:It's down to picojoules by Anonymous Coward · · Score: 0

      What you're positing here is correct then there's something very wrong with this:

      https://en.wikipedia.org/wiki/Exascale_computing

      > it is believed to be the order of processing power of the human brain at neural level (functional might be lower).

      Either exascale computing is perfectly feasible because our brains do it for about 20watt right now, using *CHEMISTRY* for energy and data exchange, or whoever made the estimation of brain processing power is wrong by like 9 orders of magnitude or whatever.

    2. Re:It's down to picojoules by Anonymous Coward · · Score: 0

      The article you posted is from 2008 and seems to show that it wouldn't be possible by 2015 but might be possible by 2020. Emerging technology is probably a bit better than 1-3 picojoules today.

  13. Why do we need this? by Anonymous Coward · · Score: 1

    I thought Trump knew everything! Why don't they just ask him what he believes that answer should be. If the research comes to a different conclusion, they'll just ignore the faulty research anyway.

    1. Re:Why do we need this? by Anonymous Coward · · Score: 0

      Your tears make me laugh.

  14. Peanuts by Anonymous Coward · · Score: 0

    Why not get serious and pour 1.5 billion into one company who can prove their worth?