Announcement

Collapse
No announcement yet.

BOINC, Seti@Home, distributed computing

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • BOINC, Seti@Home, distributed computing

    What would be a good system with respect to computations per unit energy ? And one that could regularly run all night without over-heating at this intensive task ?

  • #2
    Really no PC should overheat.
    The cooling should be designed to handle running at full load.

    As you can see from the temperature graphs here the temperature on a CPU doesn't increase much after the system has been under load for 5min.

    I don't have hard figures handy, but pretty much any chip from the current Intel's new Ivy Bridge range would be fairly efficient. In theory the mobile chips should be better than the desktop chips. So something like the Intel Core i7-3720QM would be right up there in terms of efficiency. Some of the newer Atom chips might also be OK. Low power, but also very low wattage.

    Comment


    • #3
      Originally posted by douglaswilson View Post
      What would be a good system with respect to computations per unit energy ? And one that could regularly run all night without over-heating at this intensive task ?
      Right, waste resources and donate your time helping some ungrateful dorks look for assumed evidence of aliens during an economic depression in your parents basement.

      How about something more productive like donating computing power to find a cure for cancer at Folding@Home?

      Comment


      • #4
        There is a massive amount being spent worldwide by governments and private companies into cancer research.

        On the other hand there is pretty much nothing being spent on SETI, by anyone. Some expenditure is surely justified given the potential (but unlikely I admit) result.

        What's a bigger waste. Computer gaming, or SETI?
        What about 731,000 funny cat videos stored on Youtube for wasted resources? Surely you should be on a crusade to rid the world of funny cat videos before worrying about SETI.

        Comment


        • #5
          I have deleted your post. Too much swearing and hating.
          You are welcome to re-post it if you tidy it up.

          Comment


          • #6
            Thanks a lot. The Passmark temperature graphs are interesting. Looks as if running all night in off peak electricity will not damage the computer. The RealTemp program looks useful for checking. It was my experience that the 8087 used to get very hot (no fan) and that lockups occurred if one overdid the continuous calculations but obviously this has changed. My father's Sinclair also used to go out after it got hot.

            The Atom N2600 ratio of Passmark rating to TDP looks as good as the i7-2760QM's but maybe when the peripheral loads are added the big guy would win.

            Comment


            • #7
              Recently in the news, a Sequoia supercomputer achieved 16.3E15 floating point
              operations per second using only 7.9 million watts. That is, 2063 Mops/watt. If a 3820
              laptop manages 9.0E9 on 65 watts, that would be 138 Mops/watt. This makes me think I
              would do better to donate the laptop purchase and power money to SETI so that they
              could buy time, at least so far as power efficiency goes. I am not sure how much time
              costs in money. I am starting to think that my question was not very good, though your answer correct, a good choice if one were going to go ahead.
              The essential thing I had somehow forgotten is that the main advantage of distributed
              systems like BOINC is that the donors already have the equipment and are using it
              anyway, so that the power efficiency is very high. For example, SETI could get a lot done
              between keystrokes while I spend 5 hours slowly typing a BASIC program into my putative
              3820QM, for little extra power. Everybody knows that. A question that arises is whether
              the donor ought to leave his machine turned on after finishing his own work. Would it do
              the world enough good to justify the power ? Perhaps for my P3, not any more, but for a
              3820QM, maybe it would. Very difficult question, and a subjective evaluation is involved,
              eg SETI vs cancer research, as well as the science of climate change (and resource
              depletion, this machine was manufactured, better use it as much as possible ?).
              I have research left to do on the question of whether using a gpu for BOINC science could
              up the efficiency to become competitive with the supercomputer. It is difficult to find data
              on gpu power and computations per second, and things are rapidly evolving. One of the
              BOINC projects says that the gpu can be 2 to 10 times faster at scientific calcs than the
              cpu, but they don't advise one on what models are involved, to help plan a purchase. The
              gpu may be more useful for some projects than others.
              I am an admirer of SETI. Dorks or not, one never knows what their work may lead to. The
              Schneider trophy seaplane races pre WWII led to the Spitfire. The demonstrators against
              F1 races probably would not have approved of those races. Yet, with no Spitfire, possibly
              the Battle of Britain would have seen British airfields and aircraft factories bombed out
              totally, the RAF shot out of the air, and the English Channel left open to an invasion fleet.
              Possible NAZI world conquest. No more Jewish doctors to fight cancer. One never
              knows. Maybe intelligent extra-terrestrial life forms could help us with medical problems. I
              cannot guess what will be the future of the human race nor what exactly we should be
              doing (and in any case we probably would not do that).
              And we recently have the example of the gamers solving a science problem.

              Comment

              Working...
              X