Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Supercomputing It's funny.  Laugh. Power

Students Power Supercomputer with Bicycles 148

inkslinger77 writes "A team of ten MIT students powered a supercomputer for twenty minutes by pedaling bicycles. They duly claimed the world record for human-powered computing (HPC). They powered a SiCortex SC648 supercomputer with a Linux cluster of 648 CPUs and almost 1TB of main memory in a single cabinet. The system is low-powered and draws 1,200 watts without needing special power supplies or cooling..."
This discussion has been archived. No new comments can be posted.

Students Power Supercomputer with Bicycles

Comments Filter:
  • Question (Score:5, Funny)

    by Misanthrope ( 49269 ) on Thursday December 20, 2007 @03:15AM (#21762204)
    One MIT student is how many foot pounds per second?
    • Re:Question (Score:5, Funny)

      by glomph ( 2644 ) on Thursday December 20, 2007 @03:18AM (#21762228) Homepage Journal
      MIT measures its resident nerds using the International Smoot scale.
    • Re:Question (Score:5, Funny)

      by proudfoot ( 1096177 ) on Thursday December 20, 2007 @03:45AM (#21762346)
      Imagine a beowulf cluster of these! ...i hope they wear deodorant.
      • Re: (Score:3, Funny)

        by Walt Dismal ( 534799 )
        Hell, imagine a supercomputer cluster powered by French bicycle racers on steroids. With no deodorant. And garlic. in a small room.
    • Re:Question (Score:5, Funny)

      by Valdrax ( 32670 ) on Thursday December 20, 2007 @04:33AM (#21762546)
      Well, assuming a lack of friction or air resistance and perfectly spherical MIT students...
    • Re: (Score:2, Interesting)

      10 guys, 1200watts... looool 20 minutes? even more lol... at my last performance diagnostics i had 320watts at the lactacid threshold, wich means this can be done for HOURS! ;)
      • Re: (Score:3, Informative)

        by Yetihehe ( 971185 )
        You must be superhuman. Best cyclists can endure 200 wats for few hours. I could make 200w for 5 minutes (tested on ergometer [wikipedia.org]).
        • Re:Question (Score:5, Informative)

          by Simon Brooke ( 45012 ) <stillyet@googlemail.com> on Thursday December 20, 2007 @09:00AM (#21763852) Homepage Journal

          You must be superhuman. Best cyclists can endure 200 wats for few hours. I could make 200w for 5 minutes (tested on ergometer [wikipedia.org]).

          Actually, you're well wrong. Cycling, like anything else, is about power to weight ratio. Lance Armstrong [cruciblefitness.com], in training, could output 6.8 watts per kilogramme, which on his body weight of 74Kg is just over 500 watts. Some of the big fast guys (Tom Boonen [tomboonen.com], Magnus Backstedt [magnusbackstedt.com]) can sustain really startling outputs for long periods of time. I have a friend who peaks around 600 watts and can sustain better than 400 all day; but on his 102Kg that's only 6 watts per Kg. That's strictly amateur - he's fast for us, but he couldn't compete with the pros.

          There's nothing at all hard about 200 watts. Any club cyclist who can do a ten mile time trial at evens is sustaining 4.8 watts/Kg for half an hour. I can do that, and I'm probably old enough to be your grandfather. I weigh 82Kg, so for me that's 381 watts. And I can certainly do 200 watts (13.5mph on the flat, for someone my weight) for eight hours continuous without difficulty.

          • by ahaile ( 147873 )
            www.analyticcycling.com is a cycling geek's best friend.

            And your numbers are just as bad as the guy above you, only in the opposite direction.

            6.0 w/kg, which you say your "amateur" friend can do, is enough to win many Grand Tour climbs. Lance was only ever confirmed at 6.3. Your average recreational cyclist is around 3, beginning racer around 4, local elite racer 5, pros around 6. See this chart [cyclingpeakssoftware.com].

            13.5mph is about 100 watts, not 200.
    • Says the calculator. [google.com]

      A general rule of thumb is that one barrel of oil is the energy equivalent of about one year of hard labor for a human.
    • How many pedal-flops is one MIT student?
  • Heh... (Score:5, Funny)

    by sapphire wyvern ( 1153271 ) on Thursday December 20, 2007 @03:16AM (#21762208)
    So, the students don't run Linux... they bicycle it instead!
  • by CookieOfFortune ( 955407 ) on Thursday December 20, 2007 @03:23AM (#21762244)
    imagine how long the students of any other university would be able to power them!
    • by Adambomb ( 118938 ) on Thursday December 20, 2007 @04:53AM (#21762640) Journal
      Free beer for every kilowatt-hour, Energy crisis solved.

      "Screw ethanol, we're green the hops way! Take that corn industry!"
      • by irtza ( 893217 ) on Thursday December 20, 2007 @05:56AM (#21762884) Homepage

        Free beer for every kilowatt-hour, Energy crisis solved.

        "Screw ethanol, we're green the hops way! Take that corn industry!"

        Well, your solution apparently still requires ethanol to be burned. Just that the powerplant is in human form.
        • was more ripping on the fact that the focus on corn-based ethanol is a rather back-scratchy situation, but i hear ya =)
      • by mpe ( 36238 )
        Free beer for every kilowatt-hour, Energy crisis solved.
        "Screw ethanol, we're green the hops way! Take that corn industry!"


        Maybe you could have they wash tortias down with their beer.
        Of course the real issue with the "corn industry" is it isn't necessary to make fuel ethanol from perfectly good food. Far more sensible would be to either use a a waste product or a plant which grows as a weed on non agricultural land.
      • Free beer for every kilowatt-hour, Energy crisis solved.

        "Screw ethanol, we're green the hops way! Take that corn industry!"

        And, many American brands of beer actually use corn as some of the sugar for making beer.

        So, you're not gonna get away from corn that readily.

        Cheers
  • In the future, which I imagine shall be very much like "Mad Max", this is what shall be required to run SP3 of Windows Vista....
  • Matrix? (Score:4, Funny)

    by ThreeGigs ( 239452 ) on Thursday December 20, 2007 @03:28AM (#21762272)
    Oh great, next they'll invent SkyNet.

    If one of those guys goes by the nick 'Neo', I'm gonna get worried.
  • by LaskoVortex ( 1153471 ) on Thursday December 20, 2007 @03:39AM (#21762306)
    They should have just gone over to the nearest administrative offices and unplugged all of the CRTs running "screen savers". This would have freed power to run the computer even longer and wouldn't have been as tiring.
    • You do realize that SiCortex makes mips clusters right? They are already low power and do not have a CRT attached. They typically run a modified version of Gentoo, you just ssh in and submit a job to the scheduler, just like any HPC (expect maybe Windows HPC, I don't know how that works).

      Unplugging CRTs would not help at all because there aren't any CRTs.
  • by CrackPipePls ( 1205568 ) on Thursday December 20, 2007 @03:46AM (#21762356)
    Not really that useful and interesting How about taking a look into "powered by kicking" technology, make it available to the average joe, millions of kWh of power will be saved across the globe on a daily basis from frustrated computer users
    • must...resist...making...Chuck...Norris...reference...
    • by lekhak ( 874006 )
      For "Powered by kicking" technology you need to run on Windows 98 - Not sure a whole lot of people want to go back to Windows 98. Vista seems promising on that front with the improved CPU / memory hogging technologies.
  • Oh dear. (Score:5, Funny)

    by ChePibe ( 882378 ) on Thursday December 20, 2007 @03:49AM (#21762378)
    It used to be research.

    Then little errands. Get the professor a coke, pick up his stuff.

    Then acting as tech support.

    Then doing all the prof's work for him.

    And now... running the system by the sweat of your brow.

    We must draw the line somewhere, folks. Free labor has its limits!

    This cannot bode well for graduate students... well... maybe for the chunky among us... but let's hope this doesn't catch on. I can already see profs carried about on the shoulders to and from meetings...
    • This cannot bode well for graduate students... well... maybe for the chunky among us... but let's hope this doesn't catch on. I can already see profs carried about on the shoulders to and from meetings...
      Clearly you haven't been to Cambridge...
    • Isn't the system entitled to the sweat of your brow?
  • An Excerpt from the article:: "An SC648 chip, with six processors on it, draws around 8 watts of power, which compares to a typical notebook computer CPU needing 100 watts, according to SiCortex CEO John Mucci." Yea, my sister's p4-HT 3GHZ laptop CPU only takes 88 watts max. I guess they meant the average power consumption of the whole laptop, averaging across all models on the market? Well, obviously the statement in the article is bogus. http://processorfinder.intel.com/details.aspx?sSpec=SL7DT [intel.com]
    • What I find impressive is the fact that apparently, the average power output of each cyclist was at least 120 W. I remember seeing someone use a bike generator to make a 40 W bulb dimly flicker when I was a kid. Either generators have gotten a heck of a lot more efficient or these people are serious athletes.
      • Re: (Score:3, Informative)

        What I find impressive is the fact that apparently, the average power output of each cyclist was at least 120 W

        A quick google suggests that 400 W is about right for a rider in good condition for a short time.

      • by SlashV ( 1069110 )
        In fact 120W is rather poor. A reasonably fit person will easily produce 200W, while professional cyclists will output 500W or more...
      • by zcat_NZ ( 267672 )
        My 10 and 12yo daughters managed to brightly light and then burn out a 12V 100W halogen bulb entirely by pedal power. The older one was peddling, while the younger daughter held an old DC motor against the wheel to act as a generator.

        I was quite impressed.

    • How can you claim something is bogus by stating one single example? You haven't even taken an average of a few chips like you seem to presume they have!
  • "A spokesperson said that the human-powered session produced more computations than took place in the first 3,000 years of civilization." Except that they didn't pedal enough cover the development and manufacturing costs.
    • interesting ? Rubbish! Just because a spokesperson says it doesn't make it true. Ask Scot McLellan :)
      The first 3,000 years of civilization ?? When hardly anybody was literate and counting was like 1,2, many ?
      Come off it, of all the nonsense comparisons this one really is far out. Is a year of ancient civilization now a performance benchmark ?
  • by newsdee ( 629448 ) on Thursday December 20, 2007 @04:04AM (#21762444) Homepage Journal
    Oh man, I had written the below ages ago, and now there's a relevant story on /. !
    (yes I could claim prior art but I'm not SCO ;-D )

    !!! welcome to the ENIAC democompo!!!

    We are happy to announce the opening of the first demoparty dedicated to the ENIAC.
    If you wish to participate, please agree to observe the following party rules:

    1) Bring your own ENIACs. We do not want to see viruses on our system.

    2) If you plug your ENIAC to the wall power plug, the PC, Amiga and Atari ST demomakers will saw your head off. At your own risk.

    3) To give electrical power to your machine, we provide a bicycle room with attached generators. Please let us know a week in advance of the size of your group, so that we can get enough bicycles for everyone.

    4) You are responsible of finding your own spot in the party room. Our ENIAC is already taking half of the available space.

    5) It is forbidden to step on the wires.

    6) "Flame" demo effects are forbidden. When we tried to achieve one, the vacuum tubes caught fire and we had to call 911. They were not very happy about it.

    7) It is forbidden to spray paint graffiti on our ENIAC.

    8) Domestic animals are forbidden. We will not pay for any damage caused by the stench of burned fur coming out of relay boxes.

    9) You are responsible looking after your ENIAC. Dishonest persons may want to steal it at night.

    10) It is STRICTLY forbidden to sleep on top of the ENIAC units.

    11) Bring your own spare vacuum tubes and resistors. If you forget them you can buy them at the party but we will set the price... don't say we didn't tell you...

    12) Musical creations are forbidden. Our musician tried to compose something and provoked the death of five dogs while trying to complete "Woof Woof ZAPPP !!", played with his newly created Music Tracker "LiveWireDogeeh".

    13) Graphical creations are forbidden. Our graphist found a horrible death after making a vacuum tube box explode in an attempt to automatically create a drawing of Pamela Anderson on the floor with the glass shards. The result was not so great anyways.

    14) The Bicycle Room has an excellent drink vending machine [rubs hands].

    15) The coders are not allowed to access the ENIAC switches while the demo is running.

    16) Any vacuum tube that fries during the demo cannot be replaced.

    17) The "Plasma", "Shadebobs" or "Lens" demo effects are forbidden. Our coder placed some pot in the relay box so that we were stoned by the smoke and saw all kind of weird stuff.

    18) If somebody does not respect these rules, people may be pissed off and quit the ENIAC scene !

    The competition prices are as follows:

    1. A brand new ENIAC
    2. A Z80 building kit for every member of the group.
    3. A box of General Electric vacuum tubes.

    Good luck !
  • maybe the students could save more energy by spending that time on optimizing their code....
    • Absolutely, it is indeed frightening to see students from the top american technology college show pride in doing grunt physical work instead of using their brain, and even worse to see that on /. instead of Sport Illustrated.
  • ...I really like the idea of low-powered computing. In the last week we already saw a SSD 'disk' with SATA interface presented here that only uses .3Watts. We may not call our PDA's supercomputers these days, but one day in the past, the Big Iron systems used as supercomputing stations were slower than our PDA's can emulate them.
    • Eh...

      I've done some research into it, and I figure you're probably better off buying one of the 12W foldable solar panels (or the rollable ones, but they're even more expensive, and give even worse W/area), and tying it down over your Panniers than to use some kind of generator system on a bike. On the upside, depending on the rest of system, next years ultra mobile parts based on Intel's Silverthorne ought to be pretty awesome for this kind of thing, and hopefully offer performance in the range of 500Mh

  • to solve two prominent US problems: Too much CO2 output due to excessive energy consumption and a lot of overweight people. Solution: have them power something by pedaling for a little longer than 20 minutes, though.
  • New record? (Score:5, Interesting)

    by Titoxd ( 1116095 ) on Thursday December 20, 2007 @04:26AM (#21762520) Homepage
    OK, so they broke the record for human-powered computing. But what was the previous record? Was there even such a thing as a previous record? How is this new record actually measured? You know that more people will now try to break the MIT mark, and TFA is rather scant on details...
    • by DMoylan ( 65079 )
      what was the most difficult sum calculated on an abacus?

      weren't the original adding machines hand cranked?

      at least a starting point.
  • by ishmaelflood ( 643277 ) on Thursday December 20, 2007 @04:28AM (#21762528)
    Well fuck me dead, somebody has figured out how to convert mechanical work into electrical energy. Trust those whacky kids at MIT to pull it off.

    • Moderate parent up, please. This is not news for nerds. And it doesn't matter the least.
    • "Well fuck me dead"...given the rest of your post, this brings up unpleasant images of your orifices being outfitted with devices to turn friction into electricity.
    • Now, that is funny! Thanks for the morning laugh, it is the best way to start off the day.

      Next thing, maybe they will create a device to store some of the *mechanical work* to be used later to convert into electrical energy.

      InnerWeb

  • And they thought finals time was exhausting before.
  • a typical notebook computer CPU needing 100 watts, according to SiCortex CEO John Mucci

    My laptop draws about 20Watts max when running on battery power, even with the CPU running full-throttle. Much of those watts is for the backlight and drive. A 'typical laptop CPU', an Intel Pentium Mobile, uses maximal 24W, according to the datasheet [intel.com] provided by Intel. Saying a laptop would typically use 100W is a bit overstated.

  • This will become a requirement for grad students now?
  • by sapphire wyvern ( 1153271 ) on Thursday December 20, 2007 @05:05AM (#21762676)
    Ahem. Sorry...
  • This is amazing! Just imagine what they could accomplish if they tried to do something useful, innovative, or even - technologically challenging.

    • This is amazing! Just imagine what they could accomplish if they tried to do something useful, innovative, or even - technologically challenging.
      Yes, there is nothing challenging or useful in getting 648 CPUs and 1TB of memory to work with a power draw of 1200W in the same way that you're not a gigantic fucking faggot.
  • A team of ten MIT students powered a supercomputer [...] The system is low-powered and draws 1,200 watts without needing special power supplies or cooling.

    Indeed, what's so special about a power-supply consisting of 10 MIT students? Nothing really...

  • Now that's the way to do it. Not just the environmentally friendly power generation, but the performance per Watt of the computer, too. According to an article at The Register [theregister.co.uk], the SC648 is built from MIPS (the type of CPU) cores that run at 500 MHz and execute two intructions per cycle. That should work out to about 1000 MIPS (the performance unit) per core, which, according to el Reg, the SC648 has 2916 of. In other words, these students got 2.9 million MIPS for 1200 Watts. That's some 2400 MIPS per Watt!
    • Apparently [behardware.com] my Core 2 Duo can, in ideal circumstances, run 2 + 8 (two 128-bit SSE floating point ops equal 8 single precision 32-bit ops) = 10 instructions per clock (I think). 2 Cores at 1.6 GHz would then process 32000 MIPS and use up about 10 watts (in a Lenovo X61t with some power-saving settings (e.g. no Bluetooth or WiFi, low display brightness) fully usable. If I didn't mess up something, that'd be 3200 MIPS/W.
      During actual processor load and WiFi, usage of the chipset's 3D capabilities and high disp
  • Here you find a good overview of the possibilities of wind up power and bicycle machines [lowtechmagazine.com]
  • by hyades1 ( 1149581 ) <hyades1@hotmail.com> on Thursday December 20, 2007 @06:05AM (#21762934)

    Please allow me to offer the Geekaflop, which will be defined as the number of gigaflops per bag of Chitos, as determined by the weight loss of 10 MIT students pedaling flat-out for 10 minutes to keep the supercomputer powered up.

    The new term would be abbreviated "HaM"(Hamster Megacycles), thereby incorporating the longstanding scientific truism "We stand on the shoulders of giants".

    • I would hardly call it flat out. The 1200 watt consumption divided by 10 people is 120 watts per person, which is only a light to moderate amount of power to produce while bicycling.
  • Next week it will be: Hamster Powers (Super)notebook with Hamster Wheel
  • This is a particularly satisfying story to me. When I was writing my pulp scifi novella The Bikes of New York [cheeseburgerbrown.com] (in which the poor pedal generator bicycles for spare change) I was told by many snotty self-proclaimed debunkers that human beings could never generate a meaningful amount of power using their bodies, and some of them had all sorts of intimidating mathematics to prove their points.

    This story seems to show that their rigour was limp, and their points pointless.

    Hooray for a legitimate basis fo
    • Feed the grid with your ElectroExerBike! Now you too can reduce your waistline AND your electricity bill by up to $3.50 per month with only 6 hours of work per day!

      Yours for only $299 in 3 easy monthly payments.
  • Well congrats, Tony and Wilson and Peter and all my other SiCortex buddies I hope you sell a ton of them. Wilson recently gave an interesting talk about verifying the SiCortex system and ASICs http://www.veripool.com/papers.html [veripool.com]. He's also a huge open source contributor.

    /Ed - not affiliated with SiCortex
  • As an OLPC? (Score:2, Funny)

    by qrwe ( 625937 )
    Maybe this would be a good way to powersupply future OLPC:s?
  • Then this exercise should be (HPC)^2, right?
  • I once took part in a nutrition study that required each participant to ride a stationary bicycle for 20 min or so. I asked the lead researcher how much power a typical college student could produce. He said a healthy college student could produce about 180W for the duration of a test session.

    The test itself was uncomfortable. There was no breeze to keep you cool (and, since the human body is only about 25% efficient, that meant you were dissipating > 500 W as heat), and you had to wear a mouthpiece

  • The system is low-powered and draws 1,200 watts without needing special power supplies or cooling..."

    The students' special power supply and cooling, OTOH, came from beer! (Free, of course!)

  • > They powered a SiCortex SC648 supercomputer with a Linux cluster of 648 CPUs...
  • For those that aren't sure how watts measure up on a bicycle, use this calculator to figure out what equivalent speed that is. http://www.kreuzotter.de/english/espeed.htm [kreuzotter.de]
  • Effeciency? Bah! (Score:3, Insightful)

    by Dan East ( 318230 ) on Thursday December 20, 2007 @11:31AM (#21765814) Journal
    The brain consumes somewhere around 20 watts, and the estimated processing power of the human brain is somewhere between 10 and 100000 teraflops, with a storage capacity somewhere in the terabyte range. Now, we enjoy that processing power while completely at rest, without having to exert ourselves at all. Granted, our architecture isn't suited for some of the tasks a supercomputer is put to, on the other hand there are many incredibly rudimentary thinking tasks that the computer cannot perform no matter how powerful it is.

    Dan East
  • Just due to its shear number of server farms and servers - about 3 million nodes recent estimate. On the other hand they may hold the contrarian position as lowest power consumption per peta-op/peta-byte due to their attention to decreasing power costs (and green concern).
  • I've always wondered why more gyms don't by equipment that gathers the excess electricity and sell it back to the power company. I bet the average Golds Gym probably puts out an incredible amount of power from all those machines.
  • IMO the real story is that they can run a 648-processor cluster on 1200 W.
  • To withstand a /.'ing?

    Only one way to find out....

Bus error -- please leave by the rear door.

Working...