Forgot your password?
typodupeerror

Junk Super Computer Assimilates All 182

Posted by samzenpus
from the one-mans-trash-is-another-mans-computer dept.
VonGuard writes "The ACCRC is the relatively famous computer recycling non-profit in Berkeley that builds clusters out of old hardware. Make Blog has an article about the Center's plans to build a cluster out of the equipment people bring to recycle at Make Faire later this month. The ACCRC geeks are now able to integrate PII's or better into the cluster, which will be powered by Vegetable Oil and run Parallel Knoppix."
This discussion has been archived. No new comments can be posted.

Junk Super Computer Assimilates All

Comments Filter:
  • by Anonymous Coward
    Soylent Cluster is made from equipment people!

    Soon they'll be breeding us like cattle!
  • veggie oil? (Score:2, Insightful)

    by Evoluder (669436)
    is it really cheaper than plain old power company? maybe it scales cheaper?
    • I doubt it. If you look at a typical utility bill you're talking pennies a kilowatt hour. And small systems don't scale as well as larger systems that can deliver power to an entire city. The only inefficiency is how the power plants burn coal to heat the boilers.
      • Re:veggie oil? (Score:5, Insightful)

        by blincoln (592401) on Friday April 07, 2006 @08:29PM (#15088908) Homepage Journal
        If you look at a typical utility bill you're talking pennies a kilowatt hour.

        I think their idea is to counteract the concept that for the same amount of power, they could be running much more powerful hardware. If the electricity comes from coal, they're wasting energy, but if it comes from biodiesel they're... uh... wasting energy in a way that sounds good to hippies?
        • I think their idea is to counteract the concept that for the same amount of power, they could be running much more powerful hardware. If the electricity comes from coal, they're wasting energy, but if it comes from biodiesel they're... uh... wasting energy in a way that sounds good to hippies?

          I'm a hippie. I don't know too much about biodiesel except from what I've heard and read. I'm not that terribly interested in it, but lets look at the differences in this wasted energy. Coal is pressed plants that t
          • Hydro electric power is pretty cool too. Wind is not as cool as it seems. Very variable and it disrupts weather patterns.

            Wind disrupts weather patterns? Unless you're talking about curbing global warming, I've never seen a shred of evidence that wind turbines affect the weather. As for hydroelectric power, it's one of the environmentally worst types of renewable energy. Building dams destroys valuable farmland, floods priceless natural habitat, and screws up fish migration.

            Hippie nonsense aside, I think t

      • Curious, I thought with an efficient enough fuel cell (run off natural gas) you could generate power in your home, avoid the inefficiency of losing power during the power line travel... (if I remember my physics, correctly, the power lost during transmission is proportional to i^2 where i is the current). I thought the process of transforming the energy to and from that state was fairly inefficient (but better then sending it down the power line without doing it. It's been a long time. I thought it was
        • if I remember my physics, correctly, the power lost during transmission is proportional to i^2 where i is the current

          That's why they use transformers to lower the current and increase the voltage.

        • Re:veggie oil? (Score:4, Insightful)

          by Eivind (15695) <eivindorama@gmail.com> on Saturday April 08, 2006 @06:22AM (#15090195) Homepage
          Curious, I thought with an efficient enough fuel cell (run off natural gas) you could generate power in your home, avoid the inefficiency of losing power during the power line travel...

          Yes. But you'd add the inefficiency of having to transport the natural gas somehow. Which also costs energy. No, natural gas floating in pipes is *not* obviously that much more energy-efficient than electrons floating in wires, it depends on the details. (one thick pipe offers less friction than many small ones, higher voltage power-lines give lower losses)

          if I remember my physics, correctly, the power lost during transmission is proportional to i^2 where i is the current.

          Yeah. In absolute terms. But offcourse in this case your power transmitted is higher too, so your losses, measured as a percentage, doesn't go up that rapidly.

          The oposite is also true though: If you up the voltage, then you can scale back the current needed for a certain power by the same amount, which leads to lower losses. Multiply your voltage by 10, and you can divide current by 10, and still transmit the same power. But at 1/10th the current, this means, by your formula, that the losses are now only 1/100th of what they where.

          I thought the process of transforming the energy to and from that state was fairly inefficient (but better then sending it down the power line without doing it. It's been a long time.

          Where'd you get that idea ? Large transformers achieve efficiencies in the 99.75% range, and even the small ugly wall-wart transformers that are mass-made at a buck a piece from the cheapest possible materials frequently manage to come in at 95%

          It's like the fact that modern day farms are actually far less efficient then ones from 100 years ago, from an energy perspective.

          Yes. But only from that perspective, which isn't the one we're trying to optimise for. Our current economical system optimises for production-efficiency. And a single person working on a farm produces probably 100 times more than a single person working on a farm did 100 years ago.

          Energy isn't lacking. Not even *clean* energy is lacking, there's plenty of it to go around. The only reason it's not dominant is that currently non-clean energy is cheaper. It's perfectly possible to make clean energy enough to supply current and forseeable needs. But the thing is, with current tech it costs more. I don't know the numbers for US, but for example in Norway wind-power costs double of normal power (which is hydro with us, so also clean, but let's ignore that). In Germany there's a minimum prize given for home-produced energy of $0,50 or so, which is more than enough to make it a paying proposition (i.e. you make a *profit* by installing solar-cells on your roof), but which also happens to be like 4 times the price of conventional power.

          A farm using only clean energy would still be a hell of a lot more efficient than the ones 100 years ago. But thing is: it'd be *less* (financially) efficient than the farms that burn oil. So that's what's happening.

          But the scale is slowly tipping. The price of oil and gas has raised a lot, and ist likely to raise a lot more. The price of solar, wind, hydro, thermal and so on has all been falling steadily, and will continue to do so.

    • That's the beauty of the vegetable oil: it's free. In fact, restaurants pay to have the stuff hauled away.
    • Make isn't about doing things easy. ;)

      It's cool to cluster computers together, it's cooler to make that cluster out of the odds-and-ends of PC 'junk', and it's even cooler to run it off of an alternative power source. :)

  • Make Faire = Do Do

    Bit redundant, do do you think?
  • by stevetures (656643) <stevetures@gmail . c om> on Friday April 07, 2006 @08:20PM (#15088874) Homepage Journal
    Hmm... they tried this piecemeal supercomputer at my university (university of san francisco). From what I understood, they accepted a lot of low-spec computers that actually caused more problems than they served to compute. http://www.flashmobcomputing.org/ [flashmobcomputing.org] Can anyone confirm on my specific point?
    • by Darth_brooks (180756) <clipper377@@@gmail...com> on Friday April 07, 2006 @08:34PM (#15088933) Homepage
      The biggest problem you can run into with older machines is the lack of support for things that make building a cluster easy. The two biggest things I ran into were:

      -Wake On LAN support is huge, so that rules out the old machines with the clicky power switch.
      -Trying to do anything with less than 32 megs of RAM is a PITA.

      On top of that older machines don't always have a NIC, so you're stuck scrounging for parts. Plus who knows what kind of cryptic Acer-Packard Bell-eMachines crap hardware you're going to get via donations (so building a boot image can be a pain), so you're opening youself up to tracking down odd bits of unsupported yet essential hardware drivers (PCI Controllers stand out.)

      Clustering gets way easier when you can stick to at least the same general system brand (e.g. Dell) or even better, identical systems.
      • It would seem from the ParallelKnoppix that the preferred way to build a cluster is by network booting them. It's possible to boot each node from a Live CD but the author of the tutorial at least doesn't seem very enthused about the idea.

        So it would seem that if you were going to troll for donations, your minimum spec would be something that either had a bootable NIC in it, or was capable of accepting one that you'd be able to acquire easily (i.e., has PCI slots, no old ISA garbage).

        If it was me trying to b

      • Having an almost completely heterogeneous computer system is not something I would want to use or admin. I know a little about this stuff...

    • absolutely. we had a bunch of old P3s lying around with about 256mb each. i spent some time putting them into a cluster because the departmental servers were overloaded. turned out that it just wasn't useful. the overloaded dual-opteron server with 4 gigs of ram ran a set of intensive experiments faster than the p3s could crank out their stuff. from my gross, empirical estimates, 5 p3s w/256mb each == 1 overloaded dual-opteron server at fairly I/O intensive tasks. Don't even want to think about power/heat i
      • 5 p3s w/256mb each == 1 overloaded dual-opteron server at fairly I/O intensive tasks. Don't even want to think about power/heat issues.

        Not to mention 5x+ the likelihood of a hardware failure because there are 5 of every part that can and will fail and that the machines are already old. Plus ease of use, and admin maintenance.

  • by neoshroom (324937) on Friday April 07, 2006 @08:23PM (#15088885)
    So, they have a network of recycled computers...
    ...being run by a generator using veggie oil...
    ...to render 3D images.

    So the only question remaining is: What are they rendering?

    My guess: PETA [peta.org]'s new 3D logo.

    __

    Custom Research Paper [elephantessays.com]
  • by Anonymous Coward
    ...was that they want money to take your stuff. If you don't mind a drive, Video Only will take monitors (and TV's) for FREE! And displays are the hardest things to get rid of.

    As for PC's, there is a charity in Berkeley that takes donated PC's, refurbs. them and gives them to families that can't afford a computer.

    The City of Albany (next door to Berkeley) had a day where you could take almost any kind of electronic device and dispose of it for free (no large applicances). They plan on doing this yearly duri
    • I hate to parrot the ACCRC website so blatantly, but if you aren't paying someone to take it, it's probably being shipped to China, stripped of its precious metals, and then dumped in a river.

      All this electronic equipment that we love so much is made up of some very, very unpleasant substances. Dumping it in a landfill or shipping it to a country that doesn't care about its environment doesn't count as proper disposal.

      Working equipment is fine to be donated to charities who can use or redistribute it, but
      • If it costs more to recycle something than to store in a safe location where it will not pollute, then it is not worth recycling. In the end, you're expending more energy to recycle it than you are taking in. The ACCRC's passing it on to the donater (no good deed goes unpunished) is a cheap way of hiding that cost. Obviously, we don't want lead dumped in Chinese rivers ("we" obviously doesn't include the Chinese enslavers/bureaucrats) so if you can't have it recycled efficiently, then you might as well h
    • First of all, monitors are free now at the ACCRC, as long as you are a California resident.

      This IS the place that fixes PC's and gives them to charities, non-profits, and underprivlidged individuals.

      The ACCRC was the recycler that took all of the equipment for free at the Solano Street fair, the day in Albany to which this post refers. The Solano Street fair will probably host the ACCRC for free recycling again next this year. No one from Albany has called to set this up yet, however.

      Oh, and the AC
  • What a colossal... (Score:5, Informative)

    by Harry Balls (799916) * on Friday April 07, 2006 @08:29PM (#15088911)
    ...waste of electricity.
    State-of-the-art computers are probably about 15 times as fast as Pentium II-based computers, and consume maybe twice as much electricity.
    Or take Pentium M-based computers, they consume less electricity than Pentium II-based computers and are probably about 10 times as fast.

    Just my 2 cents.

    • Funny, New systems are coming with 350 and 400 watt Power supplies, but my old PII has a 250. Yep, curse those old inefficient machines!

      It really depends on what they're doing with the machines. If they're running it as a diskless thin cluster then, yeah, newer chips draw less power. But each node is probably drawing less than a hundred watts (max) in the first place.
    • by Jeff DeMaagd (2015)
      I've found that PII/PIII based computers can be pretty efficient. I have a PIII 500MHz computer that measured to consume 40W total at the wall jack, so that's counting PSU inefficiency too. You are right that some of the latest computers can produce far more FLOPS per watt, but electricity is pretty cheap compared to the cost of a new computer. That 40W computer might consume $40 worth of electricity per year, it's hard to argue that spending money on new computers would be better from an electricity cos
      • Your $40 is about right. So, lets say I can get a Pentium M at 10x the CPU power and the same wattage. You'd spend $1200 over the course of three years (reasonable lifetime of a PC) on electricity. I'd spend $120. Plus, you have 10 old PCs that are going to break down and cause you problems. I have one new one that probably won't.

        Running old computers as supercomputers makes no sense. I do have one, though, that I run as a home network server. They are great for that.
        • WOW, Where do you get these 10 GHz (or since the gp said 500, 5 GHz) Pentium Ms? Yes, I know its not all about hertz, but It is not even possible, much less believable that a Pentium M is 10x more powerful than a PIII 1.13Ghz coppermine or even a lowly PIII 500Mhz(unless you mean for playing games, which is mostly due to the older video cards in those machines). For reference you may refer to this chart.

          65 CPUs from 100 MHz to 3066 MHz [tomshardware.com]
    • by tux_deamon (663650) on Friday April 07, 2006 @09:48PM (#15089182)
      I think the point is is that the equipment being used is old discarded hardware. New fast computers cost money. This organization is reusing this old hardware and keeping it out of a landfill.
      • by Elastri (911062)
        Electricity costs money too, and it keeps costing money even after the initial capital purchase (or lack-there-of). If you're building a cluster powerful enough to be comparable with even a couple low-mid end modern PCs you're not going to be running for too long before you hit the point where the power savings makes the modern PCs less expensive. A more thorough explanation can be found in these two posts from a beowulf mailing list:

        http://www.beowulf.org/archive/2003-March/009658.h tml [beowulf.org]
        http://www.beowulf [beowulf.org]
    • Yes.. because, after all, vegetable oil [flickr.com] is a horribly inappropriate solution for their recycled cluster. Sigh.

      -DrkShadow
    • Dr. Emmett Brown: Marty, I'm sorry. But the only power source capable of generating 1.21 gigawatts of electricity is a bolt of lightning.

      Marty McFly: [startled] What did you say?

      Dr. Emmett Brown: A bolt of lighting. Unfortunately, you never know when or where it's ever gonna strike.

      Marty McFly: Hmmmm... What about vegetable oil?

      Dr. Emmett Brown: Well of course, vegetable oil. But where are we going to get vegetable oil in 1955?
    • This was along the lines of my thought.. using old computers that consume tons of power to compute very little.. but showing your ecological awareness by using vegetable oil? Give me a few nuclear powered quad core opterons anyday. I'll show you the definition of progress. Lower power consumption, cleaner energy, and multiples high in productivity and computational gains.
    • by SuperBanana (662181) on Friday April 07, 2006 @11:52PM (#15089471)

      Just for comparison's sake, I borrowed some Kill-A-Watt meters and measured my gear.

      • Powerbook 1ghz 17": Idling, about 15w, full-bore 29W or so
      • Dual P3-850 2U rackmount: 150W
      • Dual 2Ghz G5 Xserve rackmount: 175W
      • Mac Mini 1.25ghz: Idling, 11W, full bore, about ~25W
      • Celeron (original celeron) 450mhz PC. No hard drives, one CPU fan, one PSU fan, floppy, 2 net cards, CDROM drive. 65W, idling (running a bsd-based firewall.)

      The shocker was how low the Mini's power consumption was, and how high the celeron router. Also, the Xserve, Mini, and Dual P3 all had power factors of .99, whereas the celeron had a power factor of about .6...ie, not power-factor corrected.

      Oh, and switchgear? Varied from 1W (yes, ONE watt!) to ELEVEN for an old 100BaseT switch. The lowest power consumers were newer hubs, second by a pair of gigabit switches I bought within the last year that were about 5-7W.

      • whereas the celeron had a power factor of about .6...ie, not power-factor corrected.
        That's nothing to do with the CPU or the PSU age, and everything to do with how cheaply designed the PSU was. There are tons of brand new PSUs on the market with absolutely crap power factors.
    • Yes, but that's the point. The one time cost of new hardware will eventually pay them back because they get more computations per amount of electricity.....
    • State-of-the-art computers are probably about 15 times as fast as Pentium II-based computers, and consume maybe twice as much electricity.

      Modern computers being 15 times faster than P-IIs?
      I seriously doubt it. Take the slowest P-II ever, the P-II/266, as reference. 15x that means, what, a P4 at 4GHz?
      Sorry, no way it is that faster. Maybe a P3 core at 4GHz, assuming overall improvements besides the clock speed.

      Personally, I do think that a P3/866-1GHz is worth for such tasks.
      You may say about being u
    • You know, you're absolutely right. And I'm sure if you contact them, they'll be happy to tell you who to make your check out to so that they can buy these new "state of the art" computers. I'm afraid it'll cost a bit more than your 2 cents, however... as well as a lot more than they're currently paying for this hardware, which... while not ideal... is FREE.

      • Except they still are paying for the electricity. Which was the original point.
        • They're only paying for the few chemicals needed to make biodiesel to fuel the generator that powers the cluster.

          -Jesse
          • The electricity they generate could be sold back to the grid at a profit, instead they are burning it running computers. Whether they make their own electricity or not, the value of the electricity is still something like 12c/kWh.

            To make this clearer, lets say you discovered 100t of gold on your property. Maybe you make your house out of gold, maybe you make it out of bricks. If you sold the gold you could buy a lot of bricks. Which would you do?
    • Unloved fry grease meets unloved computers to make parallel super(?)computing cluster. How is this bad in any way?
  • by Anonymous Coward on Friday April 07, 2006 @08:31PM (#15088918)
    ... the old Stone Soup Supercomputer [ornl.gov] was the first I can remember that used cast-off computers to generate (what passed for) Serious Horsepower. Tempus fugit, indeed.
  • Worth it? (Score:4, Insightful)

    by Galahad2 (517736) on Friday April 07, 2006 @08:42PM (#15088967) Homepage
    I wonder if the power consumption of a low-end Pentium 2 is actually worth the computation capability it could contribute to a network. There's definitely a point at which it costs more to run a computer than you can get out of it -- where does that line fall?
    • Re:Worth it? (Score:3, Insightful)

      by bear_phillips (165929)
      A good point, but a better question is which is more power efficient: 1) running 4 old dontated computers or 2) building and manufacturing a new computer.
    • Doesn't matter, in this situation. The power they're using is free, if they really are going to use a vegetable-oil generator. (Assumedly they are getting the vegetable oil donated, so the cost of the energy in is zero.)

      I suppose you could argue that this still contributes to global warming or carbon desequestration or something, but that's taking things a bit far. Plus the carbon in the vegetable oil is eventually going to be released anyway.

      If you're getting the energy from a source like that, which is bo
      • There is still a line in a heterogeneous cluster. There comes a point where the overhead cost needed to negotiate a job to the lowest end machine for processing surpasses the cost of a higher end machine doing the job itself. At such a point, adding lower end machines actually slows the cluster down.
    • There's definitely a point at which it costs more to run a computer than you can get out of it -- where does that line fall?

      I would guess just barely this side of solar-powered calculators...

      energy is cheap. computation is not. It's that simple - look at the laws of thermodynamics... the universe hates free energy, that much is clear, but even more than free energy it hates organization, information, computation.

      Making computation a much scarcer resource than free energy...
  • Hey, I know this place. Drive by there all the time. I always thought it was something run by the county. Never realized they were an independent non-profit. Based in a funky looking building right by the freeway.
  • Yes, but... (Score:4, Funny)

    by zpeterz63 (851922) on Friday April 07, 2006 @08:46PM (#15088978)
    will it run Windows?
  • Before everyone gets all "top500" over this idea, here is little cold water on the realities of cluster computing [clustermonkey.net]. Yes large number sof CPUs are great, but the software has to be there to use them. Plus, for anything other than rendering, a heterogeneous cluster is limited by its slowest node (for almost all applications)

    It is the cluster equivalent of how many people you can stuff into phone booth. If you have nothing better to do, well why not I guess.

  • ... powered by Vegetable Oil and run Parallel ...

    I remembered when they used to be powered by vacuum tubes and ran one instruction at a time.
    • Vacuum tubes? Hmph! In my day, we had to pass instructions by shouting them one character at a time into the ear-trumpet of a half-deaf, illiterate programmer. He would then flip the bits by wetting his fingers and toes, then shorting out the leads because we didn't have the kind of money to buy niceties like wires and switches like you young dandies nowadays. Brats!
  • What's the optimal CPU for MFLOPS per watt these days? At some point, an old system is just too slow to justify the power bill, when a cheap new system delivers much more performance at minimal cost.
    • Probably Power5. However, to use desktop processors as the example: I have some benchmarks of a code I've run on machines from a DEC Alpha 21064 through a current G5 2.0.. The timings that matter here are jobs that took 107 seconds on a PII/400 now take 7. The speedup is about constant across the series, so we're looking at a 15x speed-up on my current code versus a PII/400. Since this particular package runs at about the same speed on comparable clock-rate Opterons/G5s, we can infer a probable 15x spe
  • is a big problem for these kinds of projects.

    When computing power grows exponentially, you need an exponential (in the age of the machines) number of old machines to do the work of one new one. And that is even before you consider all the losses to parallelism, the big electricity bill, and all the know-how needed to put them together!

    But it is pretty cool.

    Tor
  • The first borg cube was created. Resistance is futile. F34r the Pentium Pros.
  • by ScrewTivo (458228) on Friday April 07, 2006 @09:22PM (#15089093) Homepage
    It was amazing what they were doing way back then. Before discussing the practicality of this you have to remember that a lot of what they do is teaching people about computers and providing refurbished computers to the poor. So now they get to learn about building super computers, it doesn't make a difference if a new multi-core system can outperform it or not, it's the lesson that is important. And tossing in a bio-diesel generator is precious!

    BTW, he was talking of building a supercomputer way back then. So the group has put some thought into this.

    If this turns out that it actually has some horsepower I can't wait to hear how it is put to use. The guy who started this is way ahead of the curve. Turning garbage into a self powered supercomputer...kewl!!
  • Experimentation (Score:2, Interesting)

    by ComputerJerk (966963)
    Personally, I think experimenting with computers is a good thing. Maybe it's not a feasible solution to a problem, but experimenting generates new ideas. I see a lot of people saying "Oh, give your burnt out computer to the poor.". Well, if you want to be so generous with the poor, give them a computer that's worth having. Personally, I think giving a computer to someone that's too poor to buy it themself is a waste of a good computer. If they can't afford to buy the computer, they probably can't afford to
  • by Xerp (768138)
    With a veggie oil powered generated, they really should have a distro called OilSlix...
  • We are one.

    Screw you, Fatass.

    Heh, screw YOU!
  • I wish them luck... (Score:4, Interesting)

    by teebob21 (947095) on Saturday April 08, 2006 @01:03AM (#15089646) Journal
    I got excited about cluster computing a couple years back. I spent about $600 on parts for a 12-node Pentium II cluster, then spent 3 weeks setting it all up. I then spent another 6 weeks with a comp sci professor trying to reverse-engineer the Folding at Home client to parallelize the data units. (Psst...don't tell Vijay!) Our solution was to use the F@H client as-is, and to network the nodes as additional drives and run a client with a different machine ID on each drive.

    As it turned out, a single 1.1GHz P3 was doing more folding than 12 350MHz P2's working in parallel. I scrapped the cluster and sold the parts on eBay. My electricity bill dropped about $100 a month afterwards. Again, I wish them luck.
  • by tsa (15680)
    The first thing I thought when I read the title was: Hex! I read too much Pratchett...
  • I did this once (Score:2, Interesting)

    by DurendalMac (736637)
    With 16 old Beige G3s, Mac OS 10.2.8, and XGrid PR2. Yeah, it was crap, but I did it just to do it. I got all sixteen running and ran some of the Xgrid scripts, but beyond that, I had no use for the damn thing. I only had 10Mbps hubs anyway. Built the thing for next to nothing, too.
  • What sort of garbage is that?
  • Not worth it at all except for slashdot-article posts. You can buy an Athlon64 cpu + motherboard + 512mb ram for $250 and thats equal in power to dozens of PII machines, which will use up $250 worth additional power in under a year. Consider issues of space, the fact that its way more work and unreliability, and its obvious the pace of development is so fast that using 10 year old machines is never worth it. If you need to run many single-threaded apps, its still better to use one Athlon64 machine... but ev
    • I think the whole point of the 'project' was to show that older hardware still had usefuless to some people, though useless to most. Mostly to (I think) try to convince people to get off their duffs, go out to the garage, grab those two old desktops and drop them off some place that could use them.

      And don't forget Linux people tend to do things simply because they can. However I agree this is not great feat. This was plugging in a bunch of computers to a switch tossing in some CD's and powering up the boxes
  • On a tangent; I got 8 PIIs working together with clusterknoppix - then having spent the day setting it up, realised I had absolutely no use for it -- any ideas?

Power corrupts. And atomic power corrupts atomically.

Working...