Japan's Petaflop Supercomputer

slashthedot writes "Japan has built the fastest supercomputer in the world. While the BlueGene/L contains 130,000 processors, Japan has managed to create the first Petaflop supercomputer, called MDGrape-3, with just 4808 chips, and it cost just $9 million to develop."

Comments

machines like this

[Neuropol]

should be used in conjunction with the topic from the previous article. Creating coutless means by which, to not only find vulnerabilities in things like Javascript, but equally, construct fixes to those vulnerabilities. Once it creates an open door, it generates the fix for closing it and keeping it closed. Machines like this can think thousands of times faster than your average black-hat-crackah, so why not use them as a fight fire with fire tool?

Every one is so concerned with internet safety, on would think that at some point massive resources with be set forth in order to effectively deal with the flaw finding few out there making it difficult for the rest of to simply enjoy the benefits of the internet.

Efficiency

[Eightyford]

The article says that this machine is much more efficient than other supercomputers. Is it actually cheaper to run large programs like SETI@HOME on a supercomputer? Electricity isn't cheap.

Re:Efficiency

[Jerry Coffin]

Is it actually cheaper to run large programs like SETI@HOME on a supercomputer?

This computer is efficient at what it does largely because it's extremely specialized. It's built specifically for working on molecular dynamics, but from the looks of things, it's probably close to useless for nearly anything else.

As such, it would probably work quite nicely for Stanford's folding@home project (which studies protein folding, i.e. molecular dynamics). It probably would not work very well for seti@home, because SETI isn't studying molecular dynamics, and it would probably be difficult to cast the problems they're working on into a form that would "look" enough like molecular dynamics to work well on this machine (this, BTW, is why this machine probably shouldn't go onto the top500 list or anything like that -- it's really not a general purpose computer at all).

As far as using other supercomputers for these kinds of jobs, here's what the folding@home FAQ has to say about it (from the F@H FAQ):

Why not just use a supercomputer? Modern supercomputers are essentially clusters of hundreds of processors linked by fast networking. The speed of these processors is comparable to (and often slower than) those found in PCs! Thus, if an algorithm (like ours) does not need the fast networking, it will run just as fast on a supercluster as a supercomputer. However, our application needs not the hundreds of processors found in modern supercomputers, but hundreds of thousands of processors. Hence, the calculations performed on Folding@Home would not be possible by any other means! Moreover, even if we were given exclusive access to all of the supercomputers in the world, we would still have fewer cycles than we do with the Folding@Home cluster! This is possible since PC processors are now very fast and there are hundreds of millions of PCs sitting idle in the world.

To put that into perspective, consider that the Blue Gene/L has 65536 processors. seti@home has over a million hosts and folding@home has a couple hundred thousand more. As the quote above notes, most supercomputers aren't drastically faster on a per-processor basis than PCs -- not nearly enough to make up this deficiency in sheer number of processors.

My guess is that the Blue Gene/L is probably somewhat more power efficient than the average contributor to seti@home or folding@home -- but mostly because the majority of the latter are probably Pentium 4's, which are notoriously inefficient in terms of power usage. As the world transitions away from the Netbust architecture, it's nearly certain that the efficiency of seti@home, folding@home, etc., will go up (considerably).

That brings up another point worth considering: the way things are right now, the computers used for seti@home, folding@home, BOINC, etc., get updated on quite a regular basis. If they spent millions of dollars for a single fast machine, it would might be more efficient right now -- but in a few years it would fall behind the curve -- but most budget committees (and such) would be reluctant to spend millions of dollars to replace it simply because something better was available.

Uses a large walk-in closet?

[StarWreck]

If this petaflop supercomputer really only costs $9 million and only occupies the space of a large walk-in closet, why don't they mass-produce it and sell it. No, not to individuals but to corporations and governments. Folding@Home and Seti@Home could suddenly be like, sorry guys we don't need you anymore - we got something better. Having hundreds of copies of this super computer could quickly solve problems across the globe that much slower supercomputers are currently having trouble with!

Re:1,500 $

[smallfries]

No? 'cos a GTX-7800 does 320Gflop/s and you could buy a few of those for $1500...

Re:Say what?!?

[bloosqr]

Yea its specialized hardware, the mdgrape basically calculates Newton's law in the hardware so it does the inverse ^2 calculation really super fast. There used to be a md-grape equivalent which did the same thing for coulombs law (as you would think there is more money in doing biosims than astrosims), but i think that died as the market was too small.

I think this was an ibm/fujitsu collaboration and ibm had md-grape and dropped it because of the market and fujitsu is still making the grape..

FYI the reason even though it is specialized, this is cool is that any simulation you want to do classically (i.e gravity, coulomb), basically goes as N^2 where N is the number of things (i.e. you have to calculation the interaction btwn each thing and every other thing, so there are lots of tricks to make approximations (clever versions of stuff far away doesn't matter so much). This goes up fast as simulations get bigger, hence the GRAPE tricks, which let people do monster simulations as if they had terahertz machines!

(On the other hand some people will object the "approximations" make real simulations go as N log N, so its not like we were all twiddling our thumbs waiting around for GRAPE)