Slashdot Mirror
The Amazing Shrinking Supercomputer
mE123 writes "It would seem that IBM is trying to change what we all think of as super computers. Their new Blue Gene family of super computers is meant to be 6 times faster, consume 1/15 of the power and be 1/10 the size of current models. The prototype is already number 73 (with 2 teraflops) on the list of the most powerful super computers and it's only "roughly the size of a 30-inch television". They are hoping to be able to make it up to 360 Teraflops using only 64 racks." We covered this a bit earlier, but without the level of details.
Should the priority be making faster supercomputers (but large) or smaller supercomputers (but the same speed)? This one seems to be a step in both directions, but I wonder if they're sacrificing speed for size (or vice-versa).
My mom wouldn't let me have one because they take up so much space!
Clif
Blogzine.net
Fortress of Insanity
clifgriffin > blog
So, how long will it be before these become commoditised for sme's ?
...
:-)
Something that fits into the space of a 30" TV set (how about dimensions, guys ?) is presumably about half to 1/3 a standard rack in a co-lo. 2 Teraflops of processing power ought to be able to comfortably shift the bottleneck to the bandwidth, even for database-orientated sites
I think people's cost expectations are going to be significantly impacted by the size of this - if it's small, it must be cheap, right ? (wrong, but try telling them...)
Fantastic acheivement, btw, kudos to the man in blue
Simon
Physicists get Hadrons!
Yes, this is amazing--but I predict that in ten more years computers will be twice as fast, ten thousand times larger, and so expensive that only the five richest kings in Europe will be able to aford one.
I'm awaiting a supercomputer affordable by a small business...something top 100 $30-$60k...then i'll be impressed. Otherwise, it makes no difference to me as I will never get to play with one. *sigh*
If you read the press release, they claim that previous 2 teraflop machines fill up entire rooms, with more than a dozen racks. I'm not so sure this is the case: for instance, Apple claims 798 gigaflops to a rack with the Xserve; by my reckoning that works out to needing 2.5 racks to get 2 teraflops. And that's just with dual 1.3 GHz G4 CPUs; I'd imagine there is an upcoming Xserve rev featuring dual 2.0 GHz G5's.
Don't get me wrong, it's still an impressive achievement (especially if it uses as much less power as claimed.)
"There is no night so forlorn, no mood so bleak, that it cannot be infused with pleasure by tender meat..." - R.W. Apple
Take your standard technology curve (aka Moore's Law), take any specification/cost point, then move ahead an arbitrary point in time and wonders of wonders, it costs less and is smaller and does more.
Yes, one day supercomputers will fit into your wristwatch! What's more, they already do! If you use an ancient measure from, say, 50 years ago.
It's very disappointing to see technology always reduced to whizz-bang figures that are in fact meaningless. What about the impact on our society? What about the capability for good and for bad? What do "good" and "bad" mean, anyhow? How do I know I even exist? What does "I" even mean?
Now, that kind of stuff is worth discussing.
OK, go ahead and mod me as a troll now, if you can't think of an intelligent answer.
Ceci n'est pas une signature
In other news, the price of petrol increases.
Cray made the Cray EL series from '94-'97, they were a "deskside" computer. See here for more info.
Trolling is a art,
I'm not a big fan of super computers. I mean, it's kind of cool, but to me, it's just throwing a whole bunch of computers at the problem, more or less.
That being the case, why aren't distributed apps considered as part of the Super Computer list? I mean, SETI@Home has got to be far and away, #1 in terms of computing power. Granted, it's not in 1 integrated piece of hardware, and Berkeley doesn't own all the hardware, but I still think these things ought to be considered, at least to make it more realistic about who actually has the most computing power.
Just my little rant.
Would sales tax on these things be called a "Blue Gene Levy"? Hahahaha. Horrible, I know. ;)
"Yeah, well, Dracula called and he's coming over tonight for you and I said okay."
was already having to figure the propagation delay of signals (traveling at near the speed of light) into their large multirack systems. I can only imagine one of things driving the desire for smaller supercomputers is to speed up the clock by reducing the delay across the physical size of the box.
--Rob
I work on the project.
We're packing 1024 compute nodes (each node having two CPU cores) into a rack. The nodes are small and based on the PowerPC 440, with beefed up floating point. It has to be air cooled - water is a PITA.
The finished machine will still be quite large - 64 racks with miles of cables. And that doesn't count disk drives. There isn't a single disk drive on the thing - the customer provides the filesystem, which will also be another beefy set of machines. It requires a new building.
The machine featured in the article is just half a rack. It is still respectable, coming in at #73 on top500.org. Might be quite useful for business and small scale scientific in it's current form. (This is far more than my alma matter had access too.)
That's one hell of a p0rn server. Now compile apache, and connect to a pair of DS-3s...
hmm... how many Counterstrike servers will it run at the same time...
(Note: The above is meant to be foolish and meaningless. Any other interpritation is pure coincidence. The names have been changes to protect the inocent)
Why worry? Each of us is wearing an unlicensed "nucular" accelerator on his back.
Sig changed for readability by G.W.
True; another poster has noted that power is relative. I love the setup I have now, and I'm amazed that I got anything done in 1983.
C|N>K
just say no to beige
The potato it is uninformed.
Many computer rooms use sprinklers. Halon is largely illegal now and many fire system contractors won't deal with it even if it is there.
We have a "dry" system, where you have to break 2 heads in separate zones for the system to flood, the room has to be almost 200F for water to actually flow.
Since the pipes are dry normally, it doesn't hurt at all if you accidentally wipe out a sprinkler head with a relay rack, or rip a pipe down in the ceiling. The rest of the building will be deeply engulfed in flame, and the computers will have already melted from ambient heat before the water system in that room kicks in.
In fact, my guess has always been that the reality, even with halon, is that halon/foam doesn't do you any good when the rest of the building falls down on top of your spiffy computer room.
The problem is, what happens if there's a LOCALIZED fire in that room. What if the PDU explodes into a million sparking pieces. What if the UPS explodes, bad things could happen. Of course, in either of those cases, the "bad things" would include probably sending a fairly deadly spike into the machines, frying them to the point that we don't care if the water is flowing or not.
I like music
You can find most of what you want to know on IBM Research or US Department of Energy (search for bluegene). I think both can survive slashdotting.
Top500.org
Top 500 Supercomputer list
Where law ends, tyranny begins -- William Pitt
It is much more difficult to use them for most applications most of us can think of. For example, VLSI CAD software (simulation/analysis/synthesis) is very compute intesive. However, these systems usually do not even take advantage of the multiple CPUs in a typical general purpose SMP system. You have to manually partition designs and sometimes loose the advantages of global optimization.
So don't run and order your new Blue Gene yet :)
Why do we need to have small, power-efficient supercomputers? Isn't the main goal of the supercomputer to be fast as hell? Granted, if this can be achieved while simultaneously minimizing power and size then by all means go for it. However, as stated by my parent, what sacrafices are being made?
The increase in speed is related to the reduction in size.
For a moment, let's pretend that electricity within a wire travels at the speed of light.
Now, let's pretend that we wish to carry pulses of electricity from one end of the computer to the other at a very high speed.
At some point, the distance the signal has to travel will become significant to the speed of the computer.
This is already happening in PCs. If you take a close look at the motherboard in your computer, chances are you'll see weird places where the traces just zig-zag back and forth (notice the angles on them, that's not by accident either, but I'm not going to try to explain a fourth-year university course in microwave and RF design here). These zig-zags add length to the traces so that they have the same length as other traces within the same bus, and all the signals on that bus arrive at the same time. Think of them as being "equal length headers", if you're into the throb of a big-block V8.
Length of interconnecting wires is non-trivial at this point. Stray capacitance and inductance caused by any conductor are non-trivial at this point. As a result, a terrific limiting factor to the speed of a computer is now its size.
Power consumption is also related. Modern ICs are made of millions of MOSFET transistors which behave as switches. These switches are not perfect: during the transition between a logic high and a logic low, the transistors spend time in the linear state where they are resistive. As a result, they waste energy as heat.
Stray capacitance and inductance - even within the junctions of the transistors themselves - slow their ability to switch instantaneously. As a result, they must be made as small as possible to reduce capacitance (C) and inductance (L).
This also explains why newer generations of a processor can run faster than their predecessors: smaller and smaller features on the IC mean less stray C and L, which means that the transistors can switch states faster, which means that they spend less time in the linear state and therefore heat up less. This means less energy wasted as heat.
Fire and Meat. Yummy.
If supercomputers were ubiquitous, more uses would be found. So I don't see how "need" comes into the picture. Now who can afford one? That is a good question. If they were affordable you'd see needs popping up all over.
Lasers Controlled Games!
Finally a computer exists that can easily fit in my apartment with enough power to play Doom III at 30 fps.
You've got 8% of my love - 8% of my love - 8/100's of the time you're the only girl I'm dreaming of.
I think it's hilarious that we still talk in terms of computers taking up a mere half a tennis court. Once upon a time, computers took up an entire room - and they still do.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
This set of sldes compares some of the architecture of the BlueGene/L to other ASCI machines.
Seems to me there was once a guy named Seymour, who could do the fast *and* small thing quite well. Since we're talking in terms of Televisions...
Big screen #1
40" HDTV and a A size perspective
We have DEFINITELY been down this road before folks. I don't see why it's so hard to do this, unless you're using COTS components. Hence, the point of "engineering" - not cramming a bunch of stuff in boxes/packages into bigger boxes and packages.
Detachment 3 Media
Exposed, Exploited, Exploded
you may have beige but not I :) silver maybe, but I'd rather keep my 300fps in quake than have a beige box
This one little computer is small and efficent and all the waste heat easily taken care of. Now imagine not just one of these, but a whole building of these. Our heat problem crops right back up.
IBM knows what it is doing.
That being the case, why aren't distributed apps considered as part of the Super Computer list?
Most of the tasks you pick a supercomputer for aren't things you can cut up into a thousand chunks and let every computer finish it's chunk of the problem independently. In particular, the benchmarks (LINPACK) that determine who goes where on that supercomputer list generally measure a computer's performance at big linear algebra problems (which are what takes up most of the compute time for huge classes of real problems), and for those problems every node needs to share results with many other nodes after essentially every iteration: this means you need high bandwidth and very low latency connecting the nodes.
Now, the supercomputer benchmarks may make things worse than they have to be: according to this they're measuring performance on dense matrices (where every node needs to talk to every other node), whereas many real world problems can be discretized into very sparse matrices (where each node only has to talk directly to a few of the others) instead - still, even in the sparse situation you want your computers to be separated by microseconds across your high speed interconnect rather than milliseconds across the low bandwidth internet.
now just make one that's PDA sized :)
-illumina+us "I put on my robe and wizard hat..."
"I think we should be developing and improving this kind of solution FOR EVERYONE, and not spending millions to develope something that just a few ones will have."
What's this "we", white man? Are you a supercomputer developer? Because, I mean, if Seymour Cray rose from the grave, got a slashdot account, and wrote your post, I would believe that you have a point. But he didn't, so you don't.
"Supercomputers", a set which today includes the supercluster computers along with traditional supers (like Cray's stuff), generally stay in service for a hell of a lot longer than your average desktop PC. There's iron in military labs and DOD facilities left over from spending bills that Reagan signed, still doing useful computing work today. The fact that it's not the FASTEST and BESTEST machine in existence doesn't mean that it's not still useful.
You forget that a computer becoming obsolete doesn't mean that the computer gets any slower. A super is at least as fast on the day it's decommissioned as the day it was first booted.
The fact that the price has dropped somewhat is irrelevant. Today, I can buy a desktop PC for $2500 with the computing power of a DEC mini that cost $10 million in 1982 (~). But in 1982, that amount of computing power was WORTH $10 million (again, ~), because computing power was so scarce. Or are you saying that they should've waited until the price came down before buying? Because if they did, they would never be able to buy, because prices are always coming down.
If I have a computing need that can only be addressed by a super, and I have the money (or I can get the money) and believe that the computer is worth it, who are you to say shit? You don't have supercomputer-level needs--you have DESKTOP level needs. So go buy a desktop, overclock it, and live with the fact that you can't afford the fastest computers known to man.
They have no hope of bucking the Earth Simulator and taking the real crown, so they're pretending the rules have changed.
I recently did a search of top500.org which has specs back to June 1993 and up to June 2003
BTW WHO THE HELL BROKE top500.org!?!? This site used to be easy to use and informative, now it is a banner add hell, that obscures the info you used to be able to get to easily, with many broken links and apologies for works in progress.
Anyway I digress, the point is that in 1993 the fastest computer was the TMC at Los Alamos with GigaFlops ratings of 59.7 Rmax 131.0 Rpeak
My Dell XPS today would rate in the top half of fastest machines in the world for 1993 if I'm reading the stats right with just over a GigaFlop of power.
Todays fastest machine is Japan's Earth Simulator rated at 35860 Rmax 40960 Rpeak
If we define a super computer as the ability to get in the top500 then 245.1 Rmax 384.0 are the numbers that indicate your machine would be a super computer by 2003 standards.
Letter To Iran
Ask yourself: why do you need a supercomputer?
Answer: To do a very sophisticated simulation that would be too difficult or costly to conduct in real life.
But if the supercomputer is so expensive to purchase and maintain, it might be easier and cheaper to use CAD and rapid prototyping to make a few doo-dads and knock them into each other for real, as an example.
So if the supercomputers can't scale with the rest of computing or manufacturing, then no one will buy them (no one who doesn't want to get fired for being thickheaded, at least)
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
It's the research that trickles down that can at some point provide solutions FOR EVERYONE(well not everyone at the same time). This is so obvious, it should'nt even have to be explained.
I mean, where do you think technology comes from. Come on, I have a Palm that has 75 times the clock speed of my first computer. This didn't come ex-nihilo. Huge expensive projects pushed the limits of computing and the resulting advances, combined with other developments in all branches of sciences and smaller projects in computing, made it possible.
And how do you think Parallel/Distributed computing came about? What do you even think a supercomputer is?
Remember the ENIAC, now look at a calculator. How's that for a solution for EVERYONE!
And the computers aren't obsolete when they come out!! They're no longer the fastest, but they aren't close to obsolete by any sense of the word.
That dosen't mean, however that technology is then available justly to anyone, but that dosen't have anything to do with it.
Don't mean to be rude, but this is really basic applied science.
Strioa
While progress in making supercomputers more efficient in terms of power usage and space, the widespread adoption of supercomputers is still really hampered by functionality. The majority of supercomputers are used for modeling, simulations, or code breaking. This limits their usage to academic and government institutions. These break through only help those kinds of institutions afford a super computer. I would think that most businesses have little use for that kind of raw computing power. Their computing bottlenecks are more related to transactions per time as opposed to calculations per time.
Well, there's spam egg sausage and spam, that's not got much spam in it.
Probably lots of people submitted that article. Slashdot editors, not being complete idiots, had the same reaction as a lot of the posters here, to paraphrase: "Shock! IBM makes smaller, faster, clusterable computer!" So they featured this in a group article on the 14th about a bunch of similar articles.
Later on after about 20 more people submitted it, they gave in and posted it directly. They generally attribute the person who causes them to post it, rather than a group.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON