TOP500 Supercomputer Sites For 2006

geaux writes to let us know about the release of the 28th TOP500 List of the world's fastest supercomputers. From the article: "The IBM BlueGene/L system, installed at DOE's Lawrence Livermore National Laboratory, retains the No. 1 spot with a Linpack performance of 280.6 teraflops (trillions of calculations per second, or Tflop/s). The new No. 2 systems is Sandia National Laboratories' Cray Red Storm supercomputer, only the second system ever to be recorded to exceed the 100 Tflops/s mark with 101.4 Tflops/s... Slipping to No. 3 is the IBM eServer Blue Gene Solution system, installed at IBM's Thomas Watson Research Center, with 91.20 Tflops/s Linpack performance." You need over 6.6 Tflop/s to make it into the top 100.

Comments

beowulf

[thejrwr]

just beowulf 100 PS3s together, that should be able to pull it off

Re:beowulf

[TinyManCan]

Shoot a couple of the Nvida G80 based GPUs should do the trick just as well :)

Re:beowulf

[Laser Lou]

just beowulf 100 PS3s together, that should be able to pull it off

That's not how to say it. You are supposed to say "Can you imagine a Beowulf cluster of these?"

Re:beowulf

[hpcanswers]

just beowulf 100 PS3s together, that should be able to pull it off

Humor aside, I'm afraid some jackass would actually try this. First off, the PS3 doesn't support high-speed networks such as Myrinet or InfiniBand. And secondly, Sony is unlikely to ever provide support for any institution that uses a video game console in this manner, unlike IBM or Cray. I have a blog post specifically about this.

Real world examples

[Anonymous Coward]

Does anyone know of any real world examples that might give us a better understanding of how fast these things really are?

Re:Real world examples

[thejrwr]

its about 1 billion p4 2.5ghz processors put together

Re:Real world examples

[glwtta]

Yeah, I can't really get a clear picture unless they put it into standard units, like Libraries of Congress, or VW Beatles... I think there's also one with stacks of stuff to the Moon - that's a good one too.

Re:Real world examples

[frank_adrian314159]

I think there's also one with stacks of stuff to the Moon - that's a good one too.

Yeah, but how many Alaskan pipelines of data can it put out? That's the one that Ted Stevens uses. It's not just truckloads of tape, you know...

Re:Real world examples

[gringer]

If you were able to stand all the current and future potential PS3 players on each others head, stretching up towards the moon...

most of them would die by asphyxiation.

Re:Real world examples

[aliquis]

Uhm, 1.5GHz P4 seems to be around 3 Gflops, so 2.5 is around 4.5 since the architecture suck ;)

4.500.000.000 Gflops = 4.400.000 Tflops or around 15700 times faster than the #1 machine, but anyway...

More like 63.000 2.5GHz P4s or so.

Re:Real world examples

[interstellar_donkey]

Wait what? I just got a fancy new AMD duel core CPU a few months ago, and you're saying they have systems out there that can outperform it by a magnitude of tens of thousands?

Gosh. I knew my PC would be obsolete, but I didn't realize it would be this soon. I'll be out by the dumpster crying.

Re:Real world examples

[Anonymous Coward]

More like 63.000 2.5GHz P4s or so.

So, a 2.5GHz P4 would be 2500 times the clock rate of the Commodore 64 (1Mhz). The C64 is only 8 bit and the P4 is 32 which would mean it would take roughly 10,000 C64's to equal 1 2.5GHz P4. 63,000 * 10,000 = 630 Million Commodore 64s in the current World's #1 machine. Didn't the space shuttle use the equivelant of 3 commodore 64s for its flight computers back when it was originally built?

Re:Real world examples

[Anonymous Coward]

630 Million Commodore 64s in the current World's #1 machine.

The Commodore 64 was 16 inches wide and 3 inches high. That would mean you could have a stack of Commodore 64s a foot high that would circle the earth 1.5 times if placed end to end.

Re:Real world examples

[Anonymous Coward]

630 Million Commodore 64s in the current World's #1 machine.

The Commodore 64 was 16 inches wide and 3 inches high. That would mean you could have a stack of Commodore 64s a foot high that would circle the earth 1.5 times if placed end to end.

Can you imagine a beowulf cluster of these?

Re:Real world examples

[Anonymous Coward]

How fast is it? It does an infinite loop in a micro-second.

Re:Real world examples

[10Ghz]

"I just got a fancy new AMD duel core CPU a few months ago..."

Two cores enter, one core leaves! Two cores enter, one core leaves!

Re:Real world examples

[MoriaOrc]

Shotgun!*

*Although I have no intention of buying a PS3 at the moment, it's always best to play it safe...

That's all good...

[DerekTomes]

...but can it run Microsoft Word? :|

Re:That's all good...

[MikeBabcock]

Sorry, didn't you notice? None of these run Windows. But I'm sure a CPU and OS emulator would run just fine.

Re:That's all good...

[Epeeist]

> ...but can it run Microsoft Word? :|

They can run the current version, but tests showed them having insufficient power and memory to run the beta versions of Vista and the associated version of Office.

Re:That's all good...

[Crayon Kid]

Yes, except Clippy becomes annoying really REALLY fast.

Don't feel bad for Blue Gene.

[BenFenner]

The way the article reads makes me feel sad for the "IBM eServer Blue Gene Solution system, installed at IBM's Thomas Watson Research Center". It slipped to number three with a mere 91.20 Tflops/s. It's like the steam shovel in that children's book. Old and outdated, no one wants it anymore. Oh wait, it's still 1,800 times faster than my new Core Two Duo machine. Apparently I'm the one with the machine that works faster the more people watch it.

Re:Don't feel bad for Blue Gene.

[Code Master]

I'll bet that Blue Gene can also retire as a building's furnace as well.

Re:Don't feel bad for Blue Gene.

[Intron]

Very similar. At the end of the book, it was just used to heat a building.

Re:Don't feel bad for Blue Gene.

[KokorHekkus]

I'll bet that Blue Gene can also retire as a building's furnace as well.

Heh.

The joke aside, in areas with district heating http://en.wikipedia.org/wiki/District_heating they sometimes also provide distric cooling so you just don't went excess heat away. In northern climates that can nice advantage since for datacenters/supercomputers/etc they often produce way more heat than they can could re-use in their own buildings

hmmm

[ILuvRamen]

I'll have to look through that list and find one near me so I can outsource my Vista booting.

100 TFPLOS is not much.

[Anonymous Coward]

Just put couple thousand GPU to work and you'll have your 100 TFLOPS:
http://fah-web.stanford.edu/cgi-bin/main.py?qtype= osstats

Btw: The FAH released a 64-bit SMP FAH clients today:
http://folding.stanford.edu/FAQ-SMP.html
Sorry, only for MacOS X and Linux.

Re:100 TFPLOS is not much.

[Anonymous Coward]

The Linpack performance benchmark is not embarassingly parallelizable.

Re:100 TFPLOS is not much.

[Anonymous Coward]

Molecular dynamic calculations are not an embarrassingly parallel problem either.

Re:100 TFPLOS is not much.

[QuantumRiff]

The really scary thing about that FaH page is that there are about 1053 computers per TerraFlop/s running windows...
and only 15 GPU's per TerraFlop..

Damn.. that is really, really impressive!

Re:100 TFPLOS is not much.

[Anonymous Coward]

The Linpack performance benchmark cannot be decomposed into uncoupled computations.

About a 50% chance.

[Anonymous Coward]

Since most of them run Linux then the x86 (or x86-64) ones probably could use Wine.

Quite the reverse

[frisket]

Is there a site for the slowest computers in the world? (My office workstation, for example :-)

--
"The best cure for sea-sickness is to go and sit under a tree" -- Spike Milligan

Re:Quite the reverse

[Ididerus]

Here's a good place to start your list (or at least see where your office comes in at)

http://www.tietokonemuseo.net/tietokoneita.htm

Re:Quite the reverse

[Anonymous Coward]

That is a truly dreadful site, at least three quarters of the information on the various machines is flat out wrong.
It looks like it was put together by a child with an attention-span disorder.

Re:Quite the reverse

[pimpimpim]

Simple trick to make your office workstation slower:

Let some PHB get the idea that it would be wise if all office PCs use _only_ virtual desktops on a central server, and put that server on a different location. So far, so good. Could be a nice idea even. But then, in a blink of true geniality, use a cheap DSL connection (1Mbit will do) to connect all PCs in one office to the server in the other location. That's a way to get your Compaq pentium III to a halt while using Word. True story :(

Look out

[jbeaupre]

Sandia's supercomputer program, along with LANL's and all the weapon and nuke work done between the two is part of New Mexico's plan to take over the world ... mañana.

Re:Look out

[timeOday]

Ah, but LLNL != LANL.

Re:Look out

[jbeaupre]

Thank you for pointing out the obvious. Note that I talked about Sandia and LANL being in New Mexico. LANL has their own supercomputer farms. No mention of LLNL because that would screw up the joke. Do you need any more explanation? Like what manana means?

Big difference between #1 and #2

[max99ted]

...No. 1 spot with a Linpack performance of 280.6 teraflops... new No. 2 system... 101.4 Tflops/s

Anyone have any insight as to why the huge difference between the top two spots? It seems that the rest (3 -> down) are a lot closer in speeds...

Re:Big difference between #1 and #2

[Anonymous Coward]

Go the link in the article, then go to the top 500 list and see the number of processors in each cluster and you will realise why they are giving so many flops.

Re:Big difference between #1 and #2

[flaming-opus]

Well #1 needs a lot of asterixes next to it. The Blue Gene architecture uses an increadible number of relatively underpowered compute nodes, each with relatively little memory, and strings them together into a cluster. It's a system architecture designed around VERY LOW COST. It works quite well for a few problems, but is difficult to use for many real world problems. Because it costs so little to build, those Department of Energy guys with the big pockets can build a VERY fast computer, at least on paper.

#2 is a more general purpose supercomputer, with a better balance of processor count, processor performance, and memory. The DOE spent a LOT of money on this machine, and thus it has a very high level of performance.

After that, you see a mix of high and low efficiency machines, but few people have the can fork over the hundreds of millions of dollars necessary for a machine that powerful. It's all about the $$$.

I'll point out, however, that the Earth Simulator is still ranked #14, 5 years after it came on-line. Of course it also cost hundreds of millions of dollars at the time.

Re:Big difference between #1 and #2

[max99ted]

Interesting...thanks for the response. Another poster pointed out that the number of processors was higher on #1 than #2 and you've explained why there are a LOT more.

Re:Big difference between #1 and #2

[Anonymous Coward]

Nope. BG isn't a cluster, at least not in the traditional sense, and has proven its versatility on many different types of work loads. All you have to do is look at the Top500 list to see in how many different types of installations it is running. It's not just the LLNL machine.

The reason there's so much difference between number one and number two is that nobody can approach BG's scalability and nobody has forked over the dough for a BG system that fits between number one and two. All it takes is buying the right number of racks.

Re:Big difference between #1 and #2

[fatphil]

Google for Zipf's law.

That's how lots of things just seem to distribute themselves.

FatPhil

Re:Big difference between #1 and #2

[flaming-opus]

It is worth noting that blue gene and red storm are very similar architectures. Both are 3D torus topologies. Both use powerpc 440 coprocessors to handle communications on the interconnect. Both use a microkernel for compute nodes, and linux on i/o nodes, with lustre as the parallel filesystem. The only tangible difference is the compute node. In blue gene, it's a second 700mhz ppc 440 with 512MB of memory. In Red storm, it's a dual-core 2.4ghz opteron with 4GB of ram. The former uses a lot less electricity, can be packed more densely, and is very inexpensive. The later is faster, and can run a larger problem set on each node.

Re:Big difference between #1 and #2

[flaming-opus]

cluster, MPP; these two things are really quite similar.

I would phrase it differently: Nobody can approach BG's low cost, both in purchase price, and in TCO. There's lots of scalable systems out there, they just cost too much.

3 out of 4 of supercomputers agree

[waif69]

Linux is the operating system to use.
FTA

Operating system Family: Linux
Count: 376
Share %: 75.20%

Re:3 out of 4 of supercomputers agree

[Anonymous Coward]

Actually, this is little misleading. Blue Gene L runs Linux in it's communications nodes, but not in the compute nodes, which run a much simpler proprietary executive. And yet it is still counted as "running Linux." Granted, one would have to be insane to pay the License fees for a 32768-node Windows XP cluster!

Re:3 out of 4 of supercomputers agree

[SnarfQuest]

And think of how much time you would spend every day verifying that your copy of Windows wasn't pirates (WGA). You'd need a very large network bandwidth just to handle that one load. Anyone remember the stories of people running shopping carts full of tubes through some of the older computers just to keep them running? With Windows, you'd have to have shopping carts full of keyboards and displays just to reboot all the BSOD's.

And, if you added another CPU, you'd probably have to buy all those Windows licenses all over again!

Password Cracker

[Bender0x7D1]

We have #141 on the list at Iowa State and we booked time on it so it could be used as a password cracker at one of our Cyber Defense Competitions.

I don't know if it actually got used, or if it was deemed "unfair" for the red team (attackers) to use it. It would have been pretty sweet if they were allowed to.

These competitions are pretty cool, and have some pretty good challenges like the red team pulling the fire alarm at 3:00AM, forcing the blue team (defenders) to evacuate the building. More info can be found at the ISU Information Assurance Student Group website, or the competition website.

Re:Password Cracker

[Anonymous Coward]

And then this one time, at band camp...

Re:Password Cracker

[h890231398021]

There's little point in using a TOP500 machine for pw cracking. PW cracking is embarassingly parallel and has almost no communication -- you'd likely get better performance by just farming out a portion of the search space to each of N normal PCs (say, on your campus network) and have each report the results back to a central coordinator. TOP500 machines are best when used for applications that need coordination among all the processing nodes, and/or that do lots of data sharing among the processing nodes.

Re:Password Cracker

[GreggBz]

The ENIAC was the first real computer, you know.

Of course, I'm kidding. Everyone should know this. :-)

Re:Password Cracker

[Bender0x7D1]

I agree. However, we can't install password cracking software on even a single machine in our labs since it is a security risk. Installing it on the supercomputer is OK since no one else has access to it during that time period.

Besides, if you had a TOP500 supercomputer sitting around, wouldn't you use it? Just for fun? Just for the nerd factor? Because you could?

Lack of people with that much cash.

[Kadin2048]

I think it has something to do with how much money people are willing to spend on supercomputers.

A lot of people are willing to throw down enough cash to get into the middle of that list, but there are only a few few people who are willing to spend the huge sums of money to build the biggest, baddest, fastest one of them all.

It's like looking at cars, and saying "huh, if we look at the most expensive class of cars, they all do 200+ MPH, but once you get down past the top price class, they all start to get about the same." It's because the market for a car that does 115 MPH is a lot bigger than for one that does 230, particularly when the latter might cost five times as much and only go twice as fast.

Re:Lack of people with that much cash.

[Venik]

Well, if all cars in that expensive class can go 200+ mph, still it would be very unusual to see one go almost 600mph. Here we have world's #2 supercomputer barely breaking 100tflops, and the world's #1 supercomputer pushing 300tflops. Quite a gap here.

Re:Lack of people with that much cash.

[Surt]

And yet, interestingly enough, there are in fact like 3 cars in the world that can go around 600mph.

Out of Date

[sogoodsofarsowhat]

The current world champion supercomputer is in Japan. and is 3~4 times faster than the IBM Blue Gene/L system. See current issue of Popular Science for details.

Re:Out of Date

[hlimethe3rd]

I assume that you're talking about the MDGRAPE machine that can do a petaflop. Actually, that machine is specialized for one type of calculation, thus it cannot run then LINPACK benchmark, and doesn't qualify for this list. It is not a general supercomputer. It's the same thing as claiming that a top-shelf GPU is faster than a top-shelf CPU: it's true for only a certain type of calculation.

Wanna know something Scary...

[hcob$]

Everyone here likes to make jokes about the Southern USA being dumb, but it's amazing how much computing power is there. Hell, Mississippi missed the top 100 by not-so-much. 115 Mississippi State University

Re:Wanna know something Scary...

[bathmatt]

Mississippi missed the top 100 by not-so-much. 115 Mississippi State University You may wanna check that list again, Mississippi has 4 in the top 100, (#26,35,48,58) It is only behind in NM in TFlops/capita.

To answer your question on why, Trent Lott.

BTW, ERDC (WES at Vicksberg) and NAVO (Stennis Space Ctr on the coast) are in MS

Re:Wanna know something Scary...

[Cutting_Crew]

are you in Mississippi?? Mississippi has been constantly in the top 10 or 15 year in and year in for total supercomputing power. i live in MS so thats why i ask. :)

Re:Wanna know something Scary...

[IvyMike]

Everyone here likes to make jokes about the Southern USA being dumb, but it's amazing how much computing power is there. Hell, Mississippi missed the top 100 by not-so-much. 115 Mississippi State University

Hell, that's a fast computer. Does it got a hemi?

Re:Wanna know something Scary...

[misanthrope101]

Not surprising at all. Somone has sunk a lot of money into trying to figure out why wrestling and NASCAR are popular. But I don't think Moore's law is going to help, really. It has something to do with the M.Q. (Mullet Quotient) of the state, and may not be a computable question.

I was going try for a Np joke, but I'm not smart enough, having gone to a public school in the south. Could someone help me out?

Re:Wanna know something Scary...

[absinthminded64]

If this were DIGG I'd DIGG your comment!!

Would you like Fries with that?

[Simon la Grue]

I'd love to know what applications are predominantly run on them... i.e. atmospheric modeling, gene sequencing, drug research, etc.

Re:Would you like Fries with that?

[bathmatt]

If you notice the largest machines are all at DoE labs, that means weapons (nuc and high energy particle) work. I know that for the DoD the largest user is geophysical (atmo/ocean modeling) This is because you need to run a lot of models every day.

What ever happened...

[Trelane]

to the one, lonely Microsoft site? Seems like it dropped off the list. Anyone know the story?

Re:What ever happened...

[p!ssa]

they upgraded it to vista, it hasnt finished booting yet...

flops per processor

[diesel66]

For all the Apple fanboys (and gals) The fastest Apple system is COLSA at #28, with 3072 CPUs making 16180 Gflops, for 5.26 Gflops per CPU overall. Meanwhile, #1 BlueGene has 131072 CPUs making 280600 Gflops for 2.14 Gflop per CPU. Clearly, BlueGene is a piece of junk :)

  I wonder how much faster the Intel versions will be in comparison to the G5s...

Re:flops per processor

[Wesley Felter]

I wonder how much faster the Intel versions will be in comparison to the G5s...

Look at #20; it's almost identical to an Intel Xserve.

Re:flops per processor

[Lussarn]

Lets just hope they didn't invested in any Altivec code.

Re:flops per processor

[MojoStan]

I wonder how much faster the Intel versions will be in comparison to the G5s...

Look at #20; it's almost identical to an Intel Xserve.

Did you notice #5? It's almost identical to a blade version of a G5 Xserve (but running SuSe Linux). Those PowerPC 970 processors still ain't bad for servers, even though Apple abandoned them.

If anybody's curious, here's Barcelona Supercomputing Center's brief description of their system: MareNostrum System Architecture

MareNostrum uses 2560 IBM BladeCenter JS21 blades.

Re:flops per processor

[clarkc3]

If you want to look at per cpu performance, look down around #69-70 and notice Hitachi has one that cranks out 112Gflop per CPU - seems the Apple and Intel ones both have a ways to go to catch up to that

Performance per processor and frequency

[Anonymous Coward]

It would seem that, despite all the negative publicity that it's got, the systems that deliver more performance on a per processor basis are those based on Itanium 2, even without correcting for its comparatively low frequency.

The one they don't tell you about...

[adrenalinekick]

is the one sitting in the NSA cracking all your passwords right now.

Re:The one they don't tell you about...

[geaux]

I thought "translator" was fried.

Re:The one they don't tell you about...

[flaming-opus]

A lot of what the NSA does is not floating-point math. In all likelihood, most of their needs are data-mining, automatic translation, and other database-intensive applications. I'm sure they have a lot of very expensive computers, but they may not be the kind that end up on the top500 list.

Just today Cray pre-announced the XMT machine a href="http://www.cray.com/products/xmt/" which is the next generation of their machine for graph-tree algorithms. The product line has been basically funded by the NSA. It won't, however, make the top500 list anytime soon.

Re:The one they don't tell you about...

[Ynsats]

Don't be so sure of yourself. There are quite a few reasons for the government and military to need all kinds of computing power. Clustered super computers can come in handy for lots of things including simulations, software testing for many systems such as guidance systems and radar systems and even things as simple as artillary trajectories. You remember those problems right? The whole reason the computer widely accepted as the FIRST computer was ever built.

Just because it doesn't seem to fit what we see as the NSA's MO doesn't mean that the NSA doesn't have use for floating-point math. The whole idea of the NSA is to make sure secrets are kept safe. To help keep those secrets safe, they don't even talk about what secrets they are keeping safe or how they are keeping them safe. There are plenty of scenarios that I can think of off the top of my head for using a floating point processor in building something like an algorithm for a statistical model to show trends apparent in data mined from internet search engines.

The realm of government secrets is an odd area to play speculation in. Just because you don't know about is the best reason to think that something like it actually exists. Will you know about something such as that? Unlikely. At least not any time soon. I wouldn't doubt that there is already technology out there so far ahead of it's time that it may never be declassified due to the dangers it could pose if it was obtained by people looking to do harm.

Re:The one they don't tell you about...

[OverlordQ]

Is it just me, or did you know the answer was three while all these smart idiots couldn't figure what the fuck the answer was.

terraflop doesnt get you the list anymore

[peter303]

A billion dollars is too small for the Forbes 400 list and a teraflop is too smal for the SC500 list.

Can you imagine..

[palindromic]

a beowulf cluster shooting you in the face?

You need over 6.6 Tflop/s to make it into the top

[robyannetta]

You need over 6.6 Tflop/s to make it into the top 100.

...or three PS3s, but I don't forsee any one of us getting our hands on at least one of these for a few months...

Tflops versus CPU's worth it?

[Saltine Cracker]

I find this list amazing simply in sheer numbers on it. But I'm called to question the nature of whether he who has the most money wins the contest or not. I mean #1 on the list has five times the number of processors that #2 does for less than 3 times the Tfops. I'm not a super computer clustering genius, the largest system I've worked on is the Aeroshark Linux System and NASA GRC ~128 nodes, but doesn't it just boil down to who can spend the most money to put the largest system together?

How bout some Kudos for the Hitachi systems with less than 100 processors making the top 100?

Re:Tflops versus CPU's worth it?

[flaming-opus]

Well, what Hitachi calls a processor, in the SR11000 series of computers, is actually made up of 8 IBM power processors. They use some special syncronization hardware to make it act like a big vector processor. Thus, if you want the cpu count to actually measure the number of chips, multiply their processor counts by 8. It's sort of the reverse of multi-core.

You are correct that the absolute number of processors does not always indicate how fast the real problem gets solved. For tasks that don't parallelize very well, #1 is going to perform less well than on other systems with a smaller number of faster processors. Of course you have to also account for memory performance, interconnect latency, interconnect bandwidth, mean-time-to-system failure, programming environment, system software, filesystem performance. Linpack is just one measure of system usability.

Re:Tflops versus CPU's worth it?

[Anonymous Coward]

BGL is a breakthrough in supercomputing because it demonstrates that low-cost, under-clocked CPUs that don't produce much heat can be used to make a very fast and relatively inexpensive supercomputer. Supercomputing is not just about the number of CPU's, or even FLOPS/CPU. I am a grad student who works with LLNL supercomputing researchers. I've actually taken a tour of BGL and #4 on the list, Purple :-). Although it has fewer processors, Purple actually cost like twice as much as BGL, takes up 3x as much space, and uses 2-3x more power - and we're talking about megawatts of power (capable of causing brown-outs) and thousands of sq feet of floor space. Half the cost of these computers is spent on the interconnects between CPU's, nodes, and racks, and BGL wins there too because the CPU's can be packed closely together in torus rings because they don't have the same cooling requirements.

Re:Tflops versus CPU's worth it?

[kramulous]

Man, you are one lucky bastard. I work with Australia's SGI machine (#73) and some others (not even worth mentioning). I'd love to be aware of shit like that. I'll continue with the apprenticeship and then go overseas.

TFOLPS/S is redundant

[gr8_phk]

Just like "ATM machine" and "CMA awards". Maybe Intel is behind this mistake - you can only get TFLOP Squared performance with a core 2 duo double dual thingy. Or perhaps they really are measuring how fast you can ramp your workload from 0 to X TFLOPS because of all the computing-on-demand hype?

Re:TFOLPS/S is redundant

[Chandon Seldon]

No, it's right. The processing power of these systems is actually increasing at that rate. Every second, BlueGene/L is able to do 280.6 trillion floating point operations more than it could do the previous second.

Re:You need over 6.6 Tflop/s to make it into the t

[Anonymous Coward]

Actually three dozen if you don't count gpu (for which real-life problems tend to be hard to convert). And more than hundred if you insist double precision.

What they use nr. 1 for - you should have guessed

[PietjeJantje]

Nuclear weapon simulation. Err, sorry, "to increase the understanding of enduring stockpile." http://www.nv.doe.gov/nationalsecurity/stewardship /default.htm

Distributed Computing Wins Again...

[jemecki]

I know this isn't a fair comparison but the SETI@Home grid runs at 250 TeraFLOPS. Many of the other massive distributed computing projects run far into the Top 500 as well. reference

Re:Distributed Computing Wins Again...

[jemecki]

D'oh! The SETI@Home grid is actually only the SECOND fastest supercomputer. I should've read the f* summary.

war is over

[Andrei D]

If you look at the operating systems statistics , you can clearly see that the war is over and Linux has won :-)

Re:war is over

[kramulous]

Completely different architectures. Linux for one type, Unix for the other. I prefer the unix any day of the week. Many problems I work on cannot be done on the clusters. You are referring to different battles.

I've got some spare time tonight

[kbob88]

You need over 6.6 Tflop/s to make it into the top 100.

Only 6.6 TFLOPS? I'll get right on that.

Learn the units...

[Anonymous Coward]

It's not Tflop/s, it's Tflops.
Terra Floatingpoint Operations Per Second.

The p and the / is the same, so if you say
Tflop/s you are really saying
Terra Floatingpoint Operations Per Per Second.
I guess Tflo/s would work, it's wierd though... /Z

Re:Learn the units...

[Anonymous Coward]

It's been backronymed as "Tera Floating-point OPerations / Second".

The SI writing style dictates that you use / for "per". "p" is already used for "pico". And when are people going to stop misspelling "tera" (= 10^12)?

How many flops?

[RealGrouchy]

You need over 6.6 Tflop/s to make it into the top 100.

So where does that put Billy Crystal?

- RG>

Only 6.6?

[gnud]

You need over 6.6 Tflop/s to make it into the top 100.
Then they forgot my botfarm! arggghhhh

Old supercomputers make great space heaters too

[lhaeh]

From http://www.deadmedia.org/notes/37/371.html

"While at Convex, a Texas-based supercomputer company, Steven Wallach, a computer designer, once used an Alliant supercomputer in his office as a conversation piece and as partial support for his desk. "But even Mr. Wallach (...) said he was surprised to learn that another Convex employee had bought a Convex C-1 for its scrap price and was using the computer to heat his garage."

Livermore

[Anonymous Coward]

I always hated living here, but I guess it kicks ass that #1,#2,#4, and #6 are located in my hometown.

Re:Livermore

[Anonymous Coward]

Oops. I guess #2 and #6 are at the other Sandia. Damnit.

Well...

[jesterpilot]

At least i found the number one: http://www.research.ibm.com/atomic/nano/roomtemp.h tml

280.6TFlops

[jandrese]

Hmm, I think that's right around the projected minimum specs for Neverwinter Nights 3.

54 and 30 top 500

[Lowlow]

Im sitting right next to number 54 right now and number 30 will be here in a couple months..