New Top500 List Released at Supercomputing '06

Guybrush_T writes "Today the 27th Edition of the Top 500 List of World's Fastest Supercomputers was released at ISC 2006. IBM BlueGene/L remains the world fastest computer with 280.6 TFlop/s. No new US system in the top10 this year, since they all come from Europe and Japan. The French Cluster at CEA (French NNSA equivalent) is number 5 with 42.9 TFlop/s. The Earth simulator (no 10) is no longer the largest system in Japan since the GSIC Center built a 38.2 TFlop/s Cluster, reaching the 7th place. The German cluster at Juelich is number 8 with 37.3 TFlop/s. The full list, and the previous 26 lists, are available on the Top500.org site."

Comments

Misleading

[Anonymous Coward]

Misleading summary. No NEW American in Top 10, but Blue Gene is #1.

Re:Misleading

[stupidfoo]

And it's inaccurate:
#3
DOE/NNSA/LLNL
United States
ASC Purple - eServer pSeries p5 575 1.9 GHz
IBM
2006

Re:Misleading

[just-a-stone]

#3 was upgraded from 10240 to 12208 CPUs in 2006, but it already was on the list in november 2005...

Re:Misleading

[Anonymous Coward]

I think I saw the Blue Gene on King of Cars.

Re:Misleading

[stevesliva]

Well, the new top10s may not be in the US, but they're both made with processors from the US. But yeah, whatever, the sky is falling, the sky is falling.

Re:Misleading

[jarom]

The change in installation date indicates that the system was upgraded to include more processors and memory since the last time the system was on the list. See similar changes to installation date for the number 1 BlueGene/L here, here and here. Compare Purple from November 2005 to the system from June 2006. Why they don't have them linked together as the same system is beyond me.

I for one....

[JayDot]

welcome our new European and Asian supercomputer overlords.

Re:I for one....

[ackthpt]

welcome our new European and Asian supercomputer overlords.

With the Chinese quest to show they're as good as anybody, if not better, they'll probably be claiming something which dwarfs the USA DOE's Blue Gene chart-topper in a few years. To be powered and cooled at the site of the Three Gorges Dam.

probably use it to run world economic models and such to plan for when the USA defaults on hits 9+ trillion $ debt.

Re:I for one....

[Anonymous Coward]

probably use it to run world economic models and such to plan for when the USA defaults on hits 9+ trillion $ debt.

We'd all better hope not.

Repeat after me: if the US goes down, so does China, the rest of Asia, Western Europe, and a chunk of Central and South America.

Re:I for one....

[kuc]

640 TFlops Should Be Enough for Anybody.

Re:I for one....

[heffeque]

Will it be able to run Windows Vista?

What, no microsoft?

[harris s newman]

Its supprising that no microsoft systems are listed....

Re:What, no microsoft?

[Kadin2048]

Its supprising that no microsoft systems are listed....

Well, they only published the requirements for Vista a few weeks ago; I'm sure they'll do better next year.

Re:What, no microsoft?

[Anonymous Coward]

I would find that suprising too, except that there are 3 Microsoft systems on the list (5 OSX systems too). I actually read the page though.

Re:What, no microsoft?

[rockytriton]

What? No OS/2???

Re:What, no microsoft?

[Kadin2048]

Just to follow up, you can get OS information here: http://www.top500.org/stats/27/osfam/ (by family)

OS (# systems) (Percent)
Linux 367 73.40%
Windows 2 0.40%
Unix 98 19.60%
BSD 4 0.80%
Mixed 24 4.80%
Mac OS 5 1.00%
Totals 500 100%

Alternately there's a more refined breakdown listing them by Operating System type and version. Oddly, "Linux" is listed both as an operating system family and as a distinct flavor/distro ... I can only assume that the systems using "Linux" as the particular operating system are using a custom-made distro, instead of one of the commercial ones (which are listed separately on the detailed chart). Unless they just failed to report one in particular.

As for the Windows-based systems, there were one each for Windows 2003 Server and Windows Compute Cluster Server 2003.

Re:What, no microsoft?

[MrFlibbs]

Also, what about massively distributed XP "supercomputers" like SETI@home? We just had a slashdot article discussing how the SETI version dominates the other spare-cycle-using background programs. If you summed up all the cycles in use at a given point in time, how would this stack up to the supercomputers on the list?

I realize internet-linked PCs are a different beast, but given the wide range of architectures on the top 500 supercomputer list, is it such a stretch to consider this a "supercomputer"? Anyone know how the SETI@home project would place?

Re:What, no microsoft?

[HermanAB]

Well, now that you mention it, *nobody* beats MS in distributed botnets...

Re:What, no microsoft?

[awing0]

With over 900,000 computers in the system, SETI@home has the ability to compute over 250 TFLOPS (as of April 17, 2006).

http://en.wikipedia.org/wiki/Seti@home

IBM's Blue Gene is faster (and more flexible). Big networks of computers like SETI's are good at crunching static radio telescope data or brute force RC5 cracking. When it comes to most real world problems, the nodes must communicate and share data, which over the internet makes it far too slow. Real supercomputers do not use any type of networking between nodes, they have a shared memory bus.

Re:What, no microsoft?

[1729]

Real supercomputers do not use any type of networking between nodes, they have a shared memory bus.

That's not true at all. Blue Gene/L, for example, doesn't use a shared memory bus across nodes.

Re:What, no microsoft?

[47Ronin]

See # 28 .. Virginia Tech .. uses Apple Xserves http://www.apple.com/science/profiles/vatech/

Re:What, no microsoft?

[47Ronin]

Also see #21 COLSA uses a cluster of Apple Xserves http://www.top500.org/system/7741

Re:What, no microsoft?

[MMC Monster]

I'm just wondering who would run Mac OS 5 on one of these things Talk about security by obscurity!

Re:What, no microsoft?

[awing0]

I should have said ethernet instead of "networking". Blue Gene/L's node interconnect looks pretty interesting though. http://www.llnl.gov/ASC/platforms/bluegenel/arch.h tml

Re:What, no microsoft?

[flaming-opus]

shared memory bus is a pretty subjective measure. Sine the mid-90's no supercomputers have used true flat-memory shared buses. Instead they are connected by some sort of switched point-to-point network. In many of the mpp machines, these networks are built into the memory controllers on each node. (blue gene, crays, earth simulator, etc) The bandwidth and latency of these networks is orders of magnitude better than gigE, but it's still a network of sorts.

Re:What, no microsoft?

[Ed Avis]

Yeah... Unix98 I can understand, and BSD 4.x, fair enough; but what about Windows 2? No supercomputer would run that... except maybe the home-made supercomputer described in Byte magazine of April 1986 built by lashing together loads of Intel 80286es.

Re:What, no microsoft?

[GunFodder]

The real question is how much work a a botnet with several thousand nodes is doing. I guess you'd have to measure it in SMS/S (Spam Messages Sent per Second) or DOSC (Denial of Service Capability). And I doubt many countries will want to take ownership of these systems.

Re:What, no microsoft?

[after fallout]

Real supercomputers do not use any type of networking between nodes, they have a shared memory bus.
Don't say that, all it does is show how little you know.

There are 2 common types of interfaces between nodes on a supercomputer: shared memory and message passing.

Shared memory is where all the nodes can access memory over some sort of network. In order for communication to happen all 2 nodes need to do is read and write to the same location in memory. There is little talk about the network protocol used at this level because for the most part it is an emulation of layer 2 of the OSI (as if all you are doing is ordering the hardware around).

Message passing would best be described in terms of layer 7. Communication occurs between 2 nodes via messages that are sent back and forth (hence the name). The most common message passing scheme is MPI. In MPI, there is a concept of a sender and a reciever. The reciever calls MPI_Recv and the sender calls MPI_Send and a message is sent from send to recv. You could almost think of this as an HTTP communication; the server is listening, the client sends information, the server sends back, except in MPI the reciever must be calling MPI_Recv and waiting for a send from a specific sender and the sender must call MPI_Send to send the information to the reciever (there really [well, sorta] isn't a concept of a timeout). In my experience, this makes MPI (I use MPICH2) difficult to debug, if A calls send to B and B calls send to A at the same time, your program blows up (often with very little useful information).

On the cluster I do my work on, the implementation of MPI sends TCP/IP packets over ethernet (much like 256 of this top500 list). The libraries could be written to do the work over Myrinet or any other network.

For future reference please learn some factual information before you go spouting bull. If you follow this link, and choose interconnect family, you would find that most of the supercomputers in the top500 list are using some standard network interconnect.

Re:What, no microsoft?

[Miles Von Blitzkrieg]

I believe the numbers next to the OS names were referring to the number of supercomputers using that OS. Easy mistake.

Re:What, no microsoft?

[after fallout]

How would you describe #27, #51, #80, #82, #87, #93, #94, ... (256 more)

Myrinet does have an order of magnitude in latency (it is actually around the same speed in terms of bandwidth), but the cost of implementing it still makes GigE attractive.

Re:What, no microsoft?

[grommit]

It's a shame that the people doing distributed movie rendering and drug discovery aren't doing "real world problems" as you call them.

Re:What, no microsoft?

[SETIGuy]

I realize internet-linked PCs are a different beast, but given the wide range of architectures on the top 500 supercomputer list, is it such a stretch to consider this a "supercomputer"? Anyone know how the SETI@home project would place?

Someone would need to port LINPACK to BOINC in order to find out. Theoretical and real world performance are two different things. The average machine connected to SETI@home has a theoretical performance of about 1.2 GFLOP/s on non-vectorized code (based upon benchmarks of a sample of 605,000 hosts that have recently returned work). Let's assume we keep about 650,000 machines running continuously (or 1.3 million running half time). That would mean our theoretical performance is somewhere on the order of 780 TFLOP/s, which would rank us at the top.

But theoretical performance doesn't count when assembling the list. In actuality, the SETI@home application gets about 28.5% of the theoretical performance, or 222 TFLOP/s. Our long term average is 167 TFLOP/s. That would still put us at the top.

But performance in a specific application doesn't count. It's only LINPACK performance that counts.

Re:What, no microsoft?

[Anonymous Coward]

Mac OS 5 1.00%

Yep, so much for OSX, the Xserver, Virginia Tech, and the lie of "cheap apple supercomputers".

Re:What, no microsoft?

[SETIGuy]

Sorry, should have said "near" rather than "at" the top when discussing the 222 TFLOP and 167 TFLOP numbers.

Re:What, no microsoft?

[Nutria]

I believe the numbers next to the OS names were referring to the number of supercomputers using that OS. Easy mistake.

Orrrr, he could have been making a joke.

Re:What, no microsoft?

[kramulous]

What? Find the solution of a matrix (dim = 5x10^7) using some Krylov subspace iterative method with some suitable parallel preconditioning strategy to accelerate convergence on a cluster? Think again.

Re:What, no microsoft?

[face411]

Its supprising that no microsoft systems are listed....

ROFL

Re:What, no microsoft?

[brsmith4]

It really depends on your problem. If you have a code that is incredibly communications-intensive, you're likely to use a very large amount of your interrupts sending and receiving messages (if you're using MPI) rather than doing computations. There are codes out there that are essentially useless over GigE (there is no performance gain realized when increasing your used resources; it does not scale) and the very _minimum_ becomes a high-end interconnect like Myrinet; therefore, your cost of implementation argument does not hold true in a number of cases. In the case of Linpack/HPL, which is used for the Top500 benchmark, http://www.top500.org/about/linpack, the algorithm works in such a manner that the limitations of GigE are not as much a problem. Linpack is simply not representative of all of the different codes out there and in many cases, the results it provides may not provide any indication whatsoever of how well your code will run on a particular computer.

Re:What, no microsoft?

[Anonymous Coward]

That's true... if by "shared memory bus" you mean a crossbar switch.

Re:What, no microsoft?

[after fallout]

I agree, there are problems that don't scale. Most of them will see performance improvements with a lower latency network interconnect. Linpack doesn't have this problem. However, these problems (ones that don't scale with resource increases) will still not scale as you increase resources on a Myrinet network. Said problems eventually have a limit to the amount of speedup they will ever achieve.

I think there should be a variety of problems included in the top500 benchmarks. Anything that can be done to display a little better just how powerful these machines are at solving a particular problem type is a good thing for potential investors. They could use those benchmarks to decide what machines they want to use for their work. For instance I am working on a parallel alpha-beta algorithm. This algorithm would likely not run to its full potential on any parallel vector machine, because it is designed to run on clusters of general purpose nodes. In fact it would probably run best on a machine where 1 node could act as a master, n nodes as sub-masters and n^m nodes as workers (or another level of masters and so on, as long as the final level is workers). The workers could be a large distributed grid like seti, all they do is receive a packet of work, do it and then return the results. The masters need better communication requirements (and a lot more memory).

Re:What, no microsoft?

[oudzeeman]

problem is, they haven't released new XServe hardware in quite some time, and it isn't as cheap as they claim... and there were bugs that had to be worked on on some of the large installations. There are a lot of small apple clusters out there, which is where they really excel (small clusters
(someone who built one of the largest apple clusters 2 years ago)

Re:What, no microsoft?

[oudzeeman]

oops. my less than sign got chopped off. that should have said (small clusters less than or equal to 32 nodes managed by a researcher that "just wants it to work")

Re:What, no microsoft?

[andrewman327]

Fluid dynamics, for example, need very fast interconnects for processing, so gigE probably would not work too well. SETI@home type projects can deal with huge latency because of the nature of the problems they are solving.

"Flop/s"???

[Anonymous Coward]

= FLoating Operations Per per second

Should be just Flops with no /

Re:"Flop/s"???

[wiz31337]

Ummm, no... Floating Point Operations Per Second

Re:"Flop/s"???

[VoxCombo]

I believe it's:
FLoating point OPerations / second, with the / representing "per"

Re:"Flop/s"???

[muellerr1]

Still wrong (FPOPS?!? wtf?). FLoating-point OPerations / second. From Wikipedia entry:

Alternatively, the singular FLOP (or flop) is used as an abbreviation for "floating-point operation"

Re:"Flop/s"???

[zlogic]

But the irony!
The better the computer, the higher its flop per second value!

Re:"Flop/s"???

[treeves]

True, but maybe they were telling us about its acceleration (like zero-to-60 in 5 seconds).

It can go from zero to 2.8 TFlops in 0.01 seconds?

Note: I'm joking, of course.

Re:"Flop/s"???

[Transcendent]

I've known it always as the P comes from "Per". Same with acronyms such as RPM. If there is no slash used, then you must specify that it is a "per" in some other fashion, hence the P.

Re:"Flop/s"???

[wiz31337]

Did you even read the link you put in your comment?

In computing, FLOPS (or flops) is an abbreviation of Floating Point Operations Per Second. This is used as a measure of a computer's performance, especially in fields of scientific calculations that make heavy use of floating point calculations.

Re:"Flop/s"???

[Anonymous Coward]

How is this asshat even remotely informative? He didn't even RTFA he referenced!

Re:"Flop/s"???

[toleraen]

Indeed, straight from the horses, uhh, website:

Gflop/s is a rate of execution, billions of floating point operations per second. (Emphasis mine)

Re:"Flop/s"???

[teebob21]

Try again. FLoating Point OPerations/Second.

280 flops not fast enough

[Anonymous Coward]

To keep up with the rate that humans make mistakes.

Damn...

[ReidMaynard]

Spanky's Cluster'O'Porn just missed the top 500 :-(

Re:Damn...

[Anonymous Coward]

690th position isn't so bad.

Re:Damn...

[Doctor Memory]

If only they'd measure performance in FAPs, they'd be on top!

From 11 to 451...

[engagebot]

Dang, when our SuperMike was built (Lousiana State University), we were 11th on the list. A quick look now and we're at 451.

I feel old... ;0)

Re:From 11 to 451...

[sconeu]

Dude, your cluster is on fire!!!!

Re:From 11 to 451...

[sootman]

VA Tech (http://www.apple.com/science/profiles/vatech/, http://www.apple.com/science/profiles/vatech2/) has dropped from the top 10 to 28th.

This is the first time I've actually looked at a Top500 list. The DOE has 131k processors? Damn.

Re:From 11 to 451...

[Hektor_Troy]

Of course it isn't you dimwit!

It's just the paperwork ...

Re:From 11 to 451...

[identity0]

Maybe it's because your nodes were scavanged by refugees to jerry-rig shelters. After all, everyone knows how sturdy cluster node cabins are...

how many aren't listed?

[rritterson]

How well does this represent the real top 500?

If you look at the list, several of the computers/clusters are known simply as "Classified". It makes me wonder if those at the top really represent the top 10 most powerful supercomputers out there. I'm willing to be the US government, for one, has a couple of military use supercomputers up there that they aren't even willing to acknowledge the existance of.

At the other end of the spectrum, how many smaller clusters aren't on the list simply because the administrator doesn't have time to shut the entire thing down to run a LINPACK benchmark? The cluster I/we use would easily make it into the top 450, and maybe higher, but our research is deemed more important than the glory that comes with being on the list.

Re:how many aren't listed?

[Entropy]

US government, for one, has a couple of military use supercomputers up there that they aren't even willing to acknowledge the existance of. Just so long as one of them isn't named SKYNET, I'm content ..

Re:how many aren't listed?

[Andrewkov]

No, it's called WOPR. Everyone knows that!

Re:how many aren't listed?

[flaming-opus]

I would say it's unlikely that the classified computers are in the top 10, and here's why: the top500 list is constructed using linpack to measure floating point performance of highly parallel computation. Much of the work done by intelligence agencies is data-mining. It's integer tasks that are probably I/O bound rather than cpu-bound. If there were a top500 list of high performance storage systems, I bet the classified systems would own the top of the list, just not for raw fp-compute power.

Take, as an example, the Cray MTA. It's a product that's not even mentioned in their products page on their website. Yet, if you surf the net very carefully, you'll find out they're building the next version for their single customer of the product-line: the NSA. Even at the maximum configuration, the machine wouldn't make the top500 list, but it has features that make it uniquely suited to a few very peculiar application kernels. (single-virtual-cycle access to any memory within the distributed system)

Sure the department of defence uses supercomputers to predict the weather, improve weapons systems and simulate, but these are probably not done on systems we don't know exist. That sort of stuff is done at AHPCRC or ERDC, or at Beoing/lochead-martin/Ratheon/etc. All of these sites have huge HPC resources, just not the hugest of the huge.

Re:how many aren't listed?

[Anonymous Coward]

It would be kind of a giveaway if they actually posted the fact that the USA has machines in the zetaflop range...

Don't ask me for details, I don't want to get "depatriated" again!

Re:how many aren't listed?

[Khyber]

If your research is too important to risk getting a spot in this list, then you're probably on the right track as far as research goes - you're there to learn things, create formulas to follow the knowledge you seek, and to serve as an example for everyone else. Who honestly gives a fuck about being the most bad-ass supercomputer? Maybe some of the overclock zealots out there, but I'll bet 95-98% f slashdot doesn't give a fuck - they just want their shit to work and perform as expected.

Re:how many aren't listed?

[0racle]

Someone, somewhere in the military has named a cluster SKYNET. They had to have.

Re:how many aren't listed?

[Anonymous Coward]

Take, as an example, the Cray MTA. It's a product that's not even mentioned in their products page on their website.

Not only that, but the MTA's lead software architect, Preston Briggs, has disappeared. He used to be a regular contributor to the comp.arch newsgroup, but right around 1996 he dropped out of sight. I found a reference to him giving a talk at a conference in 2001, but after that...nothing.

Re:how many aren't listed?

[phasm42]

United Airline's intranet is called SkyNet.

Re:how many aren't listed?

[Anonymous Coward]

The cluster I/we use [ucsf.edu] would easily make it into the top 450, and maybe higher, but our research is deemed more important than the glory that comes with being on the list.

If you outsourced the building of the cluster rather than building it internally, it's reasonably likely to have been benchmarked, for 2 reasons: (a) to let you know if you got what you paid for, and (b) the builder would want the bragging rights for their marketing material.

As such, I'd expect most (though clearly not all) of the top 500 systems do get listed.

Re:how many aren't listed?

[singularity]

Much of the work done by intelligence agencies is data-mining.

Yeah, but is the NSA sitting on a supercomputer as a "last resort" for encryption problems? I am sure the NSA definitely prefers backdoors and easy mathematical solutions to cryptography problems, but I would not put it past them to have a few supercomputers in case those methods do not work.

Re:how many aren't listed?

[BeBoxer]

Yeah, but is the NSA sitting on a supercomputer as a "last resort" for encryption problems?

Not supercomputers in the sense they are being discussed here. The top500 list is computers that excel at floating-point operations. I have never seen an encryption method which uses floating point at all. They all use integer operations. DES, RSA, AES, MD5, SHA-1, etc. All 100% integer. In most cases cracking encryption algorithms really boils down to some sort of a search algorithm, so it wouldn't suprise me if there is overlap between systems that excel at data-mining and ones that excel at cracking encryption.

Re:how many aren't listed?

[Nutria]

it's reasonably likely to have been benchmarked, for 2 reasons: (a) to let you know if you got what you paid for

Unless they did not use LINPACK as the benchmark.

Re:how many aren't listed?

[Hrshgn]

Too bad there is no PDF of the list. Otherwise we could just copy-paste it to notepad to see the secret top 500 list.

Hrshgn

Re:how many aren't listed?

[flaming-opus]

I suspect the NSA has some FPGA-derived custom processors for this purpose. Probably rewritten for each algorithm, and probably built in-house. Attach these to the I/O bus on a big sun or IBM server which acts as the I/O engine. Pure speculation of course.

Google

[celardore]

There doesn't seem to be any mention of the GoogleNet. While it may not be used for figuring out sums and what-not, it does have an estimated 126 terraflops of computing power. I'd say that's notable. I bet at least half those terraflops are devoted to advertising aswell.

Re:Google

[neonprimetime]

I was also curious. But from reading other posts ... it probably has to do with the rules excluding spread out beasts like Googleplex.

Re:Google

[zlogic]

That's not a supercomputer, that's a really large computing grid. Supercomputers are needed for tasks that require insane amounts of RAM and a really fast system bus. When you have many computers, the data transfer speed between computers is limited by the network bandwidth, and any network is slow compared to your CPU-memory bus. That may be OK for tasks like searching but not acceptable for physics simulation/capture. Sensors connected to a modern collider generate tens of gigabytes of data per second. So to simulate it you'll need to transfer all that data through your network, which will require hours or even days to simulate just one second.

Re:Google

[Rhys]

Google's full computing power is no more a supercomputer than distributed.net or seti@home is. Yes, both supercomputers and seti@home are parallel computation, but they are very different beasts.

Does google have a supercomputer? Maybe. I'm not actually clear what use a "traditional" supercomputer would be to them. For one thing, "disk" IO is genrally not the forte of a supercomputer, at least compared to processing power.

Re:Google

[Khyber]

Considering Google's ads are text-based and don't require any encode/decode process to display plain text and HTML links (as opposed to jpeg and animated gif,) that it uses *FAR* less than you estimate. I'l bet more of that is wasted on google video, than any other single google factor, streaming live video must take a shitload of resources esp. when the video has a good chance of being viewed by more than one person at a time, at any given point thru the video.

Re:Google

[Anonymous Coward]

What the hell is aswell, and why haven't I heard of it if Google's got 63 terraflops of computing power advertising it??

280 T-flops is great...

[griffeymac]

But how does it do running a distributed.net client?

fast

[Anonymous Coward]

ComData class computer: Unified Positronic Architecture Type A
Proccesing speed:- 3,200 Teraflops
Enviro-Adaptive UberThreading capability with Dynamic Quantum memory support

Release Date: Classified
Price Plan: Classified
Customer Base: Classified

EOT

131072 Processors!

[ackthpt]

Shit! I can remember when processors had that many transistors!

hello, olde programmers home, i'm enquiring for a vacancy...

Re:131072 Processors!

[Dadoo]

I can remember when processors had that many transistors!

You know, I reall like that metric, especially when you consider each of those processors probably has somewhere in the neighborhood of 100 million transistors.

Don't feel old, though. I cut my programming teeth on a processor with only 3500 transistors (6502). The transistors were probably so large, you didn't even need a clean room to manufacture it. :-)

Re:131072 Processors!

[Anonymous Coward]

By 2030, we'll have that many cores on a single die...

Re:131072 Processors!

[Anonymous Coward]

Yeah, but unfortunately, 131071 of those cores will be reserved by the OS. Think of the particle effects as you minimize a window, man!

Re:131072 Processors!

[ryszards]

I did. Vista X2 will still look like shit.

Missing Computer

[Anonymous Coward]

What about the computer that processes Bill Gates' IRS filing?

Depends on definition of a computer?

[Kadin2048]

I'm not sure that the GoogleNet counts as a single "computer." While we can argue semantics about something like BlueGene, it at least can be directed to apply all of its resources to a single problem (whether or not they actually do this in practice, I'm not sure). If the GoogleNet can be used as a supercomputer, then perhaps it should be on the list; but my understanding is that if the system can't be applied to any single arbitrary (properly programmed) task, then it's not enough of a unique entity to make it on the list.

I'm sure they have official rules that dictate what is and what isn't a 'computer' for the purposes of the list, as well.

Rmax vs Nmax

[G3ckoG33k]

Comparing the Rmax and Nmax values it seems that the list would look quite different if sorted on Nmax instead of Rmax. Can someone explain in plain English the difference, as I didn't understand their explanation. Thanks! :)

Re:Rmax vs Nmax

[SSCGWLB]

Rmax and Nmax are two measures of the capability of a supercomputer.

The first, Rmax, is the LINPACK benchmark. The LINPACK benchmark is a measure of floating point operations per second for the cluster. They usually include theoritical ((Max FLOPs for one CPU) * (number of CPUs) == Rpeak) along with the actual. Obviously, theoretical values will always be larger then actual due to wasted CPU cycles.

Nmax is the size of the problem (i.e. the dimension of the solved linear equation)

So, Rmax is the maximum performance obtained by the LINPACK benchmark during the test, Nmax is the size of the problem the best performance was achieved at.

~nate

Re:Rmax vs Nmax

[Junta]

Rmax represents the maximum acheived measured FLOPs as a result of an xhpl run.

Nmax represents the problem size. Nmax generally is aimed to be a problem that consumes as much memory as possible without swapping.

Rpeak is the theoretical max FLOPs possible according to the processor used. For example, a PPC chip is theoretically capable of 4 Flops per clock, so multiply the clock by the number of cores in the cluster. x86_64 is theoretically capable of 2 flops per clock, so multiply cores by two. Note that AMD clock for clock doesn't do any better than intel in *this* particular benchmark, so Intel clusters inherently can climb this list better, despite poor memory performance and other factors that make them less useful in a general supercomputing sense. Itanium can acheive better floating point (I believe 8 flops per clock).

And for anyone seeking to compare Rpeak/Rmax numbers with published Cell figures, keep in mind that game consoles (and by extension cell) brag about their single precision (32-bit) floating point performance, whereas this list only deals with double precision numbers (64 bit). Cell actually is nothing special at get top500 relevant benchmark results.

Many people feel this very specific benchmark is a poor indicator of the overall effectiveness of a cluster, and consider hpcc (which includes hpl as a subset) to be a better holistic method to evaluate the value of a cluster.

Re:Rmax vs Nmax

[G3ckoG33k]

Thanks, both SSCGWLB and Junta!

Comparing no 30 and 31 on the list, the Cray XT3 and IBM BlueGene/L Prototype, shows the difference clearly:

Cray XT3 - 2652 CPU:s Rmax 11810 Nmax 1158660
IBM BlueGene - 8192 CPU:s, Rmax 11680, Nmax 331775

My interpretation, the Cray manages to produce a similar Rmax with four times fewer CPU:s thanks to their better bandwidth (as indicated by the higher Nmax). Is that a correct interpretation?

US is doing badly

[twfry]

The editors comment that there are no new 10 top US based computers is an odd comment. The US has 6 out of the top 10. Thats hardly doing poorly.

Re:US is doing badly

[stupidfoo]

Actually, as I posted above, there is a new system in the US. It's currently the #3 system.
http://www.top500.org/system/8128

Re:US is doing badly

[LnxAddct]

Not only do we have 6 of the top 10, but there *is* a new American system in the top 10, and we have 300 out of the 500 total systems. In addition to that, a quick glance at some numbers looks like we have over 80% of the total computing power on the whole list. Inserting a bias into the summary was ignorant at best, and then it turns out to be wrong. It is interesting to see how some people will make up lies and believe what they want to believe so that they feel more significant than they are.
Regards,
Steve

Re:US is doing badly

[toQDuj]

well, it's not what you have, but how you use it...

B.

Can't they use water cooling

[spicydragonz]

I see water cooling rigs all time take 2ghz CPUs to +4ghz. Why not use this for these machines? Perhaps a motherboard that could be bathed in cooling fluid...

Re:Can't they use water cooling

[crunch_ca]

Why limit yourself to water? The Cray-1 used liquid freon. wikipedia

Re:Can't they use water cooling

[DAldredge]

Because, maybe just maybe, they want accurate results?

Re:Can't they use water cooling

[scoobrs]

Because quite a few of them do use liquid cooling, but they don't order hobbyist solutions. Many Crays still use 3M Fluorinert, which doesn't damage sensitive electronics when you spray it right on a processor.

Re:Can't they use water cooling

[CtrlPhreak]

Doesn't fluorinert not stand up well to extreme cooling conditions? I seem to remmeber a overclocking article having trouble with it because it froze. Although I think that was using liquid nitrogen to cool it, but still.

Re:Can't they use water cooling

[Rhys]

Given that clusters these days are made from commodity components (Xserve G5s, for instance) and how large clusters are these days, you end up with a pretty astounding failure rate. We lose roughly two peices of hardware (in order of most to least common: memory, cpu, motherboard, disk, power supply) a week, and our cluster (Turing, 640 nodes) is fairly small. We aren't even into the late-in-life crazy-disk-failure mode that most machines get at 3-5 years old. Think about the logistical nightmare if we had to try to "drain" a system of coolant before pulling it out to service it.

Plus then you'd have to have all that (very custom) cooling equipment, pumps, etc. You'd have to watch for leaks closely, which is also a problem with air cooling and the refrigerant lines, but those have a lot less surface area of pipe/connectors to go wrong: a loop per rack for rack-mounted cooling, not a loop per machine.

Plus, as other posters have said: we'd like accurate numbers.

Re:Can't they use water cooling

[flaming-opus]

Actually the sgi columbia system at nasa does use liquid cooling, as do a number of the HP clusters. They do not use liquid cooling directly to the surface of the CPU, but they do blow the system exhaust through a radiator filled with 45 degree liquid. APC also sells a similar setup for datacenters.

What happens in really dense data centers, is that you alternate hot isles and cold isles. You face all the racks so that the front of the machines are pointed at cold isles, and the backs face hot isles. Thus no machine is breathing in the hot exhaust from the next isle over. IF you pack too many machines into a machine room, the HVAC system can't injest the hot air fast enough, and you start feeding hot air into the intake for your machines. Thus you throw a relatively simple chilled-water radiator on the back of the rack, and the exhaust air isn't very hot. Hopefully you have a drip tray and a drain on each one. The setup will probably still work if one of the racks has a failed air chiller, so long as most are working, and the hot-isle doesn't get too hot.

Re:Can't they use water cooling

[Rhys]

That's like saying that my top-rack-mounted chillers are "liquid cooling" -- only in the most technical definition of the term. The (great-grand)parent post was talking about bathing machines in coolant, which is something else entierly.

You're much better off using a cooling system that doesn't just use 45 degree water (peanties for being both water and below the dew point), but does a heat exchange with something else like a standard refrigerant, and also keeps the temp of the refrigerant above the dew point.

Case in point: Liebert XDV cooling systems. Also ideally suited for a cold-hot-cold row setup, which is what we've got going -- though the air overpressure on the cold rows (air return is on the hot row ceiling, of course) tends to make the hot row more of a "oddly hot yet with cold gusts" row unless you're in the very center of it.

Googleplex?

[neonprimetime]

How does Googleplex compare with the #'s in this top 500 list? (# Processors, max, peak, etc.)

Re:Googleplex?

[UnknowingFool]

Googleplex is probably larger in terms of processors but not faster as its function is completely different. It needs to search and index web pages efficiently not crunch through lots of numbers.

Re:Cue the Vista / Linux / Beowulf cluster jokes

[bcat24]

I for one welcome our new Beowulf cluster of Linux-running, not quite Vista-capable supercomputer overlords.

In Soviet Russia, a Beowulf cluster of Linux-running, not quite Vista-capable supercomputers imagine YOU!
On the other hand, in Korea, only old people imagine a Beowulf cluster of Linux-running, not quite Vista-capable supercomputers.

Oh! Won't somebody please think of the Beowulf clusters of Linux-running, not quite Vista-capable supercomputers?

Did I forget anything?

Blue Gene?

[theheff]

You're telling me that the fastest computer in the world is a pair of pants??

Re:Blue Gene?

[blibbler]

Well I know all my thinking takes place in my pants.

Super computers on our desks?

[davonshire]

Just a passing thought looking at this list when you peek at the bottom of the list.
you see a 2.8Ghz system with 1024 processors or some such.

Sorry I remember working on repairing a Univac computer when I was in the Navy and how amazing it sounded that Cray had produced this a super computer that could do 800
million operations a second.

(Circa 1980 or so)

You could have one of these computers for I think it was 13 Million dollars.
And how fabulous that the power supply was actually under the circular bench
so you could sit on your investment.

Consider the processing power we have now a days on our desks. A lowly
3 Ghz P4 Laptop with 2 GB of dynamic ram and 60 GB of Hard drive storage.

I've yet to see a pair up with our single or dual desktop computers today
and where they sit back in the super computer days of old. If anyone
has a link or info I'd love to hear about it.

Thanks,
  Nestalgia is the romance of historic madness.

Re:Super computers on our desks?

[stratjakt]

Ever since SIMD came to the desktop with the Pentium III, they've been "supercomputers", and the term doesn't mean what it used to.

Supercomputing used to be all about vector processing, and was the domain of Cray, et al. Now we all have the tech, and a "supercomputer" is no more than a suped up version of what's sitting on your desk - basically just big-ass clusters.

Re:Super computers on our desks?

[stratjakt]

Correction, MMX added SIMD-type functionality to the original pentium line, but MMX only worked on integers, and reused floating point registers, making the proc incapable of doing FP math and SIMD at the same time..

SSE was when intel "got it right".

And it's still not that commonly used a feature.

Re:Super computers on our desks?

[Rhys]

You could just compute backwards following Moore's Law. Or compute forward: unlike the misinterpretation of "speed doubles", the "density doubles" formulation would lead to things like the cluster I manage (Turing, now #114, originally #66) sitting on your desk as a workstation in about 15 years.

It wouldn't be very useful for running today's applications, since most are not heavily threaded/parallelized, but that gives you some idea of the speed of change.

Re:Super computers on our desks?

[vought]

SSE was when intel "got it right".


I was always under the impression that SSE and subsequent revisions were more of a response to similar technologies developed previously by AMD (3dNOW!) and more robustly by Motorola (AltiVec). The timelines for development on AltiVec/SSE in particular are a little fuzzy, but it seems as though SSE was a response to competitive innovation rather than an Intel idea.

Re:Super computers on our desks?

[stratjakt]

Well, I remember that 3DNow was playing catch-up to SSE, and came later. 64 bit extensions was the first time Intel ever played catch-up to AMD, it's always been the other way around.

I'm not sure if they implemented it before or after AltiVec, but the concept of SIMD wasn't either Motorola's or Intel's design, Cray was using it in the 1970s, as per my original point, SIMD was what made a supercomputer "super".

These days, supercomputer generally means "big cluster", and I think it's about time the term got retired. It's more of a financial/logistic feat than a technical one. Most of the rigs on this list use standard gigabit ethernet interconnects, instead of something more esoteric (and faster). If I had the cash I could order 200,000 1U rackmounts from dell and make it into the top 10. Like I said, that would be quite a feat of finance and logistics, but that's all there is to it.

Re:Super computers on our desks?

[David Greene]

Supercomputer is a pretty fluid term and really hard to nail down. Everyone has his or her own definition.

However, to say that SSE/Altivec/etc. is anything close to the kind of vector processing a Cray vector machine does is ludicrous. A Cray vector compiler can vectorize a heck of a lot more than Intel's compiler because of the machine architecture.

The other thing that Crays and other high-end HPC systems have is gobs of memory bandwidth. Linpack is a completely useless measure of how well a computer will perform on things like weather forecasting and bioscience. That's because those applications tend to require lots of communication, which quickly saturates the bandwidth of inexpensive clusters.

To give you a sense of this, I've examined simulations of vector hardware on an ordinary desktop machine and the performance bottleneck is not raw compute power -- it's the lack of memory bandwidth. The simulation itself saturates the host machine's memory interface as it tries to feed its simulated function units with data.

HPC is expensive for a reason.

Re:Super computers on our desks?

[k_187]

SSE came before Altivec

Re:Super computers on our desks?

[iotaborg]

Top 500 has lists going back to 1993; back then the fastest supercomputer listed showed 60 Gigaflops. This is probably a little faster, but on the same order of magnitude, of a high-end desktop today (I am just estimating, as I cannot find general linpack scores in under 3 minutes). So perhaps give or take a 15-year span; that is, an average computer in the present is as fast as the fastest computer in the world of 15 years back.

Re:Super computers on our desks?

[davonshire]

Holy cats guys,
  I appreciate your technowiz history etc. But my primary
reason for posting was to get a sense of how amazing and
valuable our computers are today that they would have been
back then.

  Would / could my laptop beat the fastest commercial
computer if that day.

  For all the talk of SIMD / MMX etc they didn't have Windows
XP or even LINPACK back then. So it's more a question of relative
value than technical analysis.

  Still many thanks for your replies.

But does it run...

[saleenS281]

But does it run Windows Cluster Edition? (Bet you didn't see THAT one coming)

Re:But does it run...

[Donniedarkness]

Yes, actually... at least, one of them does (Windows Cluster Edition 2003, to be exact). Someone posted the OS for the computers somewhere up there in the comments =). Not surprisingly, nearly 75% of them are Linux...

I do have to admit that I'm surprised that only 2 were Windows, though. DISCLAIMER: I'm not a Windows fan.

And just so I can say it... IMAGINE A BEOWOLF CLUSTER OF THESE!

Re:Cue the Vista / Linux / Beowulf cluster jokes

[Anonymous Coward]

1. Image a Beowulf clusters of Linux-running, not quite Vista-capable supercomputers
2. ??????
3. Profit

Yeah...

[tpjunkie]

Bite my shiny Beowulf clustered, Linux-running, not quite Vista-capable supercomputing ass!

I wonder

[Itninja]

I wonder where my old Packard Bell 486/sx 33 would fall in this list. Which makes me wonder if there's a 'bottom 500' list somewhere. I would love to see a list of the slowest computer still in use.

Re:I wonder

[treeves]

. . .the slowest computer still in use.

Mechanical or electronic?
An abacus? It gets fuzzy because there is surely a hard-to-define border between what is a computer and what is not.
Is my desk calculator (non-programmable) a computer? No? How about my HP-41C calculator? My microwave oven?

I've got a PowerMac at home still running that's not much faster than your 486/33. I just remembered I upgraded the CPU on that so it's about 300MHz. I had a Mac Centris 650 68040/25MHz but I gave it away.
It had 8MB RAM and a 230MB HD. I paid about $3000 for it in 1993, and now it's worth about $10 - maybe.
I'm sure there are people out there still playing around with Altairs and homebrew computers from the 1970s.

Maybe the one the IRS uses to print refund checks!

Re:I wonder

[erikdalen]

I bet there's more than 500 Commodore 64's still running @ 1.02MHz

And a lot of even older stuff. But then again there's probably newer stuff with slow processors like coffe machines and what not. (although a modern coffe machine can even have a 66mhz processor and telnet capability...)

Re:I wonder

[MirrororriM]

Which makes me wonder if there's a 'bottom 500' list somewhere. I would love to see a list of the slowest computer still in use.

I bet a lot of them will actually run Linux. Kinda funny how the fastest computing environments run Linux...and at the same time some you can run Linux on old crusty, throw-away PCs. From one extreme to the other ;)

I got two crusty laptops running Debian 3.1 - a ThinkPad 755C and a Toshiba Satellite Pro 430CDT (75mhz and 120mhz respectively). Not that they serve much of any purpose - just a backup file server and the other is a backup mail server. Just had to do it because I could ;P

Re:I wonder

[Anonymous Coward]

We have an old HP Vectra with a bunch of add in RAM cards in use as a production system. It runs an even older bell jar chamber.

Re:I wonder

[nyctopterus]

My grandmother still uses a "blue chip" monster from 1980. It's gotta be right down there somewhere.

Her secret? Always park the heads, dear.

Is one-number comparision really meaningful?

[mi]

Even comparing simpler things, like shoes or knives, can not be reduced to a single measurement. Microwave ovens and air-conditioners are already far more complex and come with huge vectors of parameters to compare.

Can a meaningful comparision be made of computer systems based on just one number? N TFlop/s vs. M TFlop/s? I don't think so...

What? No botnets listed?

[HermanAB]

A typical MS Windows botnet will outperform any of these machines on the SOPS (SPAM operations per second) benchmark...

BlueGene assembly line...and truck route to NSA?

[Anonymous Coward]

So, BlueGene is an exercise in design for manufacturablility. I suspect they punch out some number > 10 chassis / month down in Rochester MN. What is the standing order per month for the NSA? Might be fun to try for an community based truck traffic analysis of trucks leaving southern MN and ending up in MD...:-)

Re:Cue the Vista / Linux / Beowulf cluster jokes

[Rob T Firefly]

That's no Beowulf cluster of Linux-running, not quite Vista-capable supercomputers.. it's a space station!

Re:Cue the Vista / Linux / Beowulf cluster jokes

[Dr. Eggman]

Congradulations on using nearly every cliché joke! As a reward, I present to you this award: http://img418.imageshack.us/img418/7817/rewinner9s f.jpg

Imagine a Beowulf cluster of these!

[dpbsmith]

Well, someone had to say it.

Re:Cue the Vista / Linux / Beowulf cluster jokes

[drawfour]

Did I forget anything?

Yes. You forgot to be funny.

Non-American nationalism

[amightywind]

A common story occurance on slashdot for the submitter to highlight some deep deficiency in American business, technology, or way of life, which is then inevitably followed by non-American dick-swinging nationalism of why someone else does it better. This is a perfect example.

Re:Non-American nationalism

[Anonymous Coward]

It should also be noted that this is the list unveiled at the "International Supercomputing Conference", which tends to be a much smaller ordeal than "Supercomputing" in November, which is held in the U.S.

Everyone in the rest of the world runs linpack to acheive good numbers for ISC. Everyone in the US pushes through top500 runs in late October for SC. When the Fall list is released it will include many more US systems.

The top500 list is a marketing ploy used by most organizations for a talking point in proposals for new systems. It is not a good indicator of actual performance. Take the Cray Xt3 vs. IBM BlueGene battle for instance. The BlueGene systems perform better on top500 linpack runs, but the Xt3 is actually a more parallel architecture.

Re:Non-American nationalism

[Anonymous Coward]

Right, except in this case it doesn't seem that anyone else is actually doing better. Which seems to happen a lot; some statement about how badly the US is sucking, when in fact it's doing better than anyone else is doing.

Supercomputers and Moore's law

[mollog]

I can remember when processors had that many transistors!

And I can remember when computers had far fewer vacuum-tubes in them. And I wonder why they call them fastest supercomputers. Today's calculator has more processing power than last decade's mainframes. Why not just call them today's fastest computers? In 5 years, today's 'supercomputers' will look like jokes.

Re:Supercomputers and Moore's law

[loose electron]

In case you have not noticed, the CMOS scaling of Moore is undergoing some squirming and shrinking problems. Have you noticed the clock scaling of microprocessos has been a bit more stable in the last year or two?

Latest and greatest feature size on the transistor (now starting production) is 45 nanometers. Research devices are at 10nm to 25nm.

A useful collection of info: http://www.research.ibm.com/journal/rd46-23.html

I have been on the R+D floor of Cray (Chippewa Falls Wisconsin anyone?) and the systems used for thermal cooling of these machines is impressive. When they talk "need to limit power consumption" the motivation is that the power generating stations are limited to 10-12 Megawatts.

The comment on "the big machines that the federal government has are not listed" is probably correct (I Don't know for certain either...) but then 99.95% of the engineers inside Cray don't know what their computers get used for.

Go figure...

Re:Supercomputers and Moore's law

[AvitarX]

After 5 years the Earth Simulator will presumable still be in 28th place (from 1st). I would call that respectable.

(this assumes it will go from 10th this year to 28th next like the big one built by Apple).

Re:Supercomputers and Moore's law

[Nutria]

and the systems used for thermal cooling of these machines is impressive. When they talk "need to limit power consumption" the motivation is that the power generating stations are limited to 10-12 Megawatts.

I've wondered many times if the excess heat from Opterons and scorchers like the Xeon could be recycled for some other use.

LHC?

[Anonymous Coward]

nothing in CERN on that list(yet)?

Clusters and grids, too.

[mollog]

At the other end of the spectrum, how many smaller clusters aren't on the list simply because the administrator doesn't have time to shut the entire thing down to run a LINPACK benchmark?

Good point. And what about grid computing? Grids blow away supercomputers for processing power.

Re:Clusters and grids, too.

[Wesley Felter]

Grids blow away supercomputers for processing power.

If that's true, you're welcome to run Linpack on your grid and submit it to the Top500.

The Network is the Computer

[Doc Ruby]

Isn't the Slashdot DDoS network the most powerful "computer" in the world?

Re:The Network is the Computer

[Donniedarkness]

..you apparently didn't read the "Top 10 Least Influential Tech People" thing a few days ago. Now people worry about Digg and their Claria-infested members (diggers?).

Re:Cue the Vista / Linux / Beowulf cluster jokes

[DAldredge]

Ogg didn't smash anything...
Nothing was naked and petrified...

What about...

[Cr0t]

Hmmm... where is the NSA's Super Computer?

Re:What about...

[thomasgulch]

at the NSA, silly.

Re:What about...

[Sean Riordan]

Ugh ... Ft. Meade, Maryland mayhaps. Just a guess.

Looks like I am going to need a few more.....

[i_want_you_to_throw_]

Commodore 64s in my beowulf. Maybe a few more 128's or an Amiga.

Big Blue

[Sir_Lewk]

Wow, alot of those are IBM machines.

Re:Big Blue

[Sir_Lewk]

I didn't expect that many.

Re:What, no microsoft? or modern Apple systems?

[klubar]

What happened to the Apple ads of a "super computer in a box"? No modern Apple systems (those running Intel CPU) are on the list. A guess you can cross supercomputer off the Apple ad list. And by the way... what ever happened to the Apple ads about 64-bit computers?

Apple is still holding up well

[Darth Cider]

Notice that #21 and #28 use Apple XServes still running with G5 dual processors. The Virginia Tech system, #28, has fallen only 8 places, from #20 last year.

It's too bad this list doesn't mention cost. When Virginia Tech built its first cluster, the big news was how absurdly inexpensive it was in relation to other systems. It would be interesting to learn if that still holds true.

Re:Apple is still holding up well

[Rhys]

The higher you are in the list, the slower you are to drop. Turing (also Xserve G5s) has gone down from 66 to 114 in a year. It'd be a lesser drop if we'd done a full-cluster run (as opposed to a 4/5ths cluster run), but that'd require big expensive myrinet equipment we don't have.

Re:Apple is still holding up well

[iSwitched]

I thought this too. And both are on the low side in terms of number of cpus total. Is there anywhere one can find cost info on these? I would love to look at the dollars per GFlop ratio.

Re:Apple is still holding up well

[prockcore]

When Virginia Tech built its first cluster, the big news was how absurdly inexpensive it was in relation to other systems.

That's because they neglected to mention that Mellanox donated 1100 Infiniband adapters to Virginia Tech. That's $3.3 million in free hardware. No wonder it was so inexpensive!

Re:Apple is still holding up well

[Anonymous Coward]

That, and paying for tech labor (students) with pizzas. But jokes apart, the free Infiniband adapters and the cheap labor still wouldn't come close to knocking off 90% of the cost from a comaprable system.

That's right - the BigMac cluster cost about 10% of the avarage cost of the cluster just ahead of it and just behind it in terms of flops (you can find the links - I am too lazy to Google it for you.)

And not coincidentally, the only "Self Made" clusters (at least in the top 50) are Macs.

Re:Apple is still holding up well

[Anonymous Coward]

That's right - the BigMac cluster cost about 10% of the avarage cost of the cluster just ahead of it and just behind it in terms of flops

...and "Big Mac" did about 1% of the real work the cluster just ahead of it did. The first version of Big Mac used PowerMac towers (without ECC memory) and was good for nothing except running benchmarks to get it on the Top500 List. After Apple upgraded the PowerMac towers to Xserves for free (which also isn't factored into the "real" cost of the cluster), Big Mac still does little useful work.

petaflop in 2008-2009?

[peter303]

The speed doubling time is still about 18 months (== 10x in five years). Two more doublings from the 2005 or 2006 280 TFlops is around 2008-2009. Its a version of Moore's law for supercomputing. Though processor speed hasnt been gaining as fast in recent years, improve clustering technology and software seems to be compensating.

"Exaflops in 2020!"

Re:petaflop in 2008-2009?

[bucky0]

Oak Ridge National Labratories has been talking about upgrading their unclassified cluster to a petaflop within that timeframe. Right now they're building the first 100 teraflops or so. Some of the specs are ridiculous. At full load, the cluster is planned to take several million dollars in electricity costs. And Oak Ridge gets cheap cheap cheap electricity (they have been in the past to run enrichment, which is energy-intensive)

Apple gets bumped down.

[vistic]

I noticed that the two XServe systems on the list got bumped down.

Number 21: MACH5 (Apple XServe, 2.0 GHz, Myrinet) at COLSA
Number 28: System X (1100 Dual 2.3 GHz Apple XServe/Mellanox Infiniband 4X/Cisco GigE) at Virginia Tech

MACH5 was number 15 back in 2005, and System X was ranked at number 7 back in 2004.

Someone needs to make a new, huge XServe cluster... but maybe wait until there are Intel XServes.

450K servers + 90 petabytes?

[peter303]

A recent Wired story about their twentieth-something server farm in Oregon (near cheap electricity) has them at about 450K blades. Assuming a mix of old and new commodity disks averaging 200GB per blade, gives close to a 100 petabytes. Plus MicroSoft was blathering about 800K server farms recently which hints at its estimate of a "beat-google" number might be.

Re:450K servers + 90 petabytes?

[dantheman82]

90 petabytes really doesn't seem all that much these days for 450K machines. Let's say you have 500-750 GB per machine. That's roughly 144K machines.

Other nations are catching up

[Anonymous Coward]

> The editors comment that there are no new 10 top US based computers is an odd comment.

No. It is not an odd comment, it is a noteworthy fact. For some reason, you derive from it, that this suggests, that the US is doing poorly, while in the next sentence, you are already giving a strinking counter-evidence. For someone with a less touchy national pride, the aforementioned sentence would maybe already mean, that other nations are catching up.

ISC not SC

[deanj]

The list was released at the International Super Computing (ISC) conference, not the Super Computing conference (SC). SC'06 doesn't happen until October or November.

Big Blue eaten by a Grue

[Anonymous Monkey]

I want to build a super computer called Grue. That way when it becomes the fastest computer in the world I can call IBM and say "All your super computers are about to be eaten by a Grue," and that would make me laugh.

Re:Cue the Vista / Linux / Beowulf cluster jokes

[Kesch]

But do your Beowulf cluster of Linux-running, not quite Vista-capable supercomputer overlords run Linux?

What about the NSA?

[SamAdam3d]

Does anyone realise that the NSA is not on that list?
I am just guessing here, but the NSA has to have the fastest supercomputer there is, but I am sure it is secret enough to avoid this list. How long do you think they have been trying to break codes that they would stand not having the best technology in the world to do so? Especially with this "war on terror" and whatnot. I would bet dollars to something that costs slightly less than a dollar that the NSA has had the most powerful computer for a good long time. Anyone agree?

Re:What about the NSA?

[funwithBSD]

Ask the New York Times, I am sure they can get someone to leak that info.

Re:What about the NSA?

[Anonymous Coward]

NASA is there at number 4.

Re:What, no microsoft? or modern Apple systems?

[1729]

What happened to the Apple ads of a "super computer in a box"? No modern Apple systems (those running Intel CPU) are on the list. A guess you can cross supercomputer off the Apple ad list.

Isn't this a bit premature? I mean, Apple hasn't even released an Intel XServe, or even an Intel PowerMac, yet you expect people to have built clusters of them?

Not Supercomputing '06

[stox]

The SC06 conference is not until November. This was from the International Supercomputing Conference (ISC2006) in Dresden, Germany.

Distributed Computing Wins Again!

[BrianWCarver]

Individuals contributing their spare processor cycles via BOINC are currently producing over 380 TeraFLOPS putting them clearly in first place (if such distributed systems were counted).

SETI@Home is now operated exclusively through BOINC and it alone is doing over 167 TeraFLOPS right now, putting the SETI@Home network in second place, only behind BlueGene/L (if such distributed systems were counted).

You can contribute your spare processor cycles too by downloading the BOINC client and attaching to a cool project such as Rosetta@Home which folds proteins as part of an effort to cure human diseases. Join the biggest "supercomputer" today!

Re:Cue the Vista / Linux / Beowulf cluster jokes

[ABeowulfCluster]

Wow.. A Beowulf cluster of Beowulf Cluster jokes!

Supercomputer vs Cluster

[Vorondil28]

It depends on your definitions of the terms "supercomputer" and "cluster." In a cluster, you have many less powerful machines working in concert (passing chunks of data over a network as needed), where as a "supercomputer" is one, huge machine (whith some kind of shared memory scheme). I suppose you can have systems that fall somewhere in between that share characteristics of both, but at least that's how I see it; maybe I have no idea what I'm talking about. Someone more knowledgeable than I: feel free to correct me. ;)

Re:Supercomputer vs Cluster

[1729]

It depends on your definitions of the terms "supercomputer" and "cluster." In a cluster, you have many less powerful machines working in concert (passing chunks of data over a network as needed), where as a "supercomputer" is one, huge machine (whith some kind of shared memory scheme). I suppose you can have systems that fall somewhere in between that share characteristics of both, but at least that's how I see it; maybe I have no idea what I'm talking about.

Well, the computer I'm talking about is the world's fastest supercomputer, according to this article. Blue Gene/L is basically a bunch of nodes (each a single board computer with 2 CPUs and RAM) networked together.

Re:Cue the Vista / Linux / Beowulf cluster jokes

[dantheman82]

Did I forget anything?

Yes. You forgot to be funny.

Apparently you forgot too.

Re:Cue the Vista / Linux / Beowulf cluster jokes

[mantar]

Huh... I've always wondered about this though: You know how they always say that water flushes down the toilet clockwise in the northern hemisphere, and counter-clockwise in the southern... would these clusters be little endian in the US and big endian in Africa?

.......and North Korea

[Warg! The Orcs!!]

...will RULE THE WORLD!!!!

Re:.......and North Korea

[JayDot]

Umm, no. Having no enemy to direct attention at and no more free rice to distribute (or not) to the starving masses will reduce North Korea's ability to take advantage of the situation.

Supercomputing '06

[breckinshire]

Supercomputing '06 was awesome. They threw this foam party... and well, 12 people died of electrocution, but it was still a great time. At one point, IBM's Deep Blue took off its chassis... you should've been there.

Several things

[WindBourne]

Odd that windows comes out being listed in #2 slot; Seems like they are trying to compare Windows to Linux. Perhaps it it just a hash order.

Interesting that even the mac beats it in all areas (except possibly costs). What is also interesting is that the apple is actually more powerful on a per cpu basis.

Re:Several things

[Kadin2048]

Not sure what you mean about Windows being listed in the #2 slot ... in the chart above, that's total number of systems in the top 500 that run that particular OS.

So out of 500, there are 367 Linux systems, 98 Unix systems, 24 "mixed" OS systems (whatever the heck that means), 5 Mac OS based systems, 5 BSDs, and 2 Windows-based ones. That makes Windows pretty conclusively the least-popular supercomputer OS.

Re:Several things

[Anonymous Coward]

It is listed second in the html table. Makes no sense, unless it was simply a hash order or order in a DB.

Re:Several things

[Kadin2048]

Oh, I understand what you mean now.

Yeah, I have no idea why they chose to list it that way. If you look at the detail list, it similarly doesn't seem to be in any particular order, rank, alphabetical, or otherwise.

You're probably right about it being a hash order or something.

Wrong...MS places at 131 out of 500.

[Anonymous Coward]

http://bink.nu/Article7585.bink

The National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign, where Windows Compute Cluster Server 2003 served as the underlying operating system for a new HPC cluster that recently achieved 4.1 trillion computations per second (teraflops) on 896, 64-bit Intel Xeon processors. This result, arrived at by using Dell PowerEdge 1855 blade servers, Cisco Topspin InfiniBand switches and Force10 Gigabit Ethernet (GigE) switches, was sufficient to place the system at 131st on the Top500 list. The cluster, named Lincoln, will serve strategic campus and state initiatives, with its peak performance approaching 6 teraflops.

Not bad for a first try!!!

Self - Aware

[JCOTTON]

If you link the top 10 or 100 super-computers together, does it (do they) become self-aware?
I would like to see a survey on how many /.ers are self-aware, and the level of self-awareness.

"A person starts to live when he can live outside himself." A. Einstein

I think you'll see MS in the top 100...

[SI285]

I think MS missed the publication deadline so they were excluded from the list. But that doesn't change the fact that they landed at 131 on their first try. I think they'll break 100 as their technology matures and is used in bigger clusters.

Re:Hmmmmm.....

[MrRuslan]

Short answere: NO Long Answere: Never

Why isn't MHPCC listed?

[davygrvy]

http://www.mhpcc.hpc.mil/

beowulf cluster

[drac0n1z]

Image a beowulf cluster of these super computers!

Yes - you forgot one

[tacokill]

Dude, you forgot the most important one...

2. Profit!

You need to know something about Rosetta@Home...

[Anonymous Coward]

Uhm...I hate to break it to you, but the Rosetta@Home project is no more likely to "cure" a human disease than you are likely to fold a protein with your bare hands.

The PI of this project (David Baker) has managed to propagate quite a few myths about this project, most of which are exaggerations, and a few of which are simply untrue:

First, the Rosetta method, despite being the best ab initio protein structure prediction algorithm, is still a long way from being able to produce structures that are of practical use to anyone. They can only predict the structures of extremely short, simple proteins from sequence, and even then, they can only produce usable structures in a very small minority of cases. For larger or more complicated proteins, you wouldn't use Rosetta (there are already far better methods).

Second, Baker's method is purely heuristic, and unless it produces a correct structure, it gives no real insight into the process of protein folding (unlike, for example, Folding@Home, which uses molecular dynamics, and is therefore at least somewhat grounded in basic physical theory). Given the previous point, it makes little sense to devote massive amounts of CPU time to a method that doesn't result in significant physical or scientific insights.

Third, the "applications" of the approach to human disease are sketchy, at best -- you might be able to argue that predicted structures are useful for annotating unknown genes, but even that process is highly unreliable, and has only distant applicability to "curing" anything.

To me, this last point is the most important -- the Rosetta@Home project is eager to appear relevant to "big picture" problems, but in reality, it has little to contribute to our scientific understanding of disease. It's one thing to over-sell your work to a panel of scientists (who know how to read through BS), but it's quite another to convinvce non-scientists to devote their time and energy to a project, by selling it as a way to personally combat HIV or cancer. Baker is doing the latter, and it strikes me as unethical, to say the least....

-a scientist

Don't tell anyone but...

[Ayanami Rei]

..the codename for the facility we're designing for the *** Secure Border Initiative is called SkyNET.
We were told to change it... so we did on documentation but use it everywhere internally and with our vendors. We all have a good laugh. :-D

Oh christ on crutches.

[Ayanami Rei]

That's the most board-room compliant cluster I've ever heard of.
I bet you could dress it up in a shirt and tie and put it the corner office on the accounting floor and no one would notice.

Ugh. 1855s... Cisco Infiniband? Excuse me while I retch.

re:Top 500 List of World's Fastest Supercomputers

[WhatDoIKnow]

Which one can run linux?

Yes..

[Anonymous Coward]

but does it have boobs???

Not so much...

[Junta]

You could have massive amounts of horribly slow memory and have high Nmax. To some extent large N will generally show the performance of a cluster better..

BlueGene processors are very slow, but compensate by having an incredible architecture for inter-processor communication, enabling BlueGene to scale very well. Each BlueGene processor is less than a GHz, and while power based systems are generally more clock-efficient than others, a single BlueGene processor would look puny next to even mediocre desktop chips these days, whereas most clusters are comprised of processors that would compare favorably with a typical desktop each.

Re:You need to know something about Rosetta@Home..

[BrianWCarver]

First, the Rosetta method, despite being the best ab initio protein structure prediction algorithm, is still a long way from being able to produce structures that are of practical use to anyone.

Umm, that sounds like a reason to support the project. If it's a good algorithm then anything that might refine or improve it so that we do eventually produce structures of practical use sounds great. Maybe there are algorithms giving better results now, but 1) most (if not all) of those are not available to me via BOINC--so I can't help those interested in pursuing those algorithms (at least not with spare processor cycles) and 2) that an algorithm produces better results today is no indication that the Rosetta algorithm might not produce better results in the future--in part because of the application of LOTS of processor cycles to refine and improve the algorithm.

Also, for a non-specialist, like most of those considering contributing spare processor cycles, there is always a limit to how much investigation one can do into the worthiness of one's chosen projects. BOINC allows one to mitigate this problem by allowing one to share cycles across several projects, so that the person without the time or ability to evaluate the worthiness of a project's claims can simply throw a little seed all around and see where it grows.

But further, you particularly attack Baker's scientific claims (without sufficient evidence or even a signature to back them up--I think several of your claims about the project are just simply false), but the best a non-specialist can do when evaluating a scientist's work is to rely on the reactions of the scientific community, particularly that scientist's peers. This can be best measured by whether the scientist is affiliated with a reputable institution, such as a major University (Baker is at U Washington), the amount of peer-reviewed funding such a person gets, and the amount of peer-reviewed publication that person produces. Baker seems to be doing well in peer-reviewed scientific publications as displayed here: Publications and Baker's lab gets grants from the NIH, NSF, and the Howard Hughes Medical Institute. Those journals and foundations have qualified scientists looking at the work and if they're willing to publish the results and give the lab money, then the average BOINC user who is not a specialist in the field, is justified in throwing them a few cycles. (It certainly would be more rational than deciding not to support the project based on the vague claims of someone who only identifies themself as "a scientist" on Slashdot.)

Finally you attack the entire enterprise of linking folded proteins to cures for diseases. Sure this may be unworkable at present, but, again, isn't that the point of supporting the research!? I know the likelihood of using protein research to cure diseases if no one does protein research: ZERO! So if we contribute to some research that even marginally improves those odds, that's better than what I was doing with those spare cycles before: nothing.