Slashdot Mirror

News.com is reporting that IBM has won a $290 million contract with the federal government to build what are expected to be the world's two fastest supercomputers at Lawrence Livermore National Laboratory. The other machine, the Linux powered Blue Gene/L will be 10 times faster than the current #1 , NEC's Earth Simulator with a speed of 360 teraflops, according to IBM.

I take it you have a vested interest in itanic?
I take it you have a vested interest in athlon-64?

For your information I have a 64-bit dual processor Sun Ultra 60 and a Dell PC under my desk. I also run several multi-processor 64-bit servers used daily for building Open Source and Free software.
And I service two of the largest alpha clusters in existence ranking 2 and 3 in the top 500 supercomputers, your point being?

were artificially removed from the market place to try to boost itanic's position
Actually no, itanium was co-designed by HP and is to my understanding the successor to pa-risc which was not artificially removed, it lives on. Alpha was dropped because most companies don't want to compete with themselves. Though the alpha engineers were transfered to intel (those that didn't quit because of it) so alpha will live on in itanium as well. Alpha technology is going to be around for some time yet.

Athlon 64 will succeed simply because it is an evolutionary improvement on existing technology
Read "we couldn't come up with anything better than adding 64-bit extensions to x86 microcode".

If intel made a jump to quantum processors and amd decided to add 1024-bit extensions to ia512 (just random guess) would you say the same thing? Anytime you have a major change in architecture you're going to break things, if not for these changes we'd still be running on an 8088.
DEC took that gamble with the alpha early on, apple with the powerpc.

Granted, I do have an interest in itanium, I'd like to see a couple large clusters of those running linux put in here. But if they don't then it's not really going to bother me since I'll probably be here working on the alphas for the forseable future.

From the Linux cluster alone at LLNL, they stand to make a quarter of a million dollars. What company would pass up this chance for YAIPW (Yet Another IP windfall)

Too bad the alpha is dead. It is taking years for intel and IBM to come up with a chip that comes close to alpha performance. Good thing that they are competing against old alpha designs and the EV8 has been killed. Otherwise those darn pesky spec numbers would have been embarassing.

Alpha gets added to the list of failed, technicaly better products.

how evil Nazis [were] was not discovered until after the war

Not quite true.

Leaving aside a discussion as to the definition of evil, the broad lines of the "final solution" were well known from around 1941...

It was well known from before 1933 that anti-semitic groups were active in Germany, and were on the way to taking political control.

Read Address Unknown, first published in 1938, set in 1932 - 1934.

Perhaps many people did not grasp just how enormous the "implementation of the solution" was... Industrial-scale extermination of an entire ethnic group!

Getting back to the topic, why do you think governments legislate to limit which countries can buy certain technologies?

• Can China buy an ASCI Q?
• Can China buy combination cattle-prod cum truncheons?

Raise your hand if you've put together a 300-node cluster before. I'm sure the guys at LANL, one of the most respected technical supercomputing labs in the world, have done the price/performance/stability trade-off thing.

An armchair critic on slashdot doesnt hold much weight compared to the site that has 2 of the top 5 supercomputer clusters.

Yes...here is a pretty good start for a list of mostly Linux clusters...

You might have been thinking of #8...although #5 beats it...and it was done by Linux NetworX...The related /. story is here.

However, check out #80 :)

Self-made
Pentium 4 Xeon Cluster
2.2/2.4/2.8 GHz - Giganet -
MSWindows/ 360

There doesn't seem to be any other Windoze boxen on the list...wonder what is the story of this cluster...I wonder if M$ bought themselves a cluster :)

I'd like to say USC has a machine in the top 100, but we don't. :(

I'd like to say USC has a machine in the top 100, but we don't. :(

This should put them in the lower part of the top 500 list if they go through whatever hoops necessary to get the results posted. see Top 500 List With only 33 boxes at about $4,000 a node, that's a very cheap path to serious performance.

What is interesting about the earth simulator is for the first time the world's fastest supercomputer is located outside the USA.

you dont think they can monitor all the email that traverses the internet and all the telephone calls that cross the country with the crap they have now, do you? uncle sam wants your email!

Anyone else notice that? Power4 is the current generation, and holds the 9th spot on the top-500 list with only 1280 processors!

I'm sure IBM is working hard on a new interconnect for this beast. Anyone know about the next-generation SP switch?

The press release also mentions that Purple will consist of "196 seperate computers" -- which works out to 64-processors per computer. Way to go IBM: the current Power4 systems are only to 32-way!

..rejected by Taco lamer and brought to you by the hacker underground.

News.com is reporting that IBM has won a $290 million contract with the federal government to build what are expected to be the world's two fastest supercomputers at Lawrence Livermore National Laboratory. The other machine, the Linux powered Blue Gene/L will be 10 times faster than the current #1 , NEC's Earth Simulator with a speed of 360 teraflops, according to IBM.

..rejected by Taco fucker and brought to you by the hacker underground.

News.com is reporting that IBM has won a $290 million contract with the federal government to build what are expected to be the world's two fastest supercomputers at Lawrence Livermore National Laboratory. The other machine, the Linux powered Blue Gene/L will be 10 times faster than the current #1 , NEC's Earth Simulator with a speed of 360 teraflops, according to IBM.

Mr. Shatner,

Is the priceline.com supercomputer a Cray, a beowulf cluster, made by IBM.. what is it? The Top 500 list makes no mention of this so-called "super computer"

According to the SETI@HOME stats page, SETI is running about 45 TFLOPS, which is slightly ahead of the Earth Simulator's 40 TFLOPS or the LANL 10 TFLOPS machines. This isn't real precise - Top500 uses Linpack as their benchmark, which is a lot more realistic and controlled than SETI, so your mileage may vary. And of course that's Today's measurement from SETI, which is fairly variable in its CPU speed.

Since nobody is answering your question: The Top500 supercomputers are ranked by the results of the LinPack benchmark.

Actually, I think the 11.2 Tflops figure was in regard to single precision...

On the front page, the statement: "Rmax: 5.69 Tflops"

The LINPACK performance number from beowulf-style clusters is derived from the aggregate total performance of each node.

Read the description at the TOP500 site. You are allowed to increase the size of the problem to scale better, but it is the total performance of the entire cluster _when running in parallel_ (which is significantly lower than the sum of node performances).

LLNL Linux Network/Quadrix supercluster if build out of Penguin Computing 1U Relion 140's:

$4,747,392 offering 11.2 Teraflops...

$423.87/Gigaflop...

OTOH, if you can take advantage of it, that would put this cluster at #250 in the Top 500 list of supercomputers. In fact, it is just a tick behind an IBM NetFinity cluster with 512x733MHz Pentium IIIs. Not bad for 66x1GHz G4s.

can be found here.

How about we calculate density by flops or something else useful. I mean, how difficult would it be to cram a butt load of Pentiums in a rack? Yeah well how much calculation can they do?

Lets cruise on over to the Top 500 and use their handy dandy html list to view 'most powerful chip'. This unfortunately requires a little calc work because they failed to include this number in their table.

#1 NEC Earth-Simulator 35,860.00 GFlops using 5,120 Processors -- WOW!

But that's only 7 GFlops per processor ... that thing is mamoth with 5,120 processors.

Now lets look at a little different design ...

#14 Hitachi SR8000-F1/168 1,653.00 GFlops using 168 Processors -- Hot DAMN!!

This is more like it. They're pulling 9.84 GFlops per processor. With their architecture they could pull off the Earth-Simulator's GFlop rate with 3,645 processors - That's 28% less computer doing the same amount of work. Which means if the Earth-Simulator had been constructed with Hitachi's hardware, they could have been pulling 50,380 GFlops in the same cubic footage.

Now this is all rambling that assumes that the processors are similar in size. Which probably isn't true. But they are also getting more power out of less hardware, and it is rare that THAT isn't a bonus. ... ramble ramble ...

How about we calculate density by flops or something else useful. I mean, how difficult would it be to cram a butt load of Pentiums in a rack? Yeah well how much calculation can they do?

Lets cruise on over to the Top 500 and use their handy dandy html list to view 'most powerful chip'. This unfortunately requires a little calc work because they failed to include this number in their table.

#1 NEC Earth-Simulator 35,860.00 GFlops using 5,120 Processors -- WOW!

But that's only 7 GFlops per processor ... that thing is mamoth with 5,120 processors.

Now lets look at a little different design ...

#14 Hitachi SR8000-F1/168 1,653.00 GFlops using 168 Processors -- Hot DAMN!!

This is more like it. They're pulling 9.84 GFlops per processor. With their architecture they could pull off the Earth-Simulator's GFlop rate with 3,645 processors - That's 28% less computer doing the same amount of work. Which means if the Earth-Simulator had been constructed with Hitachi's hardware, they could have been pulling 50,380 GFlops in the same cubic footage.

Now this is all rambling that assumes that the processors are similar in size. Which probably isn't true. But they are also getting more power out of less hardware, and it is rare that THAT isn't a bonus. ... ramble ramble ...

According to top500, top500 , The five most powerfull supercomputers in the world at the last update are:

1. Earth-Simulator-> 35860.00 Gfps
2. ASCI White, SP Power3 375 MHz-> 7226.00 Gfps
3. AlphaServer SC ES45/1 GHz ->4463.00 Gfps
4. AlphaServer SC ES45/1 GHz ->3980.00 Gfps
5. SP Power3 375 MHz 16 way ->3052.00 Gfps

Himmm interesting the second computer(ASCI White) runs at a lower MHz(375MHz) than the third and fourth comptuers(1GHz each).

There are real benefits for Canadian research that can come from this project - certainly there are a number of problems that are numerical and parallelizable, so there can be a lot of future to it if they do enough coordination, but most of Canada's academic supercomputing is currently driven by SETI. Besides scientific research, the other traditional users of supercomputers are weather prediction, oil exploration, and sometimes financial modelling - Canada may have more total supercomputer-based supercomputing than anybody realizes, in addition to SETI. However, the June 2002 top500.org list doesn't show anything in Canada above #227.

Other results from the Top500.org list - SETI@Home is still about 7 times as large as the largest single machine on the list , Japan's NEC Earth Simulator, which is about 5 times as large as the #2 machine, LLNL's ASCII White.

See the supercomputer top500.

I think that a year ago or so, the Japanese supercomputer for earthquake simulations had more power than the other top 499 supercomputers combined.

Sure, they'll be able to build a large, loose network of computers, but the access-speed will hardly compare to a single-site computer.

Here is a 256-CPU IA-64 Linux cluster, #53 on the top 500 list. And here is CPlant, #50. You can find more Linux boxes on the top 500 list

In retrospect, however, it would seem that the obvious cost benefits of Beowulf very nearly killed the development and use of large SMP and vector processing systems in the US. My understanding of the situation is this:
* Before Beowulf, academics had a very hard time getting time on hideously expensive HPC systems.
* When Beowulf started to prove itself, particularly with embarrassingly parallel problems using MPI, those academics who happened to sit on DARPA review panels pushed hard to choke off funding for other HPC architectures, promising that they could make distributed memory parallel systems all singing, all dancing, and cheap(er).
* They couldn't really deliver, but in the meantime, Federal dollars for large shared memory and vector processing systems vanished, and the product lines and/or vendors with it.... at least in the US.
* Eight years later, only Fujitsu and NEC make truly advanced vector systems, and Cray is only now crawling back out of the muck to deliver a new product. Evidently someone near the Beltway needs a better vector machine, and Congress ain't paying for anything made across the pond.

Cutting to the chase, did you advance a "political" stand among your peers within the public-funded HPC community, or were you just trying to get some work done with the budget available at NASA?

A supercomputer is a single system image. Some people call large clusters "supercomputers," but technically they're wrong.

Says who?

Once upon a time 'supercomputer' meant 'any computer made by Seymour Cray', and this was reasonable, because he (probably) invented the concept. Then there was the mid-80's loose but widely-accepted definition 'any computing system that can do more than 200 MIPS'. Then MIPS went out of fashion and processors got faster and it was 'anything that does more than a GigaFlop'. Or there's the US Department of Commerce definition which was 'any computing system that does more than 195 Mtops (Million theoretical operations per second)' during the 80's, which then got changed to 1500 Mtops and is probably something different now.

Note that most Linux cluster systems would meet the requirements of most of these - indeed, most single-CPU computers today would meet most of these requirements, which is how Apple manages to get away with calling the G4 a 'supercomputer'.

Really, these days 'supercomputer' means absolutely anything you want it to be, although if I had to define it, I think probably the fairest definition would be 'anything that can run the LINPACK benchmark suite and get on the Top500 list'.

Nice try at creative redefinition though.

yeah, i was wondering how many of the top 500 supercomputers could open a word document. if they can't open it without formatting problems they must really be useless, right?

I work for a company that writes software for those kinds of genomic computations (yes, it runs on Linux, MPI & SMP). We recently did a large computation on the 4th largest super computer in the world. The results are freely available.

Most of these computations are pretty intensive in CPU and memory usage. Network speed and disk speed are less important (although you need lots of storage). I would like to try one of these babies, must be fast.

the PS2s were very powerful for the time and could be networked to create a crude missile guidance system.

Well, the PS3 will not have this problem. You see, it can't create a crude missile guidance system (any better than its predecessor). On the other hand, it can simulate a crude... nuclear explosion! Number 23 on the TOP500 supercomputer list, sitting under everyone's TV. Wow.

Consider that a 1TFLOP system would currently be ranked the #24 most powerful computer in the world

As other people have said, this must be a typo.

• the fact that we have largely moved away from an algorithmic model to an event-driven one, which is inconsistant and unstable by nature.
• Program states used to be simplified representations modelling reality (or fantasy). They increasingly base their state on raw reality. It doesn't matter how predictable an algorhithm is if you can't predict the state getting fed to it.
• Due to a combination of both laziness and overwork, as well as a preponderance of reinventing the wheel, most algorhithmic software-based systems use too much code that's too poorly tested to even dream of reliability.

• You can't predict in what order sensors will trigger, making it possible to have unexpected effects when sensors trigger in an unexpected fashion
• Digital data and comparison sensors will never allow for a perfect decision to be made regarding analog reality.

"However, these add-on clustering solutions come from various sources, do not conform to any set standards, and are often implemented on a particular Linux distribution."

What are they talking about?

Hmm, MPI and PVM are standards, more so they are _the_ standards, and are supported in Linux beowulf clusters.

Even if it were true, with Linux clusters in the top50 (not a typo) of supercomputers, whatever they use for clustering is a standard on its own. Beowulf so widely used that it is a de facto standard too.

The MS stuff is not a standard and only implemented on a particular MS distribution...

Actually, LSF runs on RedHat, Suse, OpenLinux, TurboLinux (LSF v4.1), Debian, and Suse. It even says 'tested with', so it doesn't even force you to use one of those. So which particular Linux distribution did they miss (ok, mandrake and gentoo)? And which "potentially financially unstable Linux vendor" does that bind you to if it'd very well possible to run it on the other distributions, just not tested by the supplier?

Maybe MS thinks clustering is mainly failover, but that's much less valid for stable operating systems.

This new page will not survive long either.

Thus, moron qualification is more appropriate if assigned to you.

I mean, talking about the Alpha? Who the hell uses that. That has been dead for years!!

How about the third and fourth fastest supercomputers in the world? Including the fastest in Europe.

Not to forget that Titanic, by far the highest grossing motion picture of all time, was rendered with the help of an AlphaPC server farm running Red Hat Linux. Admittedly, that was five years ago. But it was still a 64 bit OS.

I must confess my bias, because I have two Alphas faithfully running Red Hat Linux at home. So far as performance is concerned, I'll be surprised if the Intel Itanium or even the AMD Hammer chips will compare favorably to comparable Alphas on floating point performance, which is very important in many high end applications.

Of course, I could be wrong....

I mean, talking about the Alpha? Who the hell uses that. That has been dead for years!!

How about the third and fourth fastest supercomputers in the world? Including the fastest in Europe.

Not to forget that Titanic, by far the highest grossing motion picture of all time, was rendered with the help of an AlphaPC server farm running Red Hat Linux. Admittedly, that was five years ago. But it was still a 64 bit OS.

I must confess my bias, because I have two Alphas faithfully running Red Hat Linux at home. So far as performance is concerned, I'll be surprised if the Intel Itanium or even the AMD Hammer chips will compare favorably to comparable Alphas on floating point performance, which is very important in many high end applications.

Of course, I could be wrong....

I guess it's some braindead idea NY Times has got. For benchmarking supercomputers (which of course necessarily doesn't say much about the performance of any particular application), flops is still the way to go. To be more exact in the form of the linpack benchmark, of which results are published at the top500 site. Incidentally, the site has recently been updated (usually 2 times per year).

I was looking at the list of supercomputer rankings, and I couldn't help thinking - yeah, but what about all the CLASSIFIED computers? I bet the US gov has secret computers that would blow that list away.

-

You can get an SX-6i

True, but if you look at the Top1000 list, you'll see significantly more IBM machines across the board then NEC, including a large number of "standard" units (sold as kick ass RS/6000's vs "supercomputers", e.g. the P690

I would think that this gives them a signficant edge in development costs as well as giving their customers more flexibility.

And I don't think you would sell many supercomputers for IBM if you were advocating Gimp and Mozilla as applications

Oh come on, nuclear physicists like to clean up photos of their dogs (probably don't have girlfriends) and surf the web just like anyone else ;) Imagine the speed in which those nerdy scientists can apply those Gimp filters to all that pr0n they download.

Imagine a Beowulf cluster of these !

Let's actually answer all these "imagine a cluster of racks!" remarks, shall we?

As shown here, the current record holder for the world's fastest supercomputer is ASCI White. It has a theoretical maximum of 12.288 teraflops and cost $110 million. Seems like a worthy contender.

Apple quotes their dual processor machine as having a theoretical max of 15 gigaflops. A module with Dual 1GHz G4s, quad 120 GB, and 512MB of ram costs $5,649.00 a pop. Doing a little bit of math, we find that we need at least 820 modules to match ASCI White. On principle, let's fill up all the racks (42U), getting a total of 20 and thus 840 modules. The racks cost $1.4k each. This brings us to a grand total of $4.8 million. Throw in a million or so for facilities, so about $6 million total.

We go back to our original number of $110 million for 12.3 teraflops. With Apple, you could theoretically get a 12.6 teraflop machine with 400 terabytes of storage for a little more than 1/20th of their cost. Not bad.

Now, imagine a beowulf cluster of those!

A lot of us snickered when Apple pitched the G4 as a "supercomputer" (using the technical export definition), but if folks like Genentech build racks of these, clustered, and land in the top 10% of the Top500 list, Steve and company will be the ones laughing.

Let's see... the *bottom* of the Top500 list is currently a 116-CPU Cray T3E 1200, with a theoretical peak of about 139 GFlops... you'd only need enough Xserves to fill 1/4 of a rack to come up with that kind of power.

Okay, okay, I guess I want some too.

I wouldn't get one for my desktop. Now, AMD's hammer is another story... :)

Because they have versions of linux running in native mode for both the Hammer and Itanium, I would like to see a comparison of speed between the two. I know that the compilers for Hammer will be weak, and I know that the silicon is only A0, but hey, I expect great things from AMD. :)

• Overview of the design: note that it's 640 nodes with 8 processors each
• Earth Simulator entry/home page
• Jamstec -- Japan Marine Science and Technology Center: Looks pretty much like the USA's NOAA
• News on Top500 Supercomputer Sites where y'all should have been getting this kind of info from to begin with

What a monster! Should dominate the number one spot at Top500 for some time! :)

I should imagine a computer system of this power will have the US Government a little concerned.. I shouldnt imagine it will be long before they announce that they have designed a 'more powerful' system though :)