Slashdot Mirror
SGI & NASA Build World's Fastest Supercomputer
GarethSwan writes "SGI and NASA have just rolled-out the new world number one fastest supercomputer. Its performance test (LINPACK) result of 42.7 teraflops easily outclasses the previous mark set by Japan's Earth Simulator of 35.86 teraflops AND that set by IBM's new BlueGene/L experiment of 36.01 teraflops. What's even more awesome is that each of the 20 512-processor systems run a single Linux image, AND Columbia was installed in only 15 weeks. Imagine having your own 20-machine cluster?"
Let's see them predict the weather.....
...when they hit the "TURBO" button on the front of the boxes they'll really scream.
I have one of those... in a spare room!
Who cares about a 20 system cluster, I want a one 512 processor machine!
or 20, I'm not that picky
Just what I need to model my next H-bom... uhh... umm.... I mean render my next feature film. I call it "Kaboom."
The World Serires outcome!
Call me and my voicemail! 914-713-6795. (wow, I have the balls to post my voip number on
I bet gentoo wouldn't be such a b**ch to get running with all of that compiling power behind it :)
According to the article it got 42.7 teraflops using only 16 of the 20 nodes, so the performance is going to be even better.
The major question is what does NASA hope to accomplish with this new setup?
With all of the new private space industry, NASA has been set free to explore the further reaches of space. The question is, where will they go next?
Imagine a beowulf cluster of those...
In Soviet Russia, LINPACK simulates YOU.
All your nodes are belong to us.
...they were *almost* able to get Longhorn to boot.
If the same software is used, its not going to make weather predictions more accurate. Its just going to give them the wrong answer, faster.
20 Machines with 512 processors? I think of that more as 10420 machines, not just twenty. Impressive!
SGI & NASA now have developed a computer that will be able to run Longhorn.
Yes! I listen to NYC Speedcore and do math at 3AM. I suggest you try it too.
....A beowulf cluster of.... holy crap!
This page contains images of the NASA Altix system. After reading the article I was curious as to how much room 10K or so processors take up.
http://www.busyweather.com/
1) This was fully deployed in only 15 weeks.
(Link)
2) This number was using only 16 of the 20 systems, so a full benchmark should be larger too.
(link)
3) The storage attached holds 44 LoC's (link)
Your hair look like poop, Bob! - Wanker.
Free Flat Screen HERE!
...single node of these...
:-)
oh wait, sorry, Cray deja-vu
...a computer that can run Doom at 60 FPS.
Prof. Jack Dongarra of UTK is the keeper of the official list in the interim between the twice-yearly Top 500 lists:
http://www.netlib.org/benchmark/performance.pdf See page 54.
And here's the current top 20 as of 10/26/04...
Your kidding right?
Sure they have plenty of processing power if you aren't running complex simulations, but if you are doing any type of scientific simulation its not hard to design a simulation that can bring a super computer to its knees.
Too bad its not running windows. They could set a world record for fastest windows crash after install. Mine's only a few minutes, image twenty 512 systems!
Computer superclusters don't even have O-rings.
They don't carry schoolteachers.
They don't fly in the air.
This runs Linux, not Windows. It won't crash.
sigs, as if you care.
Just wanted to remind you of an earlier post on slashdot about NEC's SX-8 which has peak performance of 65 TFlops. Now, which one is the fastest?
Colombia? They're going to use it to predict cocaine production? Well, that certainly explains the Genesis crash...
Cool...something that won't slow to a crawl while playing Sims 2.
Wow, I didn't know the NewAdvancedSearchAgent had such an interest or budget for super computing. I'd think they'd be able to afford their own web server though instead of being parked at domainspa.com and having to fill their entire page with advertisments.
Try NASA.GOV.
Why does it take so long to build a super computer and why do they seem to be redesigned each time a new one is desired?
It's a little like how Canada's and France's nuclear power plant system are built around standardized power stations, cookie cutter if you will. The cost to reproduce a power plant is negligble compared to the initial design and implementation, so the reuse of designs makes the whole system really cheap. The drawback is that it stagnates the technology and the newest plants may not get the newest and best technology. Contrast this with the American system of designing each power plant with the latest and greatest technology. You get really great plants each time, of course, but the cost is astronomical and uneconomical.
So to, it seems with supercomputers. We never hear about how these things are thrown into mass production, only about how the latest one gets 10 more teraflops than the last and all the slashbots wonder how well Doom 3 runs on it or whether Longhorn will run at all in such an underpowered machine.
But each design of a supercomputer is a massive success of engineering skill. How much cheaper would it become if instead of redesigning the machines each time someone wants to feel more manly than the current speed champion, that the current design be rebuilt for a generation (in computer years)?
The point is that your first statement isn't entirely true. Sure, a popular computer *can* do everything, but how long it takes to do something is another matter. Simulation programs exist (for things such as human heart beats) that tie up hours of processing time on supercomputers, let alone on your personal "popular computer". Finally, I really don't think a lone supercomputer is going to raise your taxes significantly compared to, hrm... say a war?
from floppy peckers?
Predicting weather, analysis of objects in conditions where hundreds and thousands of variables are present, etc.
I remember my dad worked on the Grand Challenge project.
A single timestep took around an hour and took up around 60 nodes on a Origin 2000 system (I think thats what it was at the time). He did his processing at the MSC (Minnesota Supercomputing Institute). But with faster computers doing more calculations, research takes less time and money basically.
Sure they have plenty of processing power if you aren't running complex simulations, but if you are doing any type of scientific simulation its not hard to design a simulation that can bring a super computer to its knees.
Ok, I expected someone would say that, and that's fine. But isn't that exactly the scenerio that clustering technologies have been created to be used in?
I seriously have a hard time imagining what kind of problem could not be solved with a cluster of pentium fours, each with 4-5 cpus (for a total of approx 12-15 GHZ each).
It certainly can't be a very commonly occuring one.
Ermm, which NASA are we talking about again?
National Aeronautics and Space Administration
New Advanced Search Agent
Nothing disturbs me more than blind loyalism towards some unrealistic and over-idealistic notion of one's nationality.
The amazing thing about it is that it's built at a fraction of the cost/space/size as the Earth simulatior. If I remember correctly, I think they already have some of the systems in place for 36 teraflops. It's the same Blue Gene/L technology from IBM, just a larger scale.
RAEM (redundant array of expensive machines) just doesn't ring right - to close to REAM.
This issue is a bit more complicated than you think.
Yes what is the point? We all know the resulting answer is going to be 42.
There's also a dark horse in the supercomputer race; a cluster of low-end IBM servers using PPC970 chips that is in between the BlueGene/L prototype and the Earth Simulator. That pushes the last Alpha machine off the top 5 list, and gives Itanium and PowerPC each two spots in the top 5. It's amazing to see the Earth Simulator's dominance broken so thoroughly. After so long on top, in one list it goes from first to fourth, and it will drop at least two more spots in 2005.
Whoever corrects a mocker invites insult;
whoever rebukes a wicked man incurs abuse.
--Proverbs 9:7
Does anyone know how much this system cost? It would be interesting to see how good of a teraflop per million dollar ratio they achieved.
For example, I know the Virginia Tech cluster (1,100 Apple Xserve G5 dual 2.3Ghz boxes) cost just under $6 million, runs at a bit over 12 teraflops, so it gets a bit over 2 teraflops per million dollars.
Other high-ranking clusters would be interesting to evaluate in terms of teraflops per million dollars, if anyone knows any.
It was here on slashdot last week , IIRC. :)
Try to plot the future of the solar system, with +10k objects with +1 km diamter (guessing) for any significant length of time, factoring in merely the gravitational pull of all the objects upon one another. You'd be hard pressed to calculate that for on your little network there. The thing is, they are not even factoring in solar flares, etc, etc.
Or what about something that can predict solar flares, or even creat a reasonably working model of the sun? All the convection currents and magnetic field simulations would bring your system to it's knees.
There's quite a few resons why they need this much power, but, as you said, it's not exactly a large percentage. But then again, these things aren't all that common, either.
Sig
They were only using 16 of those 20 servers. With all 20 they were able to peak 61 teraflops. Check the article at CNET.
A Beowolf cluster of Beowolf clusters....
ARRRGGGHHHH PEOPLE'S HEADS ARE EXPLODING!!!
Now you know that there's some engineer with acces to this thing thinking how he can jump to the front of SETI@HOME.
..........FULL STOP.
Seriously, am I on candid camera?
Can it play Doom 3?
Emulating a Centris 650 running Mac OS X at 2.5 Ghz.
ZZ
This is very surprising. SGI has been waning for the last several years and the top spot on the supercomputer list has been static for two years waiting for someone to build something better than Earth at a reasonable price. For them to get 80% of the machines working in 15 weeks and get 42TFlops out of it is very impressive.
Congratulations to the remaining engineering team at Silcon Graphics!
Stock Market sims for one...
Yes, but does it run linux?
Im dreaming ofa big bndwdth, That can resist the
Seti@home is currently reporting 70.93 TeraFLOPs/sec. It would be Number One if the list were a bit more inclusive.
Ok, so we have Linux doing tens of teraflops in processing, FreeBSD doing tens of petabits in networking,
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Or, try to find a pattern in the stock market.
Just watch out for those pesky ants and Rabbi's
( Before you mod me troll or whatever, it's a reference from the movie Pi.. which strangely had very little to do with Pi )
Slashdot sucks
uhm... Well 2560 motherboards, 'cause their quad-cpu... Altix is the SGI C-bricks that used were built to house 4 IA64 cpu's per brick. otoh... no... really it really is 20 machines with 512 processors each, because the memory is globally shared (all processors have access to all the memory, albeit at different latency and performance: NUMA (Non Uniform Memory Access). and a single linux kernel is running on the whole thing.
Really, given the fact that most popular computers have enough processing power to handle anything, and the fact that clustering technology has evolved and is usable in case they aren't...what is the point in the "super computer"?
The super computer is a cluster (10k+ processors in 20 nodes).
Not all applications/computations scale by just adding computers to the cluster.
An example would be solving for z: x=84+19, y=5*3, z=x+y
The ultimate solution z is limited by the speed x & y can be solved. You can have an individual computer solve for x and another for y in parallel. But no matter how many more computers you add, none of them can solve z until x&y are solved first, and none of them would speed up the computation of x&y.
After a certain scale, you do not get benifits of parellel processing, so the only way to speed things up is to make each individual computer faster.
D6 63 0D 70 89 81 BB 8E 7B 7C 5F 5D 54 EA AB 73
According to the press release, 65 teraflops is only the predicted theoretical performance; it hasn't actually been built and tested in Real Life.
Just about all super computers do.
I feel so dirty now...
Lawrence Person (lawrencepersonh@gmailh.com (remove all "h"s to mail)
http://www.lawrenceperson.com/
Damn who would have thought? Only SGI could make that possible!
||| I still can't believe Parkay's not butter.
The System X cluster contained 1150 machines containing 2 CPUs each which equals 2300 CPUs in total. You were saying? Not to mention you are comparing an expensive Server CPU with a desktop/workstation CPU.
Why don't we wait for IBM to build a Power 4+ or Power 5 super cluster?
Jesus was a compassionate social conservative who called individuals to sin no more.
And with all that power they feel the need to .ZIP their .JPG images which actually shaved off an entire 4K on this single 848K file. Wow. I should have thought of that.
Columbia
Where can I get one of these machines, how much will it cost, and how do they score on 3DMark 2005?
Please flee in terror in an orderly manner.
...in who owns the worlds fastest. I would bet this is the worlds fastest *publically admitted-to* supercomputer. Now, let's remove one A and contemplate what NSA instead of NASA has running someplace...
Am I the only one who get frustrated when companies/corporations/god/evil geniuses say that they're rolling out some really honking hardware, yet never shows all of us drooling geeks some really slick and sexy photos of it?
Man, that bites. I wanted some juicy photos of a Cray (or whatever hardware they're using)
(yes, this whole post is rather sarcastic. Go ahead and mod me down, see if I care)
I enjoy large posteriors and I cannot prevaricate.
NASA and computers? Someone please double check to see if they're using metric measurements instead of imperial, otherwise it could suddenly disappear without a trace. :)
The friendliest digital photography forums on the net!
Heck even a Centris 650 can do everything. Of course we are still waiting for boot to complete.
D6 63 0D 70 89 81 BB 8E 7B 7C 5F 5D 54 EA AB 73
Sorry, but they'll have to loan me the machine for a few weeks, so I can verify their results.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Does it run CherryOS?
This is all very nice, but the fastest supercomputer in the world is owned by Google... It has been known that it has "thousands of pc's", possible over 10,000 (50,000 is a reasonable number)...
"The thought of building a 10,240-processor system in a little over three months was something, especially since we still had to meet our normal manufacturing pace," said Dick Harkness, vice president of Manufacturing Operations, SGI.
:)!
Wonder What SGI's NORMAL manufacturing pace is
--> Your Wisecrack Here
They're saying the civilian computer will be slightly over three times faster than the military nuke sim. What's the betting that the US DoD's next step in Enduring Freedom is to liberate the Linux cluster?
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
yes but what would the Licenceing cost...s p
http://www.pcmag.com/article2/0,4149,1040410,00.a
I could not find pricing for datacenter...
enterprise version is $3,999 for 8cpu's.. so if we
do some math from that...
20-512 CPU Server Linceses
512 / 8 = 64
64 * $3,999 = $ 255,936
$255,936 * 20 = $ 5,118,720 (if all my math is good)
I wonder what percentage of total cost that would be....
yes i know datacenter spec's show only up to 64cpu's....
even if you half that amount it's a good chunk-o-change
Inevitably someone will say "we can finally predict the weather..." and in true Futurama Farnsworth fashion I say PSHAW! We don't even know how to properly frame the QUESTION of how to predict the weather, much less get closer to an "Answer" like "The hurricane will hit EXACTLY here, at EXACTLY this time. Only the people on these specific streets are boned."
Still, I bet I could get like 1 billion FPS on UT2004 at 3600x4800!
Seriously though, I want to see small improvements. Better, easier to grasp programming languages. More critical thinking skills taught in schools. And a cluster like this dedicated to uber-porn. I'm talking full frame, Hi Def, ggg stuff. (did I type that last part out loud?)
...But I digress. TREMBLE PUNY HUMANS!ONE DAY MY SPECIES WILL DESTROY YOU ALL!
Just great. Another machine that can do LinPack quickly. However, the odds of it working efficiently on scientific problems of interest to NASA is quite low.
This has to be a fake. All the analysts say that Linux doesn't scale, and you know how brilliant they are.
The most powerful computer ever built is the Planet Earth, and that uses MULTICS.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Get nearly as much press as the VT cluster, since it doesn't run on Apple hardware. Will Wired (aka Apple Corporate PR Monthly) even bother writing about it?
Who ever think to hear that one?
Look ma I made a funny on Slashdot!
...rubbing his hands whilst sitting in a dark corner amongst an ever-dwindling pile of Microsoft-donated cash, salivating at this.
"512 processors, 20 machines, $699 per processor. All that intellectual property, yes! No free lunch no, Linux mine, MIIIINE, BWAAAAHAHAHAHA!!!"
*dials*
"Hello, NASA? About that $7,157,760 you owe me...
I'm sorry, where do you want me to jump?"
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
It doesn't run OS X so it is, by definition, not cool.
"So, again, what is the point, exactly?"
To boldly go...
ohh bad joke.. yeah I get it but dude poor taste!
!!!BREAKING NEWS!!! The results are in: 42.
...that it still gets only 60fps in doom 3.
All those teraflops and still no aliens? How many freaking teraflops do we need? Come on, folks, I just want one goddamned spaceman!!
http://www.aggregate.org/
Just to be exact, you are BOTH wrong.
According to the numbers submitted this was...
42.707 GFlop/s for a 8064 processors...
So unless someone else is hiding a supercomputer somewhere.
1. NASA Project Columbia*
2. IBM BlueGene/L DD2 Prototype cluster*
3. Earth Simulator
4. IBM eServer Blade Center JS20+*
5. Intel Itanium2 Tiger4*
6. ASCI Q AlphaServer EV-68
7. Apple XServe platform
* - These systems are listed multiple times with different CPU's number on the current Highly Parallel Computing list (10/26/04) which may indication they are still being built up.
10240 Itanium 2 processors They're probably using all the Itaniums ever manufactured...
Ok, it's running one single Linux image. Is there any way we can download the ISO from somewhere, just for education purpose? I'd like to know what components they are running to make it scale like this.
Specs, versions, anyone?
The most amazing part of this development is that the fastest computer in the world runs Linux . All these TFLOPS increases are really evolutionary, incremental. That the OS is the popular, yet largely underground open source kernel is very encouraging for NASA, SGI, Linux, Linux developers and users, OSS, and nerds in general. Congratulations, team!
--
make install -not war
Curiously enough, we were talking about the future of computing at lunch today.
There was a time when different computers ran on different processors, and supported different OSes. Now what's happening? Itanic and Opteron running Linux seem to be the only growth players in the market; and the supercomputer world is completely dominated by throwing more processors together. Is there no room for substantial architectural changes? Have we hit the merging point of different designs?
Just some questions. Although it's not easy, I'm less excited by a supercomputer with 10k processors than I would be by one containing as few as 64.
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
Yes of course the answer is 42. This computer was built to find out what the question is.
Wouldnt you know it my mod pts just expired. Dude that was very poor taste
...oh never mind.
CBV@#$)(*?>M
free ipod and free gmail!
Weather prediction, it turns out, is *not at all* like playing chess. Chess is a deterministic linear process operating on rigid, unchanging rules. There is always a "best move" for every board state, which a sufficiently fast and capacious database could search for. Weather is chaotic, a nonlinear process. It feeds back its state into its rules, in that some processes increase the sensitivity to change of other simultaneous processes. Chaos cannot be merely "solved", like a linear equation; it must be simulated and iterated through its successive states to identify more states.
Of course, we're just getting started with chaos dynamics. We might find chaotic mathematical shortcuts, just like we found algebra to master counting. And studying weather simulation is a great way to do so. Lorenz first formally specified chaos math by modeling weather. While we're improving our modeling techniques to better cope with the weather on which we depend, we'll be sharpening our math tools. Weather applications are therefore some of the most productive apps for these new machines, now that they're fast enough to model real systems, giving results predicting not only weather, but also the future of mathematics.
--
make install -not war
I bet it has the latest version of DirectX installed ;-)
All your Sybase are belong to us.
It was great. I needed to build the kernel so I typed
# make -j 10534 bzImag
and even before I could hit the e and enter, it was done.
I was gonna build X but on this box the possible outcomes of "build World" scared me!
Yeah, I came up with the same list. It looks like Power/PPC and Itanium pretty much own the top 10. Interesting, because Opteron should be pretty close to PPC 970 on price/performance basis right now, but evidently no one has tried to pull a Virginia Tech with an Opteron cluster.
Still, it's quite exciting to see some really fast machines crop up after Earth Simulator sitting at the top for so long. For comparison, ASCI Q was 2nd and VT's System X was 3rd last year. ASCI Q stayed constant, but System X upgraded by ~20% and it's now 7th!
Wow.
But I'm sure in 6 months someone will go build a 10,000-CPU PPC or AMD-whatever cluster that'll eat this one for breakfast and still be hungry enough to eat the team that came up with it.
Village idiot in some extremely smart villages.
Scientists: is there God?
Computer: there is now!
"Its performance test (LINPACK) result of 42.7 teraflops easily outclasses the previous mark set by Japan's Earth Simulator of 35.86 teraflops"
Yes, but what is that in bogomips?
Do they really use the machine for real world application(s)?
or can I borrow it to crack password (or even run distributed.net client, hmm...).
Better yet, let's try Doom on it.
Kernel compile
.- -
make -j512
.
Now we see why they downscaled the images - they're all a bit blurry. Couldn't they find anyone with a steady hand?
I guess a project of this size leaves all the geeks too excited!
DROS - Open-Source Robot Software
Keep "uh"ing, it makes you look smarter.
Look at the link. Look at #1, look at #2. Then look at them again.
If you dont see what ac was referring to, kill yourself and save us.
Anyway, this is good -- we will now be able to simulate burning spacemen and such catastrophes in advance, and figure out the best way to cover up such mistakes in design and vehicle assembly.
I suggest you read Slashdot
steve
Oh, you're not stuck, you're just unable to let go of the onion rings.
Sorry, but I can't imagine how it be to have my own 20-machine cluster until I can visualize how large just ONE of them is. How big? the size of a fridge? a small closet?
For your example it is not really how fast the clock is but rather how fast the adder is inside. It is important to ensure that the adder is able to finish it's job in one clock cycle, but the adder is not controlled by the clock. You could build a full look ahead carry adder giving you almost instantaneous calculations and theoretically give that data to your cluster.
Super Computers on Slashdot are like murders on the local news: There's another one every day - to the point where they've become just about meaningless. And come to think about it why are they even news? They seem like they're just filler when there's nothing else to report
-P
Why have ONE conviction when you can have TWO?
"I seriously have a hard time imagining what kind of problem could not be solved with a cluster of pentium fours, each with 4-5 cpus (for a total of approx 12-15 GHZ each)."
Go run Gentoo and your opinion will change quickly.
http://saveie6.com/
Check out http://www.sgi.com/products/storage/ for some more info about the storage they are using. For those that don't want to wander around the site, there is a link under the picture of the storage array that says "Watch a Video" and it gives an overview of the technology that SGI uses in their storage solution.
They use tape storage from Storage Tek like this one
And harddrive storage from Engenio (formally LSI Logic Storage Systems) like this.
Its not what it is, its something else.
...rolling in from the west, bringing with it occasional Beowulf cluster jokes, resulting in a good chance of some "In Soviet Russia" cracks scattered throughout the region for the next few days...
but you picked a wrong example.
As it is, with 3 variables you can probably do a bunch of completely different things at the same time and with something like 30000 variables, it can be parallelized very efficiently.
But it took two weeks to load solitaire. They're giving it three days to shuffle the deck before they install BeOS.
- usability
- marketing
How is this the fastest supercomputer in the world at "only" 42.7 teraflops when NEC has already launched a 65 teraflops supercomputer?
http://www.nec.co.jp/press/en/0410/2001.html
"It is our choices, Harry, that show what we truly are, far more than our abilities." -- Prof. Dumbledore
Now maybe they can come up with plans to retrieve space probes that don't involve the use of stunt pilots.
The answer here is "complexity". I do some scientific computing (have done chemistry, then materials science, now doing photonic devices) and there's always more you want to be able to consider. Of course, the best I've used is an 8-processor SGI machine (although that one was a bit old - I think the 2-processor opteron system I'm using now is actually better). But especially with the materials studies, ideally we wanted to do everything with full quantum-mechanical calculations. which turns into gigantic matrices, even for a system of 100 atoms or so. And even then we put strict limits on what orbitals we consider and all that good stuff.
Slightly more concrete example - right now with my photonics simulations (finite element) on my dual-opteron rig the max I can handle is about 180,000 elements (which means a (4*180000)x(4*180000) matrix with complex elements needs to be diagonalized, among other things), and it takes about half an hour for a standing-wave calculation. To do any time propogation, repeat same calculation in picosecond increments. And with the gridding I can do, for a 100 micron disc resonator in 2-D I have to use light at about 40 microns. To go to the 320nm wavelength these resonators are operating at, I'd need roughly 2 orders of magnitude more memory. There's also the time factor to be considered. As with any design process, one must iterate. Tweak a little here, run the program, rinse, repeat. How long are you willing to spend in this process before you feel something is "good enough"? The faster the computer spits the answer out, the more things you can try, and the more you can think things over and hopefully make it better.
And this is a single component in what can be a fairly complex integrated-photonics chip. [And might I mention again I've been working in 2-D this entire time instead of doing a full 3-D simulation?] You give me the computational power and I'll use it. And I'm an experimentalist doing fairly basic research who just wants to check some stuff in the computer before sinking a lot of time and effort into fabricating a test device.
On the other hand, I actually don't want to have one of the T100 supercomputers in our lab. That would mean I'd be spending all day writing code and designing complex simulations instead of in the lab getting my hands dirty.
And as for the commonality of problems requiring such computational power, I think almost any sort of simulation can easily use it. Consider more terms (everything I've done to date is horribly linearized - let's see some more terms in the Taylor expansion) to account for nonlinear behavior, grid things up finer to get more accurate results, consider more possibilities when dealing with chaotic behavior... I would hope any good scientist would find the possibilties endless.
when they upgrade the whole lot to Montecito processors next year. LOOK OUT.
okay overated? I realize its not that funny but to mod me down from 2??? I was legitimately pointing out that the poster said that the 20 machine cluster was the story, I thought he was mistaken...
oh well
Planet VR.
I'll be impressed when I see it run Doom 3 smoothly @ 1600X1200, 4XFSAA, 8XAnisotropy.
Halitosis - (n.) Halle Berry's Camel Toe.
Got Linux?
GJC
Gregory Casamento
## Chief Maintainer for GNUstep
1. I bet CIA has something in order of 10-100x more powerfull, I mean if you can afford to wire up 5 full office floors of computers, say 20*512 * 5 per floor * 5 , thats a hell lot more. CIA can afford to spend 200m on it, and have 10 super clusters of 1000 tf each.
2. I bet the CIA also can change the weather, go read HARP etc... if the russians can do it in the 80s then the CIA can do anything.
Liberty freedom are no1, not dicks in suits.
As I'm RTFA...
"For instance, on NASA's previous supercomputers, simulations showing five years worth of changes in ocean temperatures and sea levels were taking a year to model. But using a single SGI Altix system, scientists can simulate decades of ocean circulation in just days, while producing simulations in greater detail than ever before. And the time required to assess flight characteristics of an aircraft design, which involves thousands of complex calculations, dropped from years to a single day."
Being the NASA fanboy I am, I have to wonder if this massive computational step up doesn't share a large number of similiarities between the punch card computing age versus the modern programming age. Because of a quantum leap or five in time reduction for the bottleneck in computation time, more experiments, more radical theories, more wild stuff can be done because it won't be tying up the supercomputer for the next year... just the week. For all the wild science articles that make us salivate here... is this not the harbinger of a new era?
Isnt http://science.slashdot.org/article.pl?sid=04/10/2 0/1727255&tid=137&tid=136&tid=14 faster or? Im probably missing something obvious as always but yeah...
you can always get 2thousand latest CPUs, stick them in a cluster and have a super computer ... but unless we move forward towards faster individual CPU, nothing is awe inspiring ...
just my 2 bytes
With all this combined computing power in the world it won't be long before we finally find the last prime number.
-- thinkyhead software and media
"Imagine having your own 20-machine cluster?"
NO! Let me tell you what I imagine. I imagine not being underemployed and making enough that I can take care of my needs AND being able to lust after all these geek gadgets. THAT'S what I imagine.
This must double the number of Itanium processors they've sold this year!
After reading the article I was curious as to how much room 10K or so processors take up.
The room is about 100 x 120 feet, and the Columbia cluster takes up about 85% of the floor (this is estimated!). What is more interesting is the temperature ranges within the room. There are different cooling techniques for the Altix 3000's and Vortex's. Within a few feet of walking, there can be a 15 degree change in temperature. I heard someone may do visualization of air flow dynamics in there, which is definately cool.
For those interested, we will be doing plenty of informative demos on Columbia at the SuperComputing 2004 conference (Pittsburgh, PA) coming up in a couple weeks, so stop by the booth!
I installed Windows XP* on my 1ghz p3** in just 2.5*** hours!
And they call it a super computer!
* Just kidding, SuSE 9.0
** Kidding again, 2800+
*** I can't remember the install time.
#hostfile 0.0.0.0 primidi.com 0.0.0.0 www.primidi.com 0.0.0.0 radio.weblogs.com
Imagine having your own 20-machine cluster?
With 512 Processors, I'd be happy with just 1.
THAT would be a Beowulf cluster. (not really, but it's a joke)
LK
"Hi. This is my friend, Jack Shit, and you don't know him." - Lord Kano
Not quads as such. A single C brick has two dual CPU nodes.
You guys are failing to ask the most IMPORTANT question. WILL IT RUN DUKE NUKEM: FOREVER?!
seriously though, does anyone know of a "computer" above this number, active allready, and one above 360TFLop/greater performance other than the following(which isn't complete yet?): http://news.zdnet.com/2100-9584_22-966312.html ? actually though the whole point of this post is just to point out that the grandparent is incorrect, whereas the parent is not.
In case anyone is skeptical about the parent post. In fact now that I think about it, seriously, guys; the linux number isn't very high. you linux users out there should contribute. join project wool clock
GENERATION 26: The first time you see this, copy it into your sig on any forum and add 1 to the generation.
Imagine how man FPS in quake3 this machine could pump out... BTW, where's all the pictures on this beuty?
God,root what's the difference? I read slashdot, there for I errr... am stupid?
Quote from theregister.co.uk Published Wednesday 20th October 2004 19:19 GMT NEC has trumped US computer makers once again by announcing a new supercomputer that destroys previous performance marks. The "SX-8" is a follow on to NEC's Earth Simulator, which held the supercomputing top spot for some time. The new machine can reach a peak performance of 65 teraflops (trillion floating point operations per second). The Earth Simulator topped out at 40 teraflops.
But it would take to long to:
Run Windows Update for each box.
Remove Windows Messenger.
Cancel the window telling you to take a tour of XP.
Cancel the window telling you to get a passport.
Run the net connection wizard.
Reboot after installing updates.
etc....
(I'm not being totally serious, I know you can deploy ghost images etc..)
$45 million/10240 on Intel box = $4400 per processor
$6 million/2200 on X-Serve box = $2730 per processor
For shit and giggles lets say we quadruple the VT "Big Mac" to equal the Tflops of the Intel box.
That's only $24 million!!! STUPID NASA!!!
In Soviet Russia you suck apple!
Cisco's MDS switches are beating the crap out of Brocade in this space, now.
Well if it is the fastest then what is this.. http://www.hpce.nec.com/445+M5043fbf9d60.0.html/ the NEC supercomputer does 65TFLOPS..
No, no, it's not funny. Really - how long would it take with all the FS distribution and stuff? I bet no less than 30 seconds.
Just when I thought my data was safe from the aliens (sigh).
I would have thought knowing the 'correct' move your oponent will make might force you to chose another path, to which there will be a correct answer!
Argh it hurts my head.
or is that too obscure?
But since I'm right and you're wrong, we can take this discussion up again in a few months when Red Storm comes online, with slightly over 10000 AMD Opteron processors.
I'm posting anonomously because I have no desire to be involved in random flaming. If you can come up with a decent argument to support your claims, and find evidence to back them up, I will do the same.
I'd be curious as to what the cost/performance ratio of this is. Obviously, it's chock full of vitamin fast, but how much did it cost? As a Mac user, I tend to follow Vigrinia Tech's G5 cluster at least a little bit, and what always impressed me was that it only cost $5.8 million (and that's AFTER the upgrade to XServes). Not too shabby for such an impressive cluster.
Again, not to diminish this new supercomputer, but I'm curious as to how the price/performance ratios stand up.
"Reality is merely an illusion, albeit a very persistent one " -Albert Einstein
Built in Chippewa Falls, WI ( http://www.chippewachamber.org/ ) - home of the fastest computers in the world and Leinekugel beer http://www.leinie.com/
We believe there's a connection.
Speed record increases of 20% or less.
Same old cluster technology. We are always reading about this or that fancy new technology, but none of that scales up into commercially viable supercomputers.
Imagine having your own 20-machine
;)
Thats like trying to imagine what two billion one dollar bills looks like...it ain't happening, now what will happen is my heart would stop if I had one of these.
Actually, with the way computer prices drop and the speed at which newer versions come out, I expect this computer to be out-done by Dell in about three weeks
I mod down so you can mod up. Your welcome.
> know how to properly frame the QUESTION of how to predict the weather
...
We'll process all the data for 7 million years to get something like 42 , and we'll have to build a BIGGER computer to get the QUESTION
And it'll be like "Think of a number. any number ?" (which I believe to be the QUESTION).
Quidquid latine dictum sit, altum videtur
I think Columbia should be considered "experimental", and definitely not a production machine. It's only fast if you are running benchmarks and can get access to all the processors (which eliminates all of us NASA researchers)!
Everyone who has tried Columbia has been really under whelmed. Long queue waits.... They have only allocated a very small portion of the processors to Aeronautics, and there are too many users in line. In addition, they make changes to the system fairly often, and we end up needing to recompile and troubleshoot codes every damn time. I'll take my puny 40-node local cluster any day!!
The fourth generation, ASCI Q at Los Alamos, is designed ultimately to reach 30 teraflops. However, it's still under construction and so far exists as two 7.7-teraflop parts.
Someone put a fork in it.
Know your pads. One time pad: good for cryptography. Two timing pad: where to take your mistress.
In that case they are out of their league, you need to create a whole planet for that.
Let's see, if VATech can hit 12TF with G5s at a cost of 5mil, then they could get to 60TF at around 25 mil, half the cost of NASA's Titanic cluster (and with half the chip count). Way to go NASA.
If I didn't have absolutely NOTHING to do, I wouldn't be here.
Let's see it predict Columbia crashes http://www.newsday.com/media/photo/2003-02/6459841 .jpg
Yeah, Irix is cool. The trouble is, SGI resisted the hardware change. They kept on betting on the wrong horse (MIPS). At the same time, Intel and AMD were beating the crap out of all other CPUs.
So, it's not the OS change they resisted, it's the hardware.
Fortunately, they made the right decision eventually.
Sun, can hold up to 106 processers in its Sunfire 15K product, or 72 dual-core processors in the E25K. SGI's Origin systems are equally large I believe.
Actually, SGI makes 1024-CPU systems using the older MIPS/Irix architecture (the Origin and Onyx platforms), and 512-CPU systems using the newer Intel/Linux (the Altix machines). And that's what they make on a regular basis. Experimental systems are even bigger. Also, the Intel/Linux SGI systems are still very young and likely to grow a lot more.
That's one order of magnitude above Sun.
SGI is leading the single-OS-image market in terms of the size of systems.
Also... am I the only one who feels that this is yet another post on world record supercomputing in a rather short space of time? IT seems like it is once every other week there's a new record, or report on a shot at one, or similar. Perhaps my memory is not so clear and I'm blowing it out of proportion - and I realize in this case the record has been seriously outstripped - but it still seems like a high number.
.
-shpoffo
In actual truth, faster machines allow to perform more iterations of the same algorithm in the same time. Each new iteration adds more precision to the prediction. So, faster computers do enable better predictions.
:-)
But yeah, nice joke.
Besides SGI and HP, is there anyone really standing behind Itanium?
HP is already making some moves away from Itanium.
Others have speculated that this is the "thin end of the wedge".
Maybe Intel will start selling their own brand of servers.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
That's a bunch of untrue FUD. Columbia stayed busy nearly the entire time benchmarking was being done. Users love it. The results are blowing people away. If you look at the time from when this sucker was started to when it was useful, you will find no cluster ready in so short a period of time.
Funny, I seem to remember a report of a NASA Space Super Columbia that crashed. Guess there was no Linux on board.
30secs has already been beat, at least on 2.4.x kernels.
I don't recall what FS was used.
Nothing to see here; Move along.
That's all coming out of RAM though - but with that much RAM, why not?
The reported record stands at about 4 seconds flat on a 32-way POWER4 machine, I think. This was at the start of the 2.5 series though, and there have been a crapload of scalability improvements since then.
You probably can't beat that by much because most of the time would be dominated by the final link phase, which is single threaded. A classic application of Ahmdal's law.
You must increase your exposure - I have used Sun 10K's with 64 CPU's in an image and we got 80% scaling, I've used 72 CPU images and and got 75% scaling and what was running- crappy Oracle parallel code. If you think 15K's, Superdome or Big IBM's max out at 8 CPU's you just haven't run in a large corporate data center. When you have 10's of thousands of processes these monsters scale well - with increasing overhead, of course, as they get bigger. Where you want to make the cutoff depends but it is way higher than 8 cpu's - I would say around 64 CPU's these days.
having it.. I just jizzed from all the excitement thinking about having this box! :)
YHBT. YHL. FOAD.