King of Supercomputers
 

The Sequoia supercomputer is a system built by IBM for the U.S. Department of Energy’s Lawrence Liver-more National Laboratory, in California, is the now the most powerful supercomputer on earth, according to rankings released today. It led thetop500.org list, which ranks the worlds supercomputers according to a standard software benchmark, delivering 16.32 petaflops (a thousand trillion [<?] floating point operations per second) using 1 572 864 processor cores. It marks the first time since November 2009 that a U.S. supercomputer has topped the charts.

The IBM machine made use of the company’s BlueGene/Q computing system, which features 18-core processors based on the PowerPC architecture. Overall, IBM systems had a good showing, accounting for 47.5 percent of the computing power in the top 500 list, easily outpacing it’s next nearest competitor Hewlett Packard.

Sequoia’s nearest competitor, Fujitsu’s K computer, has topped the charts during 2011. It managed 10.51 petaflops using 705 024 cores. It was followed by a U.S. system—the Mira supercomputer, another IBM machine, that pulled 8.1 petaflops with 786 432 cores.

European computers had a good showing, with two German machines and the first Italian top 10 system on the list, as well as France grabbing the number 9 spot with it’s homebrew Bull supercomputer.

Meanwhile, China’s Tianhe-1A took number five, and the Nebulae system, in Shenzhen, came in at number 10

But can it play Crysis?

I would imagine either the CIA/NSA or the US DOD has better super computers that they just don't talk about.

Wirelessly posted (Samsung Galaxy S: Mozilla/5.0 (Linux; U; Android 2.3.6; en-us; SGH-T959V Build/GINGERBREAD) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1)

Can you carry it on or must it be che ked? :D

Wirelessly posted (Samsung Galaxy S: Mozilla/5.0 (Linux; U; Android 2.3.6; en-us; SGH-T959V Build/GINGERBREAD) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1)

DoE also does lot of classified work on nuclear weapons and used supercomputers. There is no apriori reason to assume that DoD/NSA/CIA have a monopoly on operating super computers.

I actually met Watson at IBM's research lab the other day.. very impressive.. We are getting a BGQ.. But only the third fastest..

China will leapfrog us again soon..

No assumptions here, just suggesting that the worlds fasted computers are not publicly announced.

I have no idea what hardware they use but you'd think that the stuff used in the AUSCANNZUKUS (ECHELON) system would be totally state of the art and at least as good as the stuff they DO talk about.

Doubtful; they have the budget to build a computer optimized for any given task, and wouldn't need a big general-purpose supercomputer. If they do, the DoE can sell them time.

For cracking encryption, getting an Application-Specific Integrated Circuit (ASIC) designed and built pays off if you're going to be buying them in volume. For other military tasks, there's going to be a requirement that they distribute them, and at that point you have to decide if it all has to be in the same building to count as a supercomputer or not. Facebook and Google certainly have more processing power, but they don't call the network spanning their datacenters a "supercomputer" for some reason.

No they don't have a better supercomputer.

National Nuclear Security Administration is using this computer for nuclear bomb simulations.

In addition to national security applications, Sequoia will also be put to use study climate change, astronomy and the human genome.

Sequoia

Wirelessly posted (Samsung Galaxy S: Mozilla/5.0 (Linux; U; Android 2.3.6; en-us; SGH-T959V Build/GINGERBREAD) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1)

Finite elements calculations for CFD, complex structures, weather, etc require lots of calculations. If you took a portion of the atmosphere 20 km above the surface of the earth and divided into a trillion cells for your calculations, each cell would be a sqaure, one seventh of a mile. Accuracy of the calculations depend om the grid size. You have to make the grids smaller for better accuracy. Eaxh cell will be governed by diffetential equations and boundary conditions. The computational effort goes up as cube of the number of cells. A trillion cells model requires at least ten raised to the power thirty six calculations. By contrast deciphering a 128 bit code requires ten to tje thirty three calculations, if you have absolutely no clue. This effort is a thousand times smaller than a atrillion cell grid. If you reduce the cell size to a litlle smaller than half the original size, you increase the number of cells to ten trillion cells and the compuyational effort to ten to the 39th power, already a million times the effort of brealing a 128 bit code blind.
Your claim does not appear to be based on any reasoning, just a belief arisiing from the awe and respect for these agencies and their need for secrecy.
I can only speak for and of what I know and can reason. Hollywood may create a different impression.

ok, but I bet it can't run windows 8 either.

-David

You got me. Come to think of it, it won't fit in my carry on either.

This post is a bit muddled...

Kinda "mixing your metaphors" here; 20km above the earth but each cell being 1/7th mile (area, I presume, hence square miles). That notwithstanding, the math doesn't work. The earth's radius is about 6380 km (on average), add 20 and we get around 6400. The surface area of such a sphere is about 514,718,540km. Divide by a trillion (1,000,000,000,000) and you get a cell that's around .0005 km in area. That's a cell about .0002 square miles, which is a whole lot less than 1/7th of a mile.

It's been a very long time since I last worked on GCMs but, in those days, the typical lat/lon resolution was around 2.5/2.75 degrees (or some such), which meant that each atmospheric level had about 90x72 cells. I think that nasa/noaa/ncar are looking at GCMs with grid resolutions of around 3-5km (but I've been away from that field for many, many years, so don't hold me to those numbers) which means that each level would contain about 20-50 million cells, nowhere near a trillion.

And why 20km? The difference in surface areas of a sphere 6380 vs 6400km is less than 1%.

What makes you think NSA is bothering with 128 bit encryption? Ask instead, for example, how much computing power you need to monitor a whole lot (ie, a nation's worth) of phone conversations (especially cell and sat) and apply word spotting software to pull out "interesting" conversations in real-time. If you think this isn't happening then I would suggest you go back and review the now public disclosures of Bush's (Bush the Lesser, not Bush the Elder) warrant-less wiretaps.

This comes off sounding a bit snide, like an ad hominem attack.

This is what I was talking about with the "Five Eyes"/AUSCANNZUKUS/ECHELON system. I'm no expert (in fact, VERY far from it) but I'd say the computing power required would have to be massive. Fair enough, it's a network not a supercomputer per se but it would still have to have incredible computing power.

Sorry about "square" I meant a cube. In any case, it's should be obvious from the context that it should be a cube.
I gave an example of how one can perform back of the envelope calculations, without a calculator, to estimate the level of effort required. It illustrates a principle.

Yes a bit muddled. I should not have posted it from my phone. A proper keyboard would have been better.
Your presumption is totally off the mark. You did not have to presume anything to talk about area. I had already made the mistake of calling it a quare. I would have commended you had you presumption had led you to talk about volume.

I used mile to make my point about size of the cell, because most people here understand miles better.
This comment reminds me of your last comment here about "an ad hominem attack". ;)


Math does work. Once you get out of the mindset that what's stated is wrong, apparently misled by my use of "square" in stead of cube.
Both uses, one metre cube/square and one square/cubic metre are acceptable but mean different things. So 1/7th mile cube as well as 1/7th mile are absolutely correct.

It's astounding that someone who worked with GCMS would not consider the possibility that I might have mean cube.
I
And how accurate was it? At what level did you work? I mean what level of detail were you seeking? I have not made, and I am not making any definitive statement about what cell size is adequate.

I did not make any claim about being a GCM expert, just gave an example of how anyone, can do some simple calculations to estimate level of effort. If you worked with GCMS, which I have not, but know from simple commonsense, that those calculations are reiterative. I did not even give any details about what kind of calculation is done for each cell? I used one arithmetic operations per element, which A GCM person would know is ridiculously low, show that we are already to a very very large number of operations.

Again, I have not presented a GCM model. It's a simple illustration of estimating computational effort. A simple division of a portion of the atmosphere into cubic cells does not make a GCM. Any GCM expert would no that. I think you started with the assumption that my GCM model, which it is not, is wrong.

If I knew what NSA was using, I would not be posting it here. :D

I dodn't know, but I don;t think monitoring conversations, is a computing intensive operation, and computing power alone would not do it. Quality of sensors and their locations is just as important.


I disagree. The statement cited no reasons. It was a belief.


You took my off the cough remarks designed to show how one can estimate requirements for a computational efforts as an instruction on GSM models and an explanation of NSA functions.. I picked two simple examples. I am not a GCM expert, nor do I work for CIA/NSA etc. Are you? If I were, I would not be posting here.
Here is how I did a cell calculations. Again. DON"T challenge my assumptions because they are not directives for a GCM model.
I assumed, that most of the mass of the atmosphere is within the 20 Km layer above the surface of the earth. Hence only that layer interacts with the sun and the universe. I calculated voume of that layer, which is a spherical shell of thickness 20Km. I Chose an arbitrary number of cells, a million/billion/trillion to estimate the size of each cell. All other details were ignored as they are not necessary for my point.
For NSA, I picked a simple task that most people can identify with. Breaking a code, with brute force, trying every possible combination.
Research in a field always requires a lot more effort and expense, than running field applications based on that research.

If you were slighted by my comment that NSA probably does not have the most powerful computers, I apologize. Rest assured that I don't value a person by the size of tools they have. Nor would my opinion deprive NSA of asking for more powerful computers. It was an opinion.

http://www.washingtonpost.com/blogs/...RgsV_blog.html

this is probably a very very large computer. the photo in the post showed the 23 fans on the roof. the new photo shows a clean box. if you go to 8540 Electric Avenue, vienna va, and look to the west, the google pic will show the roof, and all the a/c's. these idiots put a building that puts out 90-100db of noise in this secret location???

this is not a real special location, getting in and out at rush hour is pure hell.(i lived in 8544 when electric ave was a dirt road)

That doesn't really look any different from any other co-location building. Of course, most of those are nowhere near residential areas. It could just be a normal data center, not a "super computer".

Here is another that hosts a lot of Microsoft, Comcast and other large firms:

https://maps.google.com/maps?q=o'har...e+airport&z=19

Maybe when the FBI says they're putting a detainee on ice they're really putting them on ice!:D

Discernible, perhaps, but not obvious. I had no idea you were talking about a spherical shell 20km thick and, quite honestly, wasn't planning to go thru a lot of "what if's" to try to determine how you came up with your numbers. Clarity is the responsibility of the poster.

As to the remainder of your response, I could go thru it and refute your statements and assumptions point by point, but that's not useful.

Instead let me just say that you have very grossly overstated the computational complexity of global weather modeling and grossly understated the computational needs of the IC (Intelligence Community, which includes DoD/NSA/CIA/etc), You then used these bad assumptions to support a conclusion (determined a priori?) that is simply wrong. If this were done intentionally, it would fall into the realm of a "strawman argument", but I suspect it was just a matter of "uncareful" estimation.

Nothing wrong with "back of the envelop" or "orders of magnitude" estimation but you need to use valid assumptions and not just pull numbers and tasks out of thin air.