IBM Retakes Fastest Supercomputer Title

dshaw858 writes "BBC News reports that IBM has unveiled its new Blue Gene/L machine. The Blue Gene project already has two of the top ten supercomputers in the world. Big news for IBM! I wonder what great things they can calculate in just seconds now... maybe I should get a stronger PGP key."

Comments

Don't worry

[FiReaNGeL]

"IBM and its partners are currently exploring a growing list of applications including hydrodynamics, quantum chemistry, molecular dynamics, climate modeling and financial modeling."

So no PGP key cracking. At least officially.

I wonder how the Fold@Home total CPU power compare to this in terms of percentage?

Re:Don't worry

[metlin]

Financial Modelling _is_ a big thing. I've worked on modelling stock and economic data using game theory and various analytic methods -- it's not as simple as that.

There's a lot of patterns, and a hell lot of data processing to be do.

However, that said, financial data modelling is not something which I think can be cracked using brute-force power. Although there has been a lot of fundamental progress in terms of using OR and GT algorithms and the like, it hasn't really had that "big breakthrough" to fundamentally determine the basis of financial data and market behaviour, and perhaps we never will.

Ofcourse, as always hope springs eternal - but that would also make markets a whole lot deterministic and bring about some serious differences in the way business is done.

Re:Don't worry

[Tim C]

Good point.

What would you do - sink a few hundred million in building a supercomputer to crack some guy's PGP key, or kidnap him, hold a gun to his head and ask for the passphrase?

You'd build the computer if it was imperative that the guy not know you'd cracked his encryption, or if you wanted to do it on a large scale. If it's just one or two guys, and secrecy isn't necessarily an issue, there are other ways...

Re:Don't worry

[fatphil]

How can something that says "this machine couldn't brute force PGP keys." and "that means you might manage 1024 bit RSA cracking if you are determined" get moderated to +5?

You're gibbering, sir. You say one thing and then the opposite.

No-one "brute forces" PGP keys, that's not how you crack them. Exactly how you crack them depends on what the underlying algorithm is, it's either GNFS factoring or discrete logarithm, but _neither_ is brute force. So your first point is wrong.

With current algorithms, 1024 bits is completely out of reach. The algorithms are mostly "embarassingly parallel", and therefore there's little gain from a tighty-coupled supercomputer (except at the LA stage at the end, but that's a fraction of the total workload). So this machine is no greater than the sum of its parts (i.e. several thousand high speed processors). Such a setup cannot crack 1024-bit keys. So your second point is wrong.

FP.

What about SGI?

[enigma32]

I recently read that SGI was to be claiming the world's most powerful supercomputer record from the Earth Simluator...

Does this mean that IBM leapfrogged SGI or does this mean that the SGI machine (to be built for NASA) wasn't all that exciting?

http://www.sgi.com/features/2004/oct/columbia/

Re:What about SGI?

[Anonymous Coward]

There will always be a competitor claiming the fastest supercomputer. It's always a game of leapfrog.

The rankings used in supercomputing measure very specific benchmarks and have very specific deadlines.

In this case, SGI has a computer reportedly faster than Blue Gene/L, but it is neither 1) in production by the deadline nor 2) independently verified results.

Sure, they might do this, but by then the next supercomputer will shame SGI's new baby. Like I said, leapfrog.

Rankings are all about a fairly arbitrary snapshot in time.

How 'bout

[iminplaya]

I wonder what great things they can calculate in just seconds now...

How 'bout this? 1,000,000! It tatkes pretty long on my P3.

Chaos Theory...

[oneiron]

They really need to get these things crackin on chaos theory... How many inhabited planets equals one amino acid chain? What are our odds of hitting the protein jackpot? You know?

No key cracking

[acidblood]

I recall reading on the RealWorldTech forums that these are highly specialized machines and particularly geared to floating point computation. As integer factorization, index calculus computation for discrete logarithm cracking, Pollard rho attacks for computing elliptic curve discrete logarithms, etc. are integer algorithms, crypto should be safe from this particular beast.

And before anyone asks about symmetric/secret-key cryptosystems and hash functions, recall that these are also based on integer operations, so they're safe from the BlueGene as well.

1000000! in hex

[3770]

I did that in hex on a 486DX266 back in the day. It took approximately a month.

I did it in hex because it was easier to write an efficient algorithm.

And then I decided to write a program which would convert that huge resulting hex number to decimal.

Only, that is when I realized that it would take more computational power to convert that number to decimal from hex, than to start from scratch and do it in decimal "natively".

Speed of the computer?

[Avuton Olrich]

Well, I'm confused, how can they figure out the speed so easily, when it's so hard to test the difference in speed between x86 AMDs & Intels? The other computers aren't faster at some things? Is it some special bench?

that we know of......

[Anonymous Coward]

I would rather get a list of top secret fatest computers in the world.

Is it needed?

[Ceriel Nosforit]

The final machine will help scientists work out the safety, security and reliability requirements for the US's nuclear weapons stockpile, without the need for underground nuclear testing.

Could someone explain to me why this task requires such a monster of a machine? And how can one address (as in write code for) the numerous unknowable factors that seems to be included in the problem that is to be solved? The definition just seems to be too abstract to be an actual solvable problem, and if it is solvable it would require an immense human resource contribution for the code it is to run. Wouldn't it be simpler to just stick those people into a room and not let them out until they've solved the problem?

I've long wondered who comes up with the code they run on these 'pooters. Anyone who can offer some insight on the usual complexity of the code that is run/problems that are solved?

Re:This begs the question:

[whataboutMike]

As mentioned in this article http://games.slashdot.org/article.pl?sid=04/01/06/ 0511220&tid=127&tid=208&tid=10 computers have no problem beating humans at chess. What would really make news is if computers would start beating humans at Go. Then again, Go is much less about brute force and deep searching; and more about pattern recognition. Something that humans seen to have a monopoly on.

Re:Quantum cracking algorithms?

[EsbenMoseHansen]

Indeed. Not so long ago I heard that a team had succesfully factorized 12 (into 4*3) using a quantum computer. (It was a 7 qubit 'puter) :)

Quantum computer has a way to go, even by paranoid standards.

Climate modeling

[LarsWestergren]

Ok, so of the three fastest computers in the world, one is almost exclusively dedicated to environmental climate models, and the other two have it as part of their tasks.

Perhaps this could bury the arguments on Slashdot that there is no hard data or serious research about global warming.

Re:Does Moore principle apply to quantum computing

[EsbenMoseHansen]

Currently, it isn't even a chip (or at least, last I heard). It was (a lot of) molekyles with 7 "mutable" spots (I think it was rotation). The state was read using NMR spectroscopy).

It is about as close to a chip as a printing press to a photocopier ;-)