McClatchy: Competition heats up for world's fastest supercomputer

[babylonsister]

Competition heats up for world's fastest supercomputer
By Robert S. Boyd | McClatchy Newspapers

* Posted on Tuesday, September 25, 2007



WASHINGTON — In the next few weeks, engineers at Argonne National Laboratory, 25 miles outside Chicago, will install the first pieces of a machine that will have more than triple the speed of the world's fastest computer.

By next summer, it will be able to perform a quadrillion — that's 1,000 trillion or 1,000,000,000,000,000 — calculations per second. Its maker, IBM, says it would take a tower of laptop computers a mile and a half high to match its power.

This speed demon is called the Blue Gene/P. It's the successor to IBM'S Blue Gene/L, the current world champion. Blue Gene/L edged out a Japanese supercomputer, the Earth Simulator, for the top rank in 2004.

The latest machine in the Blue Gene series is "another step on the never-ending journey to apply more compute power to the problems at hand,'' said Dave Turek, IBM's vice president for supercomputing.

A supercomputer's blinding speed makes it possible to solve complex problems in science, engineering, the environment, industry, finance and national security from the atomic to the cosmic level, Turek said. It can model the activity of electrons in an atom, and simulate the birth and death of the universe.

more...

http://www.mcclatchydc.com/226/story/19843.html

[MADem]

1. Petaflops!!!!!!!

:rofl:

What an amusing word...and ya learn something new EVERY DAY!

[babylonsister]

2. See? I was hoping someone would appreciate this, because it's

all geek to me!

[MADem]

3. I thought it was great--can't imagine why more aren't interested!!!

In fact, I'll not just kick it with this post, I'll "R" it too!!!

[eppur_se_muova]

4. I'm starting to wonder ... when will we reach a point that more computational power ...

no longer brings any real benefits? Surely there has to be a point beyond which the problems requiring this much computing power would almost have to be contrived, as any realistic problem was already within reach.

I've done a fair amount of ab initio compuchem myself, so I have some idea how much of a difference extra firepower makes. But at some point you're tackling supersize jobs for the heck of it, and not because they lead to any really new insight. (I don't think we're there yet, but it already seems to me there's a tendency to set up problems in a more elaborate fashion than is really necessary, simply because the firepower is there.)

http://top500.org/