Friday, June 29, 2012

What's wrong with being number two?

Hopefully, not too much - since Japan's K (δΊ¬) computer, named for the Japanese character for 1016 (flops), had a rather brief stay at the top and has now been deposed to second. The Earth Simulator was number 1 for a remarkable 2 full years a decade ago.

The DPJ swept into power a few years ago with promises of scything through wasteful Govt spending, and an MP (now cabinet minister, but I don't think she was at the time) called Renho alarmed many scientists when she said (with direct reference to the massive cost of the K computer): "What's wrong with being the world's number two?"

Of course, after much huffing and puffing, the spending cuts - where they happened at all - were pretty minimal. In response to the ignominy of being overtaken, there are of course plans afoot for the next, even bigger, computer.

Meanwhile one of the Bayesians last week was talking about running massively parallel computations on graphics processors, available for about 50¢ per CPU, ie $250 for a board with 500 on. These aren't really much use for highly parallel high resolution numerical models, but there are a lot of things they can do.

6 comments:

William Connolley said...

Never mind: at 12 MW, your machine is far and away the most power-hungry on the list, so you're still #1 in a way.

James Annan said...

Well it's not as if we are short of power, especially in the east of the country, oh no not at all.

JohnMashey said...

Well, a supercomputer with a "Tofu" interconnect can't be all bad.

JohnMashey said...

Well, a supercomputer with a "Tofu" interconnect can't be all bad.

EliRabett said...

No doubt about it the Nvida GPUs are the coming thing. They will lead to re-writing of all sorts of scientific codes which has already started in molecular modeling and ab initio computation

Martin Vermeer said...

Yawn... a friend of a friend computed planetary orbits inside a PostScript(TM) printer -- back in the eighties :-)