Cray to make 20 petaFLOPS XK6 HPC for DOE ORNL, RSN

By Richi Jennings (@richi ) - October 13, 2011.

Cray XK6

Cray is partnering with AMD and Nvidia to build the world's fastest supercomputer. Named Titan, the XK6-architecture high-performance computer combines AMD CPUs with Nvidia GPUs, for the Department of Energy at the Oak Ridge National Laboratory. In IT Blogwatch, bloggers dream of FLOPS.
Your humble blogwatcher curated these bloggy bits for your entertainment. Not to mention: Dennis Ritchie, 1941-2011. Rest in peace...
Patrick Thibodeau reports:

This new system, called Titan, is due to be completed by 2013. ... Titan will be built using GPU and CPU processors. ... the Oak Ridge system may see performance "well north of 20 petaflops" if...built out to its capability. ... A 20-petaflops system is also being built by IBM for the Lawrence Livermore National Laboratory.
Over the past several years, systems makers have begun turning to GPUs...[which] can boost the performance of simulations. ... Titan will be used...for "increasing the realism of nuclear simulations," and "improving...climate simulations." It will also be used to develop and understand "novel nanomaterials."   

Noel Randewich adds:

While traditional [CPUs]...are designed to make huge calculations very quickly, one after another...GPUs, excel at carrying out several small calculations at the same time. ... About 85 percent of Titan’s peak performance is expected to come from [GPUs].
Nvidia’s high-end Tesla GPUs...sell for around $2,000 each. ... When fully up and running, Titan...should be twice as fast as the world’s current top computer [the Japanese] K computer.   

And Timothy Prickett-Morgan has all the gory details:

Oak the US nuke lab most associated with open (rather than classified)
The Titan system, also known as the OLCF-3...will be based on the Cray XK6...[which] swaps out one of the Opteron processor sockets on the four-node XE system blade and puts in four of Nvidia's Tesla X2090 GPU[s]. ... XE6 [is] able to pass about 100 times the messages between nodes.
[T]he first thing that will be done is that [the existing] 18,688 two-socket nodes will be replaced with 4,672 four-socket...blade servers...[giving] 299,008 Opteron cores and...somewhere between 25 and 30 per cent more raw x86 flops. ... The machine will also have 600TB of main memory. ... [Later] Tesla X2090 GPUs will be about one-twentieth of the system nodes).
And yes, it will be able to play Crysis.   

Meanwhile, Larry Dignan has the news that Nvidia nvestors need:

So what does Nvidia’s win mean? Beyond a little chest thumping not much. [This] business remains a work in progress and the real payoff will be years away.   

And Finally...
Dennis Ritchie, 1941-2011. Rest in peace.
Don't miss out on IT Blogwatch:

Richi Jennings, your humble blogwatcher

Richi Jennings is an independent analyst/consultant, specializing in blogging, email, and security. He's the creator and main author of Computerworld's IT Blogwatch -- for which he has won American Society of Business Publication Editors and Jesse H. Neal awards on behalf of Computerworld. He also writes The Long View for IDG Enterprise. A cross-functional IT geek since 1985, you can follow him as @richi on Twitter, pretend to be richij's friend on Facebook, or just use good old email: You can also read Richi's full profile and disclosure of his industry affiliations.

Copyright © 2011 IDG Communications, Inc.

Shop Tech Products at Amazon