« Sample Press Coverage for the 2006 Almaden Institute on Cognitive Computing | Main | Machina Speculatrix »

Fastest supercomputer to be built by IBM

"Computer giant IBM will build the world's most powerful supercomputer at a US government laboratory."

"The machine, codenamed Roadrunner, could be four times more potent than the current fastest machine, BlueGene/L, also built by IBM."

"The new computer is a "hybrid" design, using both conventional supercomputer processors and the new "cell" chip designed for Sony's PlayStation 3."

"Roadrunner will be installed at the Los Alamos National Laboratory, New Mexico."  

"The new machine will be able to achieve "petaflop speeds," said IBM."

See original article -->

I wonder what new avenues in computational neuroscience can be opened with such computational power.  On a personal note, I would love to play on this machine!

TrackBack

TrackBack URL for this entry:
http://p9.hostingprod.com/@modha.org/blog-mt/mt-tb.fcgi/8

Comments

Being an IBMer you may have more current information, but anyone else who is interested in Cell but doesn't know much might want to take a look at the following two PDFs.

First, a talk by Barry Bolding of IBM's Deep Computing group given at a parallel computing conference in 2005:
http://www.spscicomp.org/ScicomP11/Presentations/IBM/bolding-cell.pdf

Second, a relatively recent paper out of LBNL's Future Technologies Group on using Cell for HPC. No neuroscience focus, of course, but interesting nonetheless:
http://www.cs.berkeley.edu/~samw/projects/cell/CF06.pdf

(An article about the above can also be seen here:
http://www.hpcwire.com/hpc/671376.html )

While this all looks very encouraging, I don't imagine that Cell is easy to program for and it will take a while to make the tools and techniques get near peak performance. Last I heard people (non-IBM people, that is) were raving about autoparallelisation across the SPEs via the compilers, but that is hardly a trivial task as anyone using such features in the PGI or Intel compilers on PCs can tell you. I don't imagine the situation is terribly different with IBM's XL compilers, but I'd be happy to be proven wrong. :)

On a different note, I want a Roadrunner system, too!

- Brian

How would you compare the cognitive computing effort at IBM compared to ACT-R project at CMU or SOAR at UMich (originally also from CMU)?

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)