Flutterby™! : Gravity Grid - PS3's Clustered

Next unread comment / Catchup all unread comments User Account Info | Logout | XML/Pilot/etc versions | Long version (with comments) | Weblog archives | Site Map | | Browse Topics

Gravity Grid - PS3's Clustered

2008-02-28 02:40:22.020754+00 by meuon 3 comments

http://gravity.phy.umassd.edu/ps3.html - Claims a PS3 is worth 25 IBM Blue Gene Nodes because of the Cell processor, and has a 16 PS3 Cluster working on gravity math. - All beyond my understanding except the cost savings involved and how efficient the cell processor must be at the types of math he is doing.

Makes me wonder how good of a LAMP server one would make?

[ related topics: Bioinformatics Work, productivity and environment Mathematics ]

comments in ascending chronological order (reverse):

#Comment Re: made: 2008-02-28 15:33:55.907731+00 by: Diane Reese

(Did you know I run a Blue Gene/L as a big part of my day job?) The Cell processor technology has indeed been shown to be an effective platform for high-performance computing: IBM's "Roadrunner" is scheduled for acceptance testing at Los Alamos later this year. I don't know the one-to-one comparison in terms of how a raw PS3 would equate to Blue Gene nodes, and would be interested in the UMass folks' benchmarking. Our 2-rack Blue Gene has 2,048 nodes and we don't use them in partitions smaller than 32 nodes, but I can conceive of applications that might be parallelized in just the right way to approach BG performance on a very small Cell cluster.

Writing parallelized and optimized codes that take advantage of the power and capabilities of the Blue Gene is a specialized skill but I'm hoping it'll get easier. And of course not every computational task is appropriate to this sort of platform. As such, no LAMP stack.... yet. :-)

Have I succeeded in speaking a foreign language here? Oooooh yay, maybe I earned some techie cred!

#Comment Re: made: 2008-02-28 15:44:53.789103+00 by: meuon

We are not worthy, and we are envious of your blinky light factory.

#Comment Re: made: 2008-02-28 16:44:30.760522+00 by: JT [edit history]

Stanford University has been using idle ps3s for protein folding for quite some time, and even give a bit of explanation concerning calculation of flops on their page.