news.com.com— IBM has devised a new Blue Gene supercomputer--the Blue Gene/P--that will be capable of processing more than 3 quadrillion operations a second.
Jun 26, 2007View in Crawl 4
When I toured the Blue Gene facility in Rochester, MN, I was told that I could buy a Blue Gene/L for only $5,000,000 and a discount of half off every other one I purchased. I can't imagine how much 72 Blue Gene/P's would cost.
The original cost of the Blue Gene/L was around $1.5 million/rack. The LLNL 64 rack system was built from a $60 million budget. The Blue Gene/P will probably be in the same ballpark. Note that this does not include the front-end equipment such as the service node, login nodes, and the storage nodes and disks for the filesystem.
It's possible he wrote a really inefficient program, or a program that generates a better program; humans currently break down at word sizes of about 512b (I've personally worked on a machine with 256byte words and it's confusing as hell and damned hard to get anything done other than padding out instructions with nops). Remember, Data is able to program himself on the fly, and that Soong spent his entire life's work on programming neural networks.
but they certainly aren't doin 216 racks like they suggest w/ the 3PF number. That's just IBM pulling the same old, "benchmark one rack and multiply till you hit a happy number." I don't know that a service node could even cope w/ a 216 rack system.
sweet, stony brook is getting one... too bad the school still sucks. we already have a cluster too, which was installed like a year ago... maybe we should get our dorms fixed up before starting a collection of supercomputers...
domdogg123Jun 27, 2007
When I toured the Blue Gene facility in Rochester, MN, I was told that I could buy a Blue Gene/L for only $5,000,000 and a discount of half off every other one I purchased. I can't imagine how much 72 Blue Gene/P's would cost.
monsterzeroJun 27, 2007
The original cost of the Blue Gene/L was around $1.5 million/rack. The LLNL 64 rack system was built from a $60 million budget. The Blue Gene/P will probably be in the same ballpark. Note that this does not include the front-end equipment such as the service node, login nodes, and the storage nodes and disks for the filesystem.
pitlordJun 27, 2007
I thought it was 60 trillion per micro second.X^P
geminitojanusJun 27, 2007
It's possible he wrote a really inefficient program, or a program that generates a better program; humans currently break down at word sizes of about 512b (I've personally worked on a machine with 256byte words and it's confusing as hell and damned hard to get anything done other than padding out instructions with nops). Remember, Data is able to program himself on the fly, and that Soong spent his entire life's work on programming neural networks.
Closed AccountJun 27, 2007
Ya, but only gets up to 29FPS in WoW, now the Quad core Mac can even get up to 160FPS<a class="user" href="http://www.youtube.com/watch?v=PM6s59OPJbk">http://www.youtube.com/watch?v=PM6s59OPJbk</a>
leesoongJun 27, 2007
''How About a Nice Game of Chess?'' - WOPR
gimpbullyJun 29, 2007
but they certainly aren't doin 216 racks like they suggest w/ the 3PF number. That's just IBM pulling the same old, "benchmark one rack and multiply till you hit a happy number." I don't know that a service node could even cope w/ a 216 rack system.
cannarymburnsJul 20, 2007
sweet, stony brook is getting one... too bad the school still sucks. we already have a cluster too, which was installed like a year ago... maybe we should get our dorms fixed up before starting a collection of supercomputers...