NASA confirms manned mission to 10 Petaflops
'Our Xeon binge is named Pleiades'
Cloud storage: Lower cost and increase uptime
Well, well, well. It would seem that the 20,000-core supercomputer announced yesterday by NASA will just be the first course in an ongoing relationship between the space folk, SGI and Intel.
The three organizations have revealed a project dubbed Pleiades that will see them build Petaflop-class machines in the coming years. NASA reckons that such machines will be needed to fuel the manned missions to the moon and eventually Mars. In addition, it will look to use the gear for scientific research in a number of other related fields.
We're told that the 20,000 core Xeon-based system being built now at NASA Ames in Mountain View, California will serve as the basis for the first Pleiades supercomputer. But where the new machine is set to hit 245 Teraflops, the first Pleiades system due in 2009 will hit 1,000 Teraflops or a Petaflop. The organizations look to push that to 10 Petaflops by 2012.
SGI has a long-standing relationship with NASA Ames and could certainly use this type of business. And Intel makes out pretty well too.
Up until a couple of weeks ago, AMD had been winning most of the recent massive supercomputer deals with more than 10,000 Opterons making their way to a supercomputer in Texas and another at Los Alamos National Lab.
Intel, however, just busted into the formerly AMD-only Cray account, which means that Xeon should pop up in a number of the world's largest computers come 2011.
The SGI deal adds to this momentum, placing Intel right in the midst of the Petaflop race.
The Register's Silicon Valley staff will be looking out their office windows for all the Altix-carrying trucks heading to NASA over the coming months. Do give us a tour, boys. ®
COMMENTS
@ Plankmeister
"...NASA don't do things efficiently. It took them 30 years to realise the Shuttle was the "wrong way around*". Well... that's not true. They *KNEW* it was the wrong way around, but it was a good excuse to pour time, effort and money into some really cool engineering.
*By "wrong way around" I mean that it *should* be the first stage that's re-usable, with an expendable payload delivery system (a lá ESA's ATV). Much more efficient, especially in terms of cost of repair/maintenance. It's just nowhere near as cool."
You mean that the boosters should be reusable and the shuttle not so..? Hmm real smart. Do they use disposable astronauts too..? And the orbiter is the most expensive bit of the shuttle stack.
Also, every bit of the shutttle is re-used except for the big orange external tank. The SRB's are recovered and used too.
@ Anonymous Coward
"Personally I cannot wait for the day when the Chinese get up there, wander around for a bit and then say "Where did you said you landed? Just here?""
Maybe that's exactly why the US is in such a hurry to go back to the moon - they have to get up there and plant all the stuff that was supposed to be left behind in the 60's.
An alien... because I love conspiracies and cover up stories.
Re: Necessary? To the moon?
"In the 60's when the USA first went to the moon, the computers weren't much more than what is in a cell phone (really much less). Why all the compute power? Hasn't anyone learned how to make algorithms efficient.
Oh, I forgot, they're using Windows. I'll get my coat."
The 20480 core machine is actually using Suse Linux. Read the article yesterday..
Not that I am saying Linux is a massive resource hog..
IT infrastructure monitoring strategies
What you need to know about cloud backup
Enabling efficient data center monitoring
Agentless Backup is Not a Myth
Top 10 SIEM implementer’s checklist
