- Sunday, 8 November 2009
- INQ Mobile
- RSS
An atheist is a man who has no invisible means of support - Fulton Sheen
IBM's 20 petaflops supercomputer to keep tabs on US nukes
IBM HAS TEAMED up with the US government to try and build the most powerful, most super supercomputer ever.
Boringly dubbed "Sequioa", the aim is to get the system, which will be housed at the Lawrence Livermore National Laboratory in a series of 96 racks, to log performance speeds of up to 20 petaflops.
Boasting over 1.6 million processors and some 1.6TB of memory, the Sequoia system will initially be used to keep track of the US' mountain of aging nukes and simulating nuclear explosions, er, tests to determine whether the weapons are stable and safe (irony?!) for use. The number of cores the supercomputer will have is, as of yet, a closely guarded secret.
Sequioa should be up and running by early 2012, but in the meanwhile, IBM is already getting stuck in on its plans to begin deployment of Dawn, a 500-teraflop system which will eventually serve as the delivery system for Sequoia operations. Both systems will be constructed at IBM's BlueGene facilities in Minnesota.
When unleashed, Sequoia is expected to decimate the performance record of IBM's current supercomputer heavyweight champion, the RoadRunner system, which runs at 1.105 petaflops. Only one other supercomputer has managed to break through the petaflops barrier and give IBM a run for its money, Cray's XT5 Jaguar, which sits at the Oak Ridge National Laboratory, and runs at 1.059 petaflops.
IBM reckons its 20-petaflops capable Sequoia system will outshine every single current system in the Top500 supercomputer rankings combined in terms of raw computing power.
Besides managing stockpiles of rusty old nukes, Sequoia will also be used for research into astronomy, energy, the human genome and climate change, according to IBM. The system will also act as a giant weather cock, allowing forecasters to predict local weather "events" less than one kilometer across, compared with 10 kilometers today and at speeds up to 40 times faster than current systems.
The price of the system is still unknown, but we don't think the US treasury will be paying for it in petty cash. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
...Imagine the resolution you could run Crysis in on this bad boy
1.6 Petabytes of memory?
Phil, are you sure that 1.6TB of memory isn't actually correct? I mean, every day machines (like my own) are only in the low GB of RAM (4GB myself), so 1.6TB of RAM would be HUGE. I wouldn't expect to see PB of RAM anytime soon, but for processing speed I would expect super computers to be in the petaflops.
assuming 1GB of memory per processor, 1.6 million processors means 1.6PB of memory. and that's NOT huge when it comes to super computers. they normally have a few hundred TB's of RAM
sandy bridge is out in 2010 and will have around 200 GFLOPS per processor. why not put 1.6 million of these processors? 200x(1.6 million)gflops= 320 Pflops.
1.6 million core * 850 MHZ * 4 flops/cycle is only 5.44 PetaFlops , not 20 PetaFlops. Could this mean that there is some other hardware floating assist (like Cell) in the BlueGene/P?
Ehmmm great computer, mayor typo: "Weather cock"
How is "Weather cock" a typo? What do you think it is supposed to be?
Yeah, that is a lame name. If it had been up to me, I'd call it Napalm. What do you mean, that's insensitive ? Hey, stop pushing ! Okay, I've got my coat already !
Rather more than 'decimate', I say.
Other news around the web mention 1.6 million CPU cores, and that Sequoia will be based on IBM's Power architecture. Currently Power6 processor runs at 5GHz, Power7 is expected to have at least 8 cores and run at a similar frequency...
"Could this mean that there is some other hardware floating assist (like Cell) in the BlueGene/P?" There is a project of the same name in Stanford using an explicit hierarchical memory model and the Cell processors, among others..
Finally, we'll get to see real Terminators, centrally controlled roaming the Earth. Cool.
96 racks, 1,600,000 processors = 16,666 processors per rack. Most racks are 42U in height, which gives us 396 processors per rack U.... even with Quad core processors this seems... optimistic!
It's likely that they will use Cell Processors for this new supercomputer
http://en.wikipedia.org/wiki/Cell_(microprocessor)
I hope the writer realizes that the word decimate means to destroy a tenth of something. If this is only going to "decimate" something, it's not that remarkable.
Apparently they don't give a damn about the amount of electricity it will eat. Is there really a need for such supercomputers?
Decimate originally referred to the killing of every tenth person, a punishment used in the Roman army for mutinous legions. Today this meaning is commonly extended to include the killing of any large proportion of a group. Seems like he used 'decimate' quite reasonably.
What I want to know is why they need to simulate nuclear explosions. We know how to make them, let's try not to.
If they are going to use that to run the current climate models, that just means it will come to the wrong answer faster. A faster computer does not mean a more correct answer if the code is not corrected to come to a more correct solution...ya know.
Jeff
But will it run Crysis in 3D, that my question.
This is one serious computer, so for how much electricity its going to use, who cares considering how much it cost, which will be in 10 of millions if older system are any indication, I think they will be able to pay the bill.
I am glad its going to be used for not only Nukes but other productive human endeavors, good to see people putting things like this to work that could lead to medical break-thru.