Fujitsu Ships Parts for the Fastest Supercomputer
10 petaflops... a lot of flops.
Fujitsu is now shipping the first parts to its next-generation supercomputer to the Japanese government-funded RIKEN research institute.
The supercomputer, called K, will have 800 racks with 80,000 Fujitsu SPARC 64 VIIIfx processors running at 2.2GHz.
Fujitsu announced this new processor during May of 2009, and great things are expected of it when it'll be the brains behind what looks to be the world's fastest supercomputer – by a fair margin. The only problem is that the computer isn't set to be up and fully operational until 2012.
The Fujitsu K is projected to be capable of 10 petaflops. Right now, the fastest supercomputer in the world is Cray's Jaguar system at Oak Ridge National Laboratory with 1.75 petaflops.
(Source: Cnet.)
- Fujitsu,
- Business Computing,
- sparc ,
- venus ,
- supercomputer ,
- cray ,
- riken
- Voting Time! Choose This Quarter’s Best Configs
- Origin's EON17 Laptop Has Maxed Out Specs
- Law Firm Facing Investigation After Details of 5,000 Broadband Users are Leaked to Pirate Bay
- Wi-Fi Seen as Important Part of Relationships
- Intel Staff Base Jumping Off a Six Story Building
- Gigabyte Remains Mum on P67, LGA 2011
- LG Announces First ''Near'' Full HD 3D Notebook
- Acer Adds New Predator Gaming Series, More
- New Onyx 2 SSDs from OCZ Spell Affordability
- ASRock Wants the Best Mobo Ideas in the World
- Microsoft May Buy Second Life Developer
- Galaxy Adding Wireless HD to GeForce GTX 460
- Apple TV Teardown Shows A4 CPU, 256MB of RAM
- Nvidia GPU Settlement Covers Apple, Dell, and HP
- Oh Snap! New HP CEO Comes From Oracle Rival
- Blizzard Banned Over 5,000 StarCraft II Players
- Report: AMD Launching Radeon HD 6000 Oct. 18
- Google Makes WebP in Effort to Make JPEG Extinct
woah...petaflops
Please, don't ask if it can play Crysis.
what would they need all that computing power for?
How much power will it consume?
what would they need all that computing power for?
Things like discovering how the universe works and curing cancer
Both of which I am ok with
I really can't wait to see what uses they have for that much raw power.
How much power will it consume?
I am waiting for the first Ass to say "Not more than a GTX 480!"....
What a monster. That is over 5x faster than the jaguar. Amazing!
The Jaguar only beat the IBM roadrunner by .71 petaflop's and that was only after a CPU upgrade.
what would they need all that computing power for?
Figuring out how they will pay for all that power draw?
Figuring out how they will pay for all that power draw?
Or maybe trying to figure out how they'll pay for the Fujitsu K? ;-)
Why not a rack of cuda ready cards? Same thing. Cheaper.
that is a crazy amount of processing... but i'll wait till we get actual results before i believe the 10 petaflop claim
just PetaFlops doesnt mean anything performance wise.
and it's simply called "K". they don't waste time on such trivial task as coming up with a sophisticated name. that's Japanese efficiency.
and it's simply called "K". they don't waste time on such trivial task as coming up with a sophisticated name. that's Japanese efficiency.
This might be true, except every conversation about it is going to have to include:
"It doesn't have as much power as K. K? You don't know K? You know, the terribly named non-descriptive title of the Japanese supercomputer."
Very efficient.
what would they need all that computing power for?
computing
We were told a while back that Intel was falling behind... looks like it is upon us.
@mlopinto2k1
these arent Intel or even AMD chips, they SPARCs which are designed to be good at number crunching, architecturally not too dissimilar to the way GPGPU works in fact
what would they need all that computing power for?
For calculating the ultimate answer to life the universe and everything
You do know a petaflop is a quadrillion computations per second. That's some pretty beefy performance if you ask me.
For calculating the ultimate answer to life the universe and everything
Easy, that's 42. Can I have all the money spent on this project?
Supercomputer for 2012... Desktop by 2016... Phone by 2020.
Imagine the folding on that one.
Easy, that's 42. Can I have all the money spent on this project?
Maybe they are trying to find the ultimate question to which the answer is '42'
Supercomputer for 2012... Desktop by 2016... Phone by 2020.Imagine the folding on that one.
It might take a little longer than 10 years to have 10 petaflops in a cellphone.
The desktop to cellphone divide is a good 10 years, the supercomputer to desktop divide is probably around 20 years. By the time my kids have cellphones (hell they will probably be on their 5th cellphone) they will have this kind of power.
IBM's Sequoia supercomputer at Lawrence Livermore Labs is expected to be 20Pflops by next year. That is an x86 machine, but all these new GPGPU machines that are planned will be over 10Pflops as well.
@joebob2000
if the hype is to be believed, we wont need a 10 petraflop chip in our cell phone, but we will have access to it through the cloud..... we have the tech for that today
Supercomputer for 2012... Desktop by 2016... Phone by 2020.
Imagine the folding on that one.
Instead of paying for a supercomputer they should rent computing power from the general public. There is an insane amount of computing power out there doing nothing. A 100,000 or more average home PCs connected to some distributed super computing service through the internet.
The folding @ home program has the basic idea but a company should start up that actually pays people for their computing power when they arent using it.
Im sure this would be a much cheaper route as there is no power or hardware costs. How much did is this K Beast going to cost again?
Things like discovering how the universe works and curing cancerBoth of which I am ok with
You and me both. It really is incredible how they're starting to use computers more and more in these research studies, taking a good load of the tedious workload off of the scientists. To all of you running folding-@-Home, cheers and thank you
Instead of paying for a supercomputer they should rent computing power from the general public. There is an insane amount of computing power out there doing nothing. A 100,000 or more average home PCs connected to some distributed super computing service through the internet. The folding @ home program has the basic idea but a company should start up that actually pays people for their computing power when they arent using it.I'm sure this would be a much cheaper route as there is no power or hardware costs. How much did is this K Beast going to cost again?
There is a fundamental problem there, with folding and Seti@home, those are easily distributed since work units are nice and separable, one unit does not depend on another. Not all problems can be separated out that way. Climate science and model prediction is one, where there are so many variables so interdependent on each other that if you're waiting a week for half the work units to come in before the next grouping goes out, the latency will negate all the gain from using so many computers.
Oil exploration, large scale cosmology, massive physics simulations, high speed material analysis, there are many many problems that are just very large and not easily broken down. Plus if someone is renting the time out and the company who is buying the computer time and the data or work they are doing is trade secret, what better way to do so on a system you can control.