HPC
Researchers break records with MILLION-CORE calculation
Stanford’s Engineering Center for Turbulence Research (SECTR) has claimed a new record in computer science by running a fluid dynamics problem using a code named CharLES that utilised more than one million cores in the hulking great IBM Sequoia at once. According to the Stanford researchers, it’s the first time this many cores …
This topic is closed for new posts.
Maybe :-)
Signed - Supercomputer manager of an F1 team.
I would also like the trip out to California as I'm absolutely sure somehow that remote access won't work until I've had a couple of weeks out there on the beach. Ahem. In the server room.
Erm..
Is there an echo in here ?
http://www.theregister.co.uk/2013/01/28/stanford_uni_million_core_calculation/
The future of home computing
Modern smartphones today have more computing power today than the Cray-1 had 30 years ago.
Can we therefore conlcude that in about 15 years time this will be the kind of power we will have in home computers.
Shit, will we even have "home" computers in 15 years or will we all be directly connected to the Cloud/Web via Matrix like cervical cords.
<<<---- There is no IT Angle becaue this moves into the realm of the unimaginable.
Re: The future of home computing
Probably. But they still wont be able to get the weather right 5 days ahead or the climate in 5 years...:-)
Re: The future of home computing
I did a calculation a couple of years ago and found it took about 12 years for the MIPs value of the worlds top supercomputer to be available on a single graphics card.
Re: The future of home computing
I'm sure this is just a satirical jibe at the met office, but something sciency in me compells me to inform you that the mathematics of chaos, not the ineptitude of the met office, is the reason they can't predict the weather.
What the met office actually do is run a large number of simulations, all with perturbations from the current weather situation and determine the probability of certain events. E.g. say they run 20 simulations and it rains in 15 of them, then they say there's a 75% chance of rain. This was taken away from our TV weather reports because people are stupid and don't what that means. So they say "It will rain today". Then, when it's bright sunshine (as predicted in 25% of the simulations), people say "blah blah, crap forcasters!" and laugh.
Re: The future of home computing
But they are crap. They give you a 5 day forecast and change it every couple of hours.
They give a % because the whole process is so complicated they can't give a definitive answer, so why bother?
I have given up trying to predict lottery numbers, it's too complicated. So I guess, and I'm not too proud to admit it.
Re: The future of home computing
I somehow feel things are heading on a mundane route, where we may have a lot of power, but we're doing nothing more with it than we are now.
1 Million FPS iPhone transition animations! 16000x9000 phone screens that can run 3D games that still don't have good collision detection!
Re: The future of home computing
Here across the pond, about the only thing the National Weather Service gets wrong to the point that people gripe is exact snowfall locations and amounts (perhaps not something you need to worry about in the UK, but here, there is a bit of a difference between 2" and 6") and tornado warnings that turn out to be false alarms. The former is because a slight shift (a dozen miles, if even) in the upper atmosphere can change both the type and the amount of snowfall, while the latter is erring on the side of caution, as tornadoes not just ruin your picnic, but probably put the sandwiches through the house down the street (maybe a slight exaggeration).
Maybe it's the island location that plays havoc with the weather forecast, or maybe the expectations are so high, it's absurd. Given than the NWS is able to predict the general path and area of impact of most hurricanes five days in advance within a 300 nautical mile error, I'd say we're doing pretty good. Just because it rains on you when you forgot your umbrella doesn't mean it's the end of the world.
Re: The future of home computing
300 miles is fine across the pond but thats halfway across the entire country here...
Re: The future of home computing
There is a reason why weather forecasting is one of the first practical studies that led to much of the discovery of chaos theory. You change one variable slightly in weather models and you find totally different results 5 days later. Basically the same concept behind cryptographic hash functions where changing a single input bit causes the output to totally change.
http://en.wikipedia.org/wiki/Edward_Norton_Lorenz
Re: The future of home computing
>> Can we therefore conlcude that in about 15 years time this will be the kind of power we will have in home computers.
Yes. We can also conclude these computers will have performance issues when running the latest version of Windows.
Re: The future of home computing
And HTML 5 will still NOT be standardized.
Can we have a piping hot update on the following problems in computation, please:
1) Classical General Relativity problems. Computing the behaviour of a 4-D warping, lorentzian spacetime? Hell yeah.
2) Lattice Quantum Chromodynamics, which, AFAIK is impossible to parallelize efficiently.
Googling (2) was fun, but didn't answer my question ... why it is hard to parallelize efficiently? The FermiQCD toolset seems to be based on parallel matrix algorithms but I couldn't get a quick idea of the limitations there.
What sort of million ?
Would that be 1000 x 1000, or 1024 x 1024 ?
Re: Thank God..
The only decent machine to run EDLIN is one that is switched off...
Re: Thank God..
> The only decent machine to run EDLIN is one that is switched off...
Yeah, real programmers use vi... in line mode... with a nokia 6310i as a terminal... over an ir link...
Re: Thank God..
> >The only decent machine to run EDLIN is one that is switched off...
>Yeah, real programmers use vi... in line mode... with a nokia 6310i as a terminal... over an ir link...
Okay you got it right until you mentioned an ir link. Real programmers use an RS-232 cable. Homemade of course.
Can it model the problem with the 787? I think Boeing would love some help.
"Can it model the problem with the 787? I think Boeing would love some help."
If you put in a small enough power supply. But why would you want your computer to overheat and catch on fire?
Veni vidi duci
To quote and old Internet saying: " Uunngghh!"
This topic is closed for new posts.
