Recent Comment
Share
Energy savings of 99% over previous methods probably sound like snake oil. But some math geeks have been able to find a way so that computers can use only 1% of the energy (and the time) necessary for some tasks.
IBM has announced a new data-processing algorithm that enables large sets of data to be processed in a fraction of the time, and with only a fraction of the electricity, as was previously needed. Supercomputer testing of the algorithm showed that the process speeded the calculations and reduced the power consumption by two orders of magnitude.
"The new method was tested on the fourth largest supercomputer in the world and what would normally have taken a day, was crunched in 20 minutes. In terms of energy savings, the analysis required 700 kilowatt-hours total, compared with 52800 kilowatt-hours total."
Not only does this mean that far less electricity is needed for data-intensive operations, but also more work will be able to be scheduled onto the same number of machines. And while the demonstration was carried out on a powerful supercomputer, the intent of the research was to make intensive data-processing activities more readily available and accessible to scientists who do not have supercomputers.
via: Slashdot
image: IBM press release
Hits: 18840
Comments (8)
Subscribe to this comment's feed
written by Carl Hage, March 04, 2010
It's good that people are working on more efficient software as a solution to excess computing (energy and networking) costs. Usually, complex math isn't even required. Software seems to get more and more bloated and inefficient. Because computers are cheap and very fast (billions and billions of ops/second), it's easier to buy a faster processor than rewrite bad code. Also, programmers tend to buy really fast machines, so don't notice when their software sucks.
Also, note that typically a computer runs at 98% idle, but power usage remains almost the same. A computer might have multiple cores, but the idle cores still burn power. Why not have a multicore system with a low-power processor with other high performance cores usually powered off until needed.
A related project is the sort benchmark contest http://sortbenchmark.org/ sorting the most records per joule for 10GB-10TB. Though algorithms are part of the solution, the main thing is computer system architecture.
Also, note that typically a computer runs at 98% idle, but power usage remains almost the same. A computer might have multiple cores, but the idle cores still burn power. Why not have a multicore system with a low-power processor with other high performance cores usually powered off until needed.
A related project is the sort benchmark contest http://sortbenchmark.org/ sorting the most records per joule for 10GB-10TB. Though algorithms are part of the solution, the main thing is computer system architecture.
-
- +1
-
-
written by Doc, March 05, 2010
Remember when you wrote your code, then compiled it to run? Man was it small and fast! Now it's Windows bloatware. Now programs are 10,000 larger, run no faster, despite CPU power orders of magnitude higher than we had 20 years ago. Granted, it's doing more, but one would think it should be nearly instantaneous when my multiple core CPU is running on a gigahertz level, no megahertz single core 16-bit CPU's in the past.
Remember when Windows 3 took nearly a minute to boot, and Netscape took 10-20 seconds to load? We haven't sped that up too much in the last 15 years, have we? My Windows 7 machine still takes about 45 seconds to boot to a working desktop, and Firefox and Explorer can take up to 10-20 seconds to load fully.
Not complaining, just making the observation that I wish the speed of application opening was orders of magnitude quicker than it is...
Cheers,
Doc
We're using faster computers and more power to make up for sloppy code and
Remember when Windows 3 took nearly a minute to boot, and Netscape took 10-20 seconds to load? We haven't sped that up too much in the last 15 years, have we? My Windows 7 machine still takes about 45 seconds to boot to a working desktop, and Firefox and Explorer can take up to 10-20 seconds to load fully.
Not complaining, just making the observation that I wish the speed of application opening was orders of magnitude quicker than it is...
Cheers,
Doc
We're using faster computers and more power to make up for sloppy code and
-
- +0
-
-
written by jake, March 07, 2010
im not sure i really understand this as well as i would like. is this the sort of thing that can be used by any computer to make it faster and more efficient? or is it just for supercomputers? if it can be used for regular computers will they be selling it as an update or what? i want it
-
- +0
-
-
written by Bob F, March 11, 2010
I don't want to detract in any way from this programming accomplishment or these programmers' skill, but I can recall several times in my programming career when a better algorithm improved processing speed (and hence, energy consumption of the computer system) by a similar impressive factor. True, I wasn't working on a supercomputer, but I was working on software that would be used many times by many people. This is why algorithm design is very important. I have a feeling that there are many situations in everyday computing in which the optimal algorithm isn't sought simply because the immediate benefit seems insignificant.
-
- +0
-
-
written by paul, April 09, 2010
suppose this will work. it only means that more tasks can be done in a day. but everybody will still consume that same amount of energy. it's not you'll pack up and go home early because you have a superfast computer. your tasks will just increase.
-
- +0
-
-
written by Ray, June 18, 2010
The energy savings is not in a single job, but in the fact that you can run a hundred more jobs in the same amount of time (& energy).
Ever wondered how fast the original Windows would load on todays machines? (2-seconds?)
Ever wondered how fast the original Windows would load on todays machines? (2-seconds?)
-
- +0
-
-
written by Beverly Hills Lasik Eye Surgery, June 23, 2010
I find revelations in technology like this absolutely fascinating. It just blows me away that a more refined intelligent algorithm can pivot humanity in such a quantum way. I'm definitely a fan of efficiency (always have been). It joyously boggles my mind to think about how a string of numbers will be leveraged in myriad ways, and the by-product of just a unique sequence.
-
- +0
-
-
Write comment
| < Prev | Next > |
|---|
so it's kind of like CFLs for computer software, kinda...okay not really, but still, this is a giant leap for supercomputing ^_^ I wonder if they can get it to work for cluster computing.