Reducing Computer Power Consumption

Wow, this is a very cool result:

Researchers at IBM's Thomas J. Watson Research Center in Yorkstown, NY have announced a breakthrough which they feel could revolutionize power consumption in computers. Today's computers are power hungry: a typical computer consumes hundreds of watts of power. Not only does this power consumption add up to a lot of wasted power, but increasingly the amount of heat generated by the machines is a significant barrier to building faster more powerful computers. The researchers at IBM say they've made a breakthrough in how computers consume power which will dramatically lower power consumption at the cost of only slightly longer time to perform computations.

"Computation is inherently a power consuming process. Every time a transistor performs a calculation in your computer, you generate a little bit of heat. Want to add two plus two? That will cost you some energy," said lead researcher Charlie Tennett in explaining the teams new approach. "But what we realized was, that this is a two way street. What happens if, instead of adding two plus two, you take four and break it up into two twos?"

The heart of the IBM researchers breakthrough is the observation that if "computing" costs energy then "uncomputing" can be used to generate energy. "By performing a computation, copying the answer, and then unperforming the computation, we only consume power during the copying procees" says Dr. Tennett. "Since that copying doesn't consume much energy we get huge efficiencies in power consumption."

As a demonstration of this breakthrough, Dr. Tennett showed a prototype computer built by team members John Swolin and Barbara Shareal. "We took an ordinary laptop and changed its programming. Every time we execute a program on the computer, the program copies over the answer and then does the same program, but running backward. We then hooked up our laptop to a power meter, and watched, in amazement, as almost all of the power consumed by the program was fed back into the electrical system when we ran the program backward!"

The team believes that there is a great potential for the use of their technology even in fields outside of computation. "Think of all the vast amounts of computation that has been performed over the last few decades," dreams Dr. Tennett, "if we could just undo all of those computations, we could easily ween America off of its oil addition."


More like this

good one, Dave!

Not a bad attempt at an April fools day joke.

Nice! I found myself wondering why they were claiming a new result, when we've known about revesable computation for years, but then reached the penultimate paragraph and the joke was revealed.

I'm trying not to read anything serious today as I keep thinking legimate results might be April fools' jokes.

I am tempted to erase your comments and thereby gain the energy resulting from this un-computation.

I vote for Erasure!

I thought I understood how you arrived at "Tennett" and "Swolin", but I am still pondering the far more subtle transformation "Terhal" -> "Shareal". Probably this is a clever pop-culture reference inaccessible to people over 50.

Ilove April fool's day

Thanks for the publicity concerning our new results about which IBM is pretty excited. One thing to note is that in our demo the regenerated energy was actually dumped in the computer's past. This means that it cannot be used to solve the world's current energy problems, still we are hopeful that it can retroactively avoid our society getting addicted to oil in the first place.

By Barbara Shareal (not verified) on 01 Apr 2009 #permalink

Spent the first two paragraphs wondering how this was new, and thinking that maybe it was just the usual PR fluff that precedes any new advance in an established field.

Didn't realize until paragraph three that they weren't talking about:

One of these years I'll remember what day it is before I go read news ...

> One thing to note is that in our demo the regenerated energy > was actually dumped in the computer's past.

Oo, does this mean they've managed to harness a closed time-like curve?!?!