Imagine an electric car that gets 100 miles per charge and costs $50,000. That isn’t a particularly attractive equation for most people. But now imagine that same car getting 200 miles per charge. It’s still relatively pricey, but better. But at 400 miles per charge it suddenly looks like a different value proposition—particularly if it can be supplemented with a small gasoline engine that can extend the range of the vehicle.

 

And that’s just the beginning. As the price becomes more attractive to more people, economies of scale set in, dropping the price of cars by as much as half. Similar things have happened in the computer industry over the years to the point where you now need a really good justification to pay more than $1,000 for a laptop computer.

 

In every aspect of our lives, reducing power has a major impact. On an environmental scale, this can be a huge win. It can reduce everything from carbon dioxide emissions to the number of batteries thrown away in landfills. And with better battery technology—and better power management techniques—those batteries can last longer because they won’t have to be charged as often.

 

Power savings are needed everywhere—in thermostats, which can be remotely powered by an iPhone or some other handheld device all the way up to data centers, which now account for about 3% of all the power consumed in the United States. It costs money to power computers and it costs money to cool racks of servers, which have become more densely packed together with every node on the Moore’s Law road map.

 

The effects of this work are even more far-reaching. Reducing the number of fans and the sounds of motors will significantly drop the decibel level inside of buildings. Having power-aware circuits that shut down when not in use can slash home utility bills. And all of this has a cultural impact. Companies that are “green” are considered better than those that are “brown.”

 

All of this begins at the big-picture level, but it has to be solved at the chip level or even the sub-chip level. Power budgets are now an issue for everyone involved in system-level, chip-level and even board-level design, and the problems engineers face today will only grow larger over the next few years as we push from 45nm to 32nm and then to 22 nm. Multiple power islands already are an issue in everything from timing to verification, and new design strategies are under discussion to radically alter how we approach these problems.

 

Never has the weight of the world rested so heavily on the brainpower of systems engineers and architects. Welcome to the world of low-power design.

 

–Ed Sperling

This entry was posted on Friday, March 13th, 2009 at 6:39 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.