March 28, 2008:

Today I was presented with a much prettier Google homepage, for it was on a black background, in an effort to promote . This in turn linked to a blog entry in which they reported that a study had shown the power consumption difference between a black-background page and a white-background page to be negligible, for an assortment of reasons - most notably, that market penetration of LCDs is estimated to exceed 75%.

The original study ( came to much the same conclusion.

"There’s no argument that on CRT monitors, Blackle does reduce the power consumption but it’s not by the 15-watts claimed. We tested the four CRT monitors we could get our hands on and found that only one unit, an older 22-inch Compaq, showed the 15-watts or more power differential." "But with the LCD monitor market penetration worldwide now beyond 75%, it’s the LCD monitor power consumption that’s just as, if not more, important."

Of the 27 monitors they tested (23 LCDs, four CRTs), the average power consumption change was +100 milliwatts - displaying mostly black, on an LCD, uses a tenth of a watt more power than displaying mostly white. And almost all monitors sold nowadays are LCDs. So... open and shut case in favour of white backgrounds?

Not so fast. The average power consumption change on the CRTs they tested was 10.8 watts.

So... what web design standard really saves the most power?

If we take their numbers as an acceptable generalization for all monitors installed worldwide, a black background saves power (on average) as long as CRTs make up about 1% of the total installed monitor base.

Mostly, it's the magnitude of the difference. It's impressive on the CRT, it's genuinely negligible on the LCD.

Granted, this does ignore the main thrust of both Google and Yates - that we would save much more power by turning off things when they aren't in use more often. Which is true - but remember not to fall into the trap of assuming that since Y is more important than X, nothing need be done about X until Y is solved. Because most of the time, doing something about X does not actually prevent you from doing something about Y.

A final note - while Blackle presents itself as a power savings over Google, has anyone calculated the serverside and routerside power cost of routing search requests through Blackle on their way to Google?

(The above is a slightly snotty question; I don't _know_ what the results are. I suspect, however, that the power cost per packet is truly negligible and so the most significant extra cost of using Blackle is Blackle's own electric bill.)


{ Add Comment }