Home Green Tech Power Saving

Koomey's Law Predicts Computer Energy Efficiency Doubling Every 18 Months

448
2

If you remember, in August I published an article about a researcher from Stanford who had discovered that the electricity consumption of datacenters worldwide didn’t double between 2005 and 2010, versus the period between 2000 and 2005, when it did. His name is Jonathan Koomey, and by now there’s a law wearing his name.

Koomey’s law,” as it was named, says that the energy efficiency of computers doubles every 18 months. The theory resembles Moore’s law, but with a twist for efficiency.

Gordon Moore had been working at Intel when observed that since 1946, starting with the first general purpose computer (called ENIAC, which consumed 150 kW and only did a few hundred calculations per second), the number of transistors that can be placed inexpensively on an integrated circuit doubles approximately every two years (source: wikipedia).

It seems that even the energy efficiency is following the same rule. In a world where the importance of battery power overtakes performance in tablets, smartphones and laptops there’s a constant need for optimization rather than for innovating more powerful processors.

Theoretically, the energy efficiency of computers working with transistors, as it was assessed in 1985 by physicist Richard Feynman, could have improved by as much as 100 billion times before reaching its limit.  By now it has only reached 40,000 times, so there’s a long way ahead.

Of course, all of the rules above could be changed by the eventual introduction of quantum computing, whose evolution in terms of both power and efficiency would be on another scale.

(Visited 519 times, 1 visits today)

2 COMMENTS

  1. It’s an honor to have you commenting on our website. Thanks for the remarks. I corrected the name typo, and although I knew Moore’s law, the “performance” interpretation had been given by MIT’s Technology Review. I nevertheless corrected the context. Thanks again.

  2. Thanks for the shout out. Note that my name is Jonathan (no “h” in the third letter slot).

    Moore didn’t talk about performance, to my knowledge, only component density doubling every year (in his 1965 paper) or every 2 years (in his 1975 paper). And he didn’t talk about ENIAC in this context, because he was focused on microprocessors, not tubes or discrete transistors.

    Our paper identifies the trend in power efficiency going back to ENIAC, so that trend has proceeded longer than Moore’s law. It appears to be an inherent feature of information technology equipment that uses electrons for switching.

    For an excellent historical summary of what Moore’s law is and isn’t, check out Mollick, Ethan. 2006. “Establishing Moore’s Law.” IEEE Annals of the History of Computing (Published by the IEEE Computer Society). July-September. pp. 62-75.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.