from slate.msn.com

By Paul Boutin*
Posted Thursday, March 13, 2003, at 1:59 PM PT


On the Top 10 list for misquoted statements, Moore's Law comes in a close second, right behind "Alas, poor Yorick, I knew him well." So many differing definitions abound for Gordon Moore's edict that two years ago, in a fit of desperation, I e-mailed the Intel co-founder and asked for the original. To my surprise, he replied. "The complexity for minimum component costs has increased at a rate of roughly a factor of two per year" was his statement in a 1965 issue of Electronics magazine.

Moore meant that the number of transistors and other parts that could be crammed onto one silicon chip, while still keeping to the lowest cost, would double annually. (In the '70s, he raised the doubling time to 18 months.) To sort out the confusion, Intel has since posted Moore's article.


Moore's point four decades ago was that integrated chips the technology that replaced separately wired transistors were rapidly incorporating more and more components onto one chunk of silicon. How that would benefit human computer users, Moore didn't specify. Yet his law became an all-purpose catchphrase for the rapid doubling of CPU clock speed. It's also been applied to memory size, disk drive capacity, modem bandwidth anything related to computing. Even trained engineers knowingly misuse the term as conversational shorthand. I've been guilty myself.

The doubling-the-clock-speed definition of Moore's Law was also embraced by the company that Moore co-founded. For a decade, Intel has let us believe it was scientific fact that the company's CPU chips would double in clock speed every 18 months. Silicon Valley insiders gripe that Intel has made Moore's Law a self-fulfilling prophecy: Because customers will buy a twice-as-fast chip every 18 months, the money is available to make it happen. Red Herring's final issue last month devoted its cover to "The Tyranny of Moore's Law," a claim that along with its chip speeds, Intel's research and development budget was doubling every year and a half, a fast-growing fiscal burden crushing not only Intel but the tech industry (and, presumably, Red Herring) with it.

The confusion over the meaning of Moore's Law led some industry watchers to raise their eyebrows at Intel's new, unspoken shift in strategy: With the launch of the Centrino mobile chip set, Intel has abandoned the shorthand definition of Moore's Law. For the first time in its history, Intel isn't touting the clock speed of a new CPU. The Pentium M central processor at the core of Centrino ticks over at a lazy 1.6 gigahertz, 20 percent slower than last year's mobile version of the Pentium 4. But despite its slower clock speed, Centrino doesn't mean that Intel has given up on Gordon.

With Centrino, Intel proves that all those transistors can be used for lots of things, not just sheer speed. Pentium M's all-new design beats the P4's count of 54 million transistors on one chip with a new high of 77 million. It's not double the old count, but it's a big leap. Instead of cranking up the clock speed and then hunting for reasons for PC owners to upgrade, Intel has turned around to meet its customers' biggest grievance: laptops that run out of juice. The extra transistors on the Pentium M bring more memory cache onto the same chip, saving precious battery power. Other new circuits are dedicated to controlling and conserving power. Centrino's built-in Wi-Fi is handled not by the Pentium M but on a separate chip. Still, integrating it next to the CPU reduces battery drain.

If Centrino-equipped laptops really run five, six, or eight hours on one battery charge, as claimed, that will be a doubling of another sort. It's unlikely that the doubling of battery life will become the next shorthand meaning of Moore's Law, but for laptop users, it's something they need more than another couple of gigahertz.


* Paul Boutin is a Silicon Valley writer who spent 15 years as a software engineer and manager.