At the time, I asked an Intel production planner if she'd consider a rating scheme. No, she said, we'll just weather the storm. Intel has invested too much in megahertz.

Indeed.

I recalled her words when Intel disclosed the Pentium M processor, the heart of the Centrino brand. To burn less power, Intel designed the Pentium M to perform more work at lower speeds.

That's great for mobile computing. But the part will be hard to market against higher-clock chips, I said. And, unlike last time, the positioning dilemma isn't transient; it's built in.

No problem, Intel said. Pentium M will have megahertz supremacy over other low-power parts, such as mobile Pentium III and Transmeta processors.

But the low-power segment of the mobile market has not exploded, despite a $300 million Centrino media blitz. Mushrooming instead are sales of desktop replacements. To reach sales goals, then, Pentium M must fit in larger systems, and that means competing with higher-speed chips after all.

Enter processor numbering, Intel's recently concocted answer to the problem. Centrino will be marked as 700-series parts, while mobile Pentium 4s and Celerons will fill out the 500 series and 300 series, respectively.

Processor numbers swim upstream against Intel's own success. For better or worse, clock rate is the de facto proxy for performance. Intel designed the Pentium 4 specifically to woo buyers with high clock rates. Intel's Xscale group also touts clock rate as its edge.

And buyers dutifully oblige. Home users constantly tell me they have a 1.4-GHz HP or a 933-MHz Dell. Even Windows XP's stated requirements are a "PC with 300-MHz or higher processor."

That's why consumers today typically choose a 9-pound mobile PC — boasting a higher clock rate, a larger display and a lower price — over a Centrino lightweight.

Think processor numbers will change that? Neither do I.

So, Intel is left with a processor that is architecturally innovative but perceptually impaired. That will make it tough for Centrino to expand much beyond its smaller-than-expected niche.

When you think about it, the planner was right: Intel really has invested too much in megahertz.

Note: In the interest of full disclosure, I am currently doing some forecasting work for Advanced Micro Devices.

Mike Feibus is principal analyst at TechKnowledge Strategies Inc., a market re-search firm in Scottsdale, Ariz., that focuses on components for mobile systems. Reach him at mike@techknowledge-group.com.