The question I hear over and over from readers and friends goes something like this: "I need a computer for running Microsoft Office and browsing the Web, not much else. How fast of a processor should I get?"
It's not hard to see where that query comes from. Decades of computing history have taught one lesson very well: The most important part of the computer is the processor, and the higher the processor's clock speed, the better.
But this lesson plan badly needs a rewrite. First, ever-accelerating chip speeds have blown past the needs of many applications -- even the slowest machine sold today can handle most everyday computing tasks without perceptible delay. And for those chores where chip performance does matter, clock speed isn't the right measurement anymore.
While few people in the hardware business seem eager to say that processors are becoming a commodity, they are thinking anew about how to describe processing power. Over the past few years, AMD and Apple (whose chips had fallen behind Intel's in the clock-speed contest) have been vigorously advancing the "clock speed is obsolete" argument. AMD, Intel's chief rival, even went to the trouble of labeling all its processors with speed ratings that indicate which Intel chips they match up with best.
Now Intel -- the foremost exponent of megahertz and gigahertz as the prime numbers of personal computing -- is making its own move away from clock speeds to a new, three-digit numbering scheme.
So far, Intel has shipped seven processors under this system: the Pentium M 735, 745 and 755; the Mobile Pentium M 518, 532 and 538; and the Celeron M 340.
From eyeballing those numbers, picking a processor would seem quite simple: The Celeron is obviously the slowest, while the 755 Pentium M runs more than twice as fast.
Wrong. These numbers don't describe performance in any way.
Intel says they're supposed to reflect five processor features -- architecture (in terms of a chip's "die size," its dimensions when first cut from a wafer of silicon, with smaller sizes offering more efficient operation), Level 2 cache (a buffer of high-speed memory to store frequently used snippets of code), clock speed, front-side bus (which sends data between the processor and the rest of the system), and a generic category of "future Intel technologies."
But no formula determined that the Pentium M 755 would bear that number and not 734 or 737. As Intel says on its Web site (www.intel.com/products/processor_number/), "the digits themselves have no inherent meaning."
Intel could have christened its chips after flavors of ice cream, baseball teams or one-word adjectives ("good," "better," "best") and communicated the same amount of useful data to customers. For that matter, the price of each processor gets this point across just as well.
But even if each number did clearly convey such details as a chip's clock and bus speeds, would that matter? Unless you're a chip designer yourself, why should you care how many kilobytes of cache it carries, as opposed to how fast it runs in a particular computer? It's as if a car were advertised with all sorts of data about its engine's displacement, valves per cylinder and so forth -- but no information about its 0-to-60 performance.
Becky Brown, Intel's consumer desktop product marketing manager, unintentionally reinforced this point in an interview Tuesday. After a discussion of the ways in which people might find themselves needing a more capable processor, I remarked on how often she had used the word "performance" in the last five minutes -- while, as she admitted, Intel's three-digit numbers are silent on that topic.
What Intel should have done is find a reasonably simple, broadly applicable measure of performance and sell its hardware by that.
But I can't be too hard on Intel here: No other major manufacturer has managed to do that, either. When AMD began labeling its chips with its new speed ratings (for example, an Athlon XP 3200+ is supposed to perform as well as a 3.2 GHz Intel Pentium 4), the company also developed a suite of benchmark tests, which it said it would work to get other manufacturers and retailers to adopt.
AMD, however, has yet to persuade anybody else to adopt that idea. And even its own speed ratings can't be used to compare AMD chips of different AMD families (say, budget-priced Duron chips vs. high-end Athlons).
It shouldn't be this hard, to judge from other industries' track records. AMD's aptly named marketing director, Hal Speed, cited the example of light bulbs in an interview Thursday: Instead of comparing performance of incandescent and compact-fluorescent bulbs by the watts of electricity they consume, we measure their light output in lumens.
Until the computer industry can come to a similar agreement, the best way for consumers to shed some light of their own on the situation is to remember this: A processor's clock speed is only one number, and in many cases, it's the least important number.
If a computer's going to be used just for those stereotypical tasks of word processing and Web browsing, don't worry about its chip at all. Those types of programs, as well as the operating systems they run on, have either changed little since 2001 (in the case of Windows XP) or have actually gotten faster with successive updates (in the case of Mac OS X).
If, on the other hand, a new desktop or laptop will be put to work at such computing-intensive activities as editing digital photos or videos or playing computer games, the chip's clock speed can't be considered without also assessing such details as the bus speed, cache size and, if you're a gamer, the graphics card (see related story below).
If you do find your computer is feeling slow, look to the other chips in the computer -- the memory installed. Simply going from 256 megabytes of memory (the minimum I'd get on any machine these days) to 512 megabytes can yield an eye-opening speed boost. People might even think you upgraded the processor.
Living with technology, or trying to? E-mail Rob Pegoraro at firstname.lastname@example.org.