Clock Speed Is Not Output

By Peter Coffee  |  Posted 2001-09-10

Have you ever bought a car because it had an 8,000-rpm engine? No, you havent. So why would you ever buy a PC because it has a 2GHz CPU?

Processor clock speed measures how hard the CPU is being flogged, not how much work it does. As chips depend on ever-more-complex logic (such as Pentium 4 "hyperthreading") and as throughput depends ever more upon a PCs I/O subsystems, its becoming ever more important to watch our speedometers instead of our tachometers.

Apple has been working for years to wean PC buyers away from their megahertz mania, with campaigns that focus on CPU hogs. I approve. As shown at, a pair of 500MHz G4 chips outrun a 1GHz Athlon in tests of both graphics and gaming; either configuration outperforms a 1.4GHz Pentium 4. Clearly, clock cycles alone are not enough.

Price/performance ratios must not be ignored: Even the Bare Feats analysis, overtly Macintosh-centric, admits that Apple has an uphill task "to convince Wintel lovers that they should spend more just for the right to be abused by their friends. ... The Athlon system is clearly the overall winner." On this point, I emphatically agree.

But performance comparisons must not end at the level of CPU throughput. Connectivity, inside and outside the box, has a major impact on how much work gets done.

Fans of Apple (speaking of abuse) were quick to show their displeasure when I failed to give Apple the credit for birthing the FireWire protocol, which is now more widely known as IEEE 1394; last month, Apple received an Emmy award for its role in devising this digital video pipeline thats now being employed for many other tasks.

IEEE 1394b, appearing in hardware by years end, will improve the protocols current speed of 400M bps, with capacity of up to 3.2G bps. In the meantime, Bare Feats reports that the low-end Apple iBook has twice the throughput to FireWire mass storage devices than that of the Titanium PowerBook. Price is yet another poor predictor.

Rocket Fuel