November 4, 2003

Underclocking: what's the rush?

Author: Joe Barr

Underclocking's a story that would remain forever inexplicable if it were not for the string unification theory. An old trend is slowly wending its way into and around the Linux crowd: underclocking. For years I have taken it as an article of faith that all geeks want more speed. The more, the better, thank you very much. That's certainly been true with me. But now I've learned that this is not universal. Over the years, a small group of geeks have eschewed the smoking-edge in favor of fewer clock cycles. Here's why.

I first read about underclocking on a local LUG mailing list. It piqued my interest so I began to look around for more information on the subject. While some sites have reported underclocking as a fad or the latest craze, it is neither. Not using the Google Newsgroup search results as an indicator, at least. Interest in underclocking began in 1993 when there were 6 mentions of the word in USENET newsgroups. The number of mentions of underclocking peaked in the year 2000 with at just over 2,300. Since then, it has been in a slow decline. Based on extrapolation of year-to-date figures, this year's number should be roughly 1,300.

In terms of popularity, underclocking definitely runs against the grain when compared to overclocking. Searching the newsgroups for hits on the term "overclocking" this year indicates the total for the year will exceed 50,000. Nearly a 50-to-1 difference between the two.

Why underclock?

Heat and faulty power are the two greatest causes of death in electronics. The certified geek Olympic sport of overclocking usually shortens the life of CPUs and perhaps the mainboard itself simply because running a chip at a faster frequency makes it run hotter. You may have seen some of the extremes that overclockers put into cooling their high-revving Athlons and Pentiums. But no matter. Maybe the overclocking yielded an extra FPS or two in an FPS like Quake or UT2003. (That first FPS is Frames per second, the second is First Person Shooter.)

While many geeks with PCs have been consumed by squeezing more speed out of their system than the law allows, more mature mainframe geeks have long been focused on adding 9s. As in uptime. Speed has been secondary to finding ways to make hardware more reliable and less prone to failure. And that's what underclocking is all about. Usually, that is.

Quiet is cool

Terry Gray wants quiet, not speed. He got into underclocking in his quest to build a quiet PC. Specifically, a PC without fans on the CPU and powersupply. His site documents his experiments with underclocking a Pentium III in order to be able to run it at an acceptable temperature without need of a CPU fan. Other underclockers have done the same.

CPUs are not the only chips being underclocked to attack heat and noise problems: video cards are also game. Leo Velikovich details the hows and whys of such an effort on a GeForce4 Ti4200 in his article in SilentPCReview.

Speed is born not made

Under and overclocking are both possible because of the way microprocessors are manufactured and sold. The speed of your new Athlon XP or Intel Xeon is not predetermined, it's the result of QA. Post-manufacturing tests by the manufacturer assign each chip the highest speed rating they are comfortable with. One overclocking site suggests that Intel chips allow a little more headroom for overclocking than does AMD. But chips from both manufacturers are regularly overclocked. And underclocked, too, though to a much lesser degree.


Underclocking, like overclocking, may meet the specific needs of some users, but they are definitely not the right thing for most users. Overclocking can shorten component life, sometimes dramatically. Underclocking is usually done to make the hardware more reliable and extend its useful life. But it can still cause problems, and I have to wonder how many people keep the same hardware for the number of years required to benefit from its extended life.

Keep in mind the point Terry Gray makes on his website: modern Intel and AMD processors are multiplier-locked: CPU clock-time is a multiple of the bus speed. Any time you make a change to the speed of the processor, no matter whether the new speed is faster or slower than before, it must also be a multiple. If you're not sure what that means, I suggest you leave it alone awhile longer. Maybe the urge to tweak will pass.

Click Here!