When appropriately impressed gawkers check out the game rig I built, their first question is invariably, "How fast is it?" Meaning, how powerful is its CPU? Historically, a processor's power has been measured by its MHz--and, more recently, its GHz--rating. That, however, is no longer true.

Back in 2001, AMD changed its CPU model numbering to reflect performance levels as opposed to actual clock speeds. This occurred with the release of the Athlon XP 1800+, 1700+, 1600+, and 1500+ processors. According to AMD, its QuantiSpeed architecture makes clock speed less relevant because the chip executes more instructions per clock. Since that change in nomenclature and chip design, AMD has tried fervently to garner industry and consumer support for a more accurate performance measurement than what's offered by clock speed alone.

For example, in relation to the recent CPUs, AMD's Athlon XP 2800+, 3000+, and 3200+ are understood to offer performance equal to Intel's 2.8GHz, 3.0GHz, and the imminent 3.2GHz processors. Naturally, the question must be posed whether that assumption is accurate. The proof, in this case, is in the benchmarks. And lest you think Intel would never agree that clock speed doesn't tell the whole story, consider its new Centrino mobile processors. In recent tests, a 1.6GHz Centrino-based laptop outperformed a 2.4GHz Pentium 4-M system on the exact same applications.

In fact, with the release of the Centrino, as well as the latest Pentium 4 desktop CPUs supporting the new 800MHz Front Side Bus (FSB), Intel is marketing its processors and chipsets as multichip performance packages referred to as "solutions." The issue is no longer whether other factors beyond clock speed belong in the equation when gauging system performance. The issue is now whether the claims and results are accurate.

So while processor speed is still the most significant measure, other considerations must be taken into account. This includes the amount of onboard L2 cache and the FSB speed, as well as accompanying chipsets.

As of this writing, the current speed demons are the AMD Athlon XP 3200+ and Intel Pentium 4 3.0GHz. The 3200+, based on the AMD's Barton core, is a 2.2GHz CPU that runs on a 400MHz FSB, features 512K of L2 cache, and is built on a 0.13-micron copper fabrication process. Intel's 3.0GHz CPU operates on an 800MHz, quad-pumped FSB, and features 512 KB of full-speed L2 cache and is built on 0.13-micron technology.

Understandably, based on its moniker, you might assume that the 3200+ is the speedier CPU. According to AMD's claims, it's "the world's highest-performing desktop PC processor" and it "outperforms its closest competitor by an average of 6 percent on a variety of industry-standard software benchmarks." So, what do the benchmarks really tell us? The results are interesting, to say the least.

Our benchmarks included Futuremark's 3DMark 2001 SE, 3DMark03, and PCMark2002, as well as games Quake 3 Arena, Unreal Tournament 2003, Comanche 4, and Serious Sam SE. Each system was similarly configured and sported a Radeon 9700 Pro video card with all tests set to a resolution of 1024-by-768 in 32-bit color.

3DMark 2001 SE, 3DMark03, and PCMark2002 all paint a similar picture, favoring Intel.

Intel's 3.0GHz CPU outshines AMD's 3200+ across the board. In 3DMark 2001, the P4 scored 4.2 percent above the XP. Meanwhile, in 3DMark03, an increase of 1.8 percent in the 3DMark Score and 8.9 percent in the CPU Score prevailed for the P4. PCMark2002 shows analogous results with an 8.6 percent lead in the CPU Score and a whopping 54.8 percent in the Memory Score. Thus, in synthetic tests, the Pentium 4 processor exhibits a decided advantage. Moreover, Intel clearly leads the way at this juncture in memory performance.

In gaming benchmarks, the story is comparable, but with a twist: the 3.0GHz P4 outperformed AMD's XP overall. In Quake 3 Arena, the 3.0GHz bested the 3200+ by 19.5 percent. Unreal Tournament 2003 was closer with Intel leading by a minimal 3 percent. Comanche 4 evidenced a wider margin of 15.6 percent to the P4's advantage. However, in Serious Sam SE, the Athlon trounced the P4 with a clear 12.9 percent lead. That's not unusual for this game, though, since it has repeatedly favored the Athlon XP.

What can we determine from these tests? First, Intel's Pentium 4 3.0GHz processor with 800MHz FSB trumps the 400MHz FSB Athlon XP 3200+ in most games. AMD's claim of an advantage on "average of 6 percent" didn't bear out. Some games, such as Serious Sam SE, benefit from being AMD-powered. But for most titles, Intel's CPU is today's high-end processor of choice.

Benchmarks also suggest that AMD's processor label is somewhat overaggressive. The XP 3200+ would have been better labeled, at best, an XP 3000+ 400 based on performance. It's no slouch by any means. However, the 3200+ nomenclature will be undeniably inaccurate when Intel ships its 3.2GHz CPU. The Pentium 4 3.0GHz CPU, in conjunction with Intel's 875P chipset, is a faster overall performer than the 3200+. The XP's lower clock speed and slower FSB hamper performance. That's a shame, since the 3200+ is capable of stable overclocking up to 2.4GHz. It's really a wonder that AMD didn't crank it up more than 2.2GHz.

Additionally, AMD's chip carries a list price higher than Intel's. The XP 3200+ is currently selling at $457 compared to the P4 3.0GHz CPU at $423. In the past, AMD's processors have been the price leaders. The 3200+ reverses that trend. Add to the equation Intel's use of the same fast dual-channel DDR400 memory, and the long-standing cost differential is gone. AMD has lost its title as value champ.

Is now a good time to upgrade? If you need to be on the bleeding edge, sure. Both CPUs offer stellar performance. Nonetheless, if you can wait, do so. Intel fans should hold out for the 3.2GHz P4 rather than opting for its 3.0GHz sibling. Concerning AMD, you may be best waiting for the Athlon 64 to arrive, unless you can swing a super deal on a 3200+ or faster Athlon XP.