You'd have to be utterly asleep at the wheel to not see the situation we face today when comparing the AthlonXP to the P4. It's not a cut and dry affair anymore, and the performance metrics used to make meaningful decisions just arent straightforward for a number of reasons, both good and bad in my view. Some of these are....

-The AthlonXP's now sport a "Performance Rating" system. While I truly understand AMD's reasoning, it's created alot of confusion. AMD's current PR system is supposedly index'd to the relative performance of their previous Thunderbird CPU's. The POINT however seems to be to help "show" the XP's performance roughly relative to P4 CPU's. Indeed the AthlonXP's on a clock for clock basis are mostly faster, in some areas considerably so. But the PR rating system is already showing it's age, relative to the advancements in P4 L2 cache and raw speed.

-The P4's Netburst architecture is the first real radical departure from standard x86 design. Much more deeply pipelined than the P3 (or the AthlonXP for that matter), and with different inherent performance properties, the P4 at initial launch was a disappointment in many areas. It was really bad timing for Intel for the initial launch to go against a truly competetive effort from AMD. This I feel is the PRIMARY reason (more than price) that AMD had such huge gains in market share last year.

-Since there is architectural divergence, there's also divergence in chipset support. Although companies like VIA, SiS, and Ali (not to mention AMD's efforts here themselves) are stronger in their offerings than they were in the past, it hasnt been a bump-free evolution. Things are excellent for AMD at this point, primarily with VIA and SIS, but it's been a somewhat troubled march. Intel still dominates chipsets for their CPU's despite offerings from VIA and SiS, and although the PX4 and SiS6xx are strong, Intel rules their own roost inasfar as chipsets are concerned, and continues to be the preferred chipset/subsystem by vendors to place underneath Intel's CPU's.

The gist of this, as we stand here today, is that clockspeed isnt the whole story, and that side by side comparisons of the architectures here (AMD Palomino vs P4 Northwood) arent the straightforward affairs we've been used to. This has of course been the situation for the last year, but I reiterate it for a reason.

Intel's latest Northwood P4 2.4 400bus CPU regained the overall performance crown, via the raw speed approach. With AMD's fastest currently available CPU, the XP2100 which "really" runs at 1.733Ghz not too terribly far behind, it's still Intel's show here. Of course the 2.4 P4 had to have a nearly 700mhz lead to pull it off, but the REAL story is how well Northwood CPU's are scaling in speed. Aircooled overclocked Northwoods are reaching 3.0Ghz plus, and some of the more radical efforts with watercooling are going considerably higher.

With the adoption of the 533bus (133fsb quad pumped), things are only going to heat up more. The adoption of a faster bus should increase overall memory bandwidth slightly (and the P4 architecture really LIKES bandwidth), and allow Intel to avoid ridiculous clock multipliers as their speeds scale up through this year and next.

Of course AMD's Thoroughbred cores are imminent as well. To what speeds these scale, and how well they overclock will be indicators of how competetive this year is going to be, but for now, Intel is firmly in the drivers seat, and the ball is in AMD's court in my view.