There are two sides to this issue as I see it:
To the average consumer shopping for a PC clock speed alone is considered to be a valid measure of performance. This is of course wrong. Due to the design of the P4 with its "hyper pipeline" ( 20 stage pipeline) enabling it to hit high clock speeds, but reducing performance per clock versus the Athlon's conventional 11 stage pipeline AMD is going to have a hard time competing clock for clock. This means that even though the Athlon greatly outperforms the P4 at the same clockspeed only the educated computer buyer knows that. AMD has thus (at least according to this article) decided to borrow a page from the old Cyrix playbook and bring back the PR rating. I can see the reasoning behind it, but truthfully I think it's a bad idea. My main objection is the choice of benchmark(s) used to obtain the PR rating. Now if what Tom says is true that a 1400 = PR1600 then the benchmark is very conservative. In office and FPU engineering type situations the 1400 will beat even a 2 GHz P4 pretty handily. If the benchmark is Quake 3 then the Athlon 1400 only is the equal of about a 1.5 P4. Now remember that the Palomino includes a prefetch unit, which accounts for much of the P4s dominance in Q3, so the gap will narrow there and widen elsewhere.
The part that makes me angry is the AMD requirement that the true clockspeed not be shown anywhere - this is wrong. That's how Cyrix got themselves in trouble.
Intel, though never using the PR rating themselves, does play marketing games like this though. Examples being the SX vs DX, the original Celeron (A) vs P2 performance, and "netburst, hyperpipelining" etc.
