While we would argue that being disappointed with Prescott's performance is perhaps a slightly severe reaction, we would certainly agree that Intel's latest creation is perhaps not as impressive as we had hoped. Especially if we see that a 3.06GHz Pentium 4, running at a mere 533MHz FSB, manages to keep pace, or sometimes surpasses, the 3.2Ghz Prescott which runs off a 800MHz FSB and has a 150MHz clockspeed advantage.

The bottom line is that, clock-for-clock, Prescott performs just about as well as Northwood. Slightly better in some cases, slightly worse in others - but overall, describing the two as relatively equal isn't a stretch. Intel hasn't appreciably 'devalued' the MHz here, although they certainly haven't improved its value, as we hoped the larger caches, better branch predictor, and other various improvements might (like they did with Northwood). Instead, as a result of the deeper pipeline, we have essentially a processor with larger caches that performs like one with smaller caches (but that Intel hopes will scale better).

The real test of the wisdom of Intel's design will come in the near future, as Intel attempts to ramp Prescott's clock speed up to 4 GHz. Though Intel denies any difficulties, the deeper pipeline is irrefutable evidence that the process shrink alone wasn't resulting in the usual clock scaling benefits we've come to expect.

We may be observing the first signs that Moore's Law is becoming increasingly difficult to obey.

Dan Mepham