SiSoft Sandra's CPU Arithmetic tests show yet again what we would epxect - the deeper pipeline hurts the performance of Prescott versus the Northwood core. In 'pure math' tests, or those where the branch predictor and cache size improvements have no impact, we expect to see Northwood continue to trounce Prescott.


Of course, those types of tasks aren't all too frequent. FutureMark's PCMark2004 test gives Prescott a very significant edge. Even taking into account our slower-speed Northwood used for testing, and correcting for the lower clock and bus speeds, Prescott still holds a small edge here, illustrating that Intel's tradeoff of cache speed for cache size may indeed have been a smart move.