Interesting test. Comparing the two 32M times, you get 6.66% performance improvement at same clockspeed (similar to the 6.5% the xbit lab article quoted as average performance gain). 6.66% at 3600 is ~240 mhz, so theoretically a 3.84 Q6600 should equal a 3.6 QX9450, but would be nice to see tested. (1M superpi showed 8.6% improvement, but thats more prone to variability error)