SPEC's viewperf continues to be a favorite professional-level OpenGL benchmark of ours, and in fact, treats Prescott more favorably than most. Even correcting for improvements in clock and bus speeds, Prescott still holds a notable lead here. We've seen in the past that viewperf absolutely loves higher memory bandwidth, so what we're likely seeing here are positive effects of the large 1MB L2 cache, sufficiently large to offset any negative impacts of Prescott's deeper pipeline.