Does anyone actually have any evidence that the benchmarks are wrong? If raytracying is enabled I can definitely see there being that much of a pref gap.