You can always do things like cram more transistors in to help with branch prediction, that's probably where most of the tiny gains from the last few generations came from.your not going to gain a lot over what youve got, unless you go 6-8 core
maybe we have reached a limit on physical switching inside chips, much like valves on a engine? and they presumeably cant reduce the pipeline or improve IDC someway, so we are stuck at this per core speed?