All of the articles coming out about IBM/Intel high-k metal transistors say power consumption decreases by 10x in off state and 1.5-2x in on state, and on the entire chip a 30-40% decrease in energy consumption or increase in clock rates should be expected. It is unclear if the improved efficiency is for 65nm or 45nm process, but if the metal process allows 30% increase in speed at 65nm, the shrink would also add to the speed.
There are all of these little unsourced one-liners in articles talking about how Conroe isn't close to maxed out on performance, and how Intel labs have C2D's at 4Ghz at 65nm (though most likely not < 65W), so that would give metal gate 45nm chips lots of room to run.
Of course, why would Intel release the chips at their fastest functional speed when it has no pressure from AMD? As long as Intel has no pressure on the high end, they can hold that 1 Ghz of headroom for when AMD releases a competitive product, then have a 'breakthrough' that suddenly boosts their entire product line by 20%. That's the only reason we see these $180 e6300 (and soon to be cheaper e4300) that overclock well past the $1000 chips.