Re: I know several people who are running the Phenom II 940 chip at 4G on air with a 5% increase in vCore.
The first thing you need to know about overclocking is that it doesn't apply to volume shipments that need to adhere to lifespans covered by the warrantee. Processors slowly degrade over years of use, so vendors have to guardband the frequencies. That's what gives overclocking headroom, but after degrading, these processors will eventually start yielding failures, or stop functioning altogether.
There's also the fact that raising frequencies and voltages raises power, sometimes significantly. AMD can't launch a product with 200W power envelope, as it will have as much appeal as lead laced bubblegum. Just raising the voltage and frequency by 5% turns into a 16% increase in power.
Re: I think wbmw's read is informative - AMD could do higher speeds, but with marginal gains for a lot of work. I don't have the knowledge to evaluate how much effort those kind of things would take.
You can put 20 experienced designers on speed path optimizing, add 6 months after the first production tapeout, and eke out another 10% of frequency gain. The next 20 designers and 6 months might only get 5%. If that's all AMD has, then 20 people for 10% (or 40 for 15%) might be worth it. Then again, if you're Intel and have dozens of projects (3 Itanium, 4+ "Core" tick-tock, 6+ Atom and SoC, and 2+ Larrabee designs in parallel, etc.), it might make sense to use that precious expertise differently. That is to say, Intel obviously has more people than AMD with the expertise to optimize speed paths, but probably not as many as they have total design projects in the company. Intel has decided to put their expertise into strategies like tick-tock to have a larger total count of designs, as opposed to having fewer designs that clock higher. And personally, I think that's what has given them such a strong product portfolio, capable of hitting more markets with better segment coverage.