Longer scientific explanation. Solid state physics for the win.
It's not an ohmic conductor, so ohms law doesn't strictly apply. The power used is proportional to (speed in mhz/stock speed), and to (overclocked voltage/stock voltage)^2. Efficiency is near enough constant, so yes power consumption and therefore temperature are closely linked to voltage. However this is not the only thing that kills chips.
Heat kills chips by encouraging the lattice impurities to move around, if the carefully placed doping atoms shift too much you've rewired your processor and it doesn't work so well. This is a slow process, solid state diffusion at only 100 degrees takes a very long time. However the chip is fragile, and slightly rewiring it may prove fatal. Rather depends on luck.
There's a feature called electron migration. This is why a chip running at 1.8V will die however cold it is. Inside a processor are lots of little tracks, 45nm apart. Each time the chip is die shrunk, the tracks move closer together and efficiency increases, generally so do transister switching speeds. However, the closer they get, the more likely an electron will jump from one to another.
That there is insulation inbetween makes little difference, electrons will still jump across. It's quantum tunnelling, which is not popular in chips at present. This leads to currents and potentials in the wrong place. Now, this effect is strongly aggravated by voltage, and I'm personally certain by temperature as well. Both are forms of energy the electron can use to jump the gap, so both contribute. The common view in overclocking is that migration is independent of heat. Either the community is wrong, or temperature has no effect on electrons. Either way, the effect definitely does damage, and is tied to voltage.
It is also very strongly tied to (directly proportional to) current, as this is how many electrons you push through. That the D0 chips use less voltage and more current for a given clock speed then C1 suggests they will suffer more from this effect, and will particularly dislike being severely overvolted. However they're still too new for any data to be available.
So, yeah. Quantum tunnelling rapes your chip. However the consensus for a long time has been that it takes both temperature and voltage to kill a chip, and it'll tolerate much higher values of one if the other is kept low. It'll also need less power at lower temperatures for a given stable clock. Enter water, chilled water, thermoelectrics and phase.
The conclusion I think is that its not going faster unless you cool it more or feed it more volts. Bear in mind that you might manage to join the elite few who have managed to kill a processor if you go down this road, but then it'll probably still be good for a couple of years at which point the hex core chips will be out and hopefully affordable. If you can fund replacing it, thrash it
