In about seven to eight years of overclocking (since the Athlon XP series that me and Vonblade (Bryan) used to overclock) I have only ever seen two casualties.
One was a Core 2 duo (one of the cores died) and another Core 2 duo (one of the cores died and the board also blew the VRMS. And, it's terribly likely they were caused by unstable VRMS on the motherboard taking a spike and providing too much juice to the CPU.
In the loudspeaker world 99% of the time when a speaker dies it is because it was not being given enough power. Quick explanation.. When you have say, a 400w subwoofer it is strongly recommended that you feed it with AT LEAST 800w of amplification. An amplifier is only good until about 2 oclock on the dial. After that it has reached its safe levels of power and after that it will start what is called clipping electronically. This clipping can send an over watted/volted signal to the speaker and damage the voice coil.
It's the same with all electronics. If you push them to or past their safety zones things begin to become unstable. For example let's say you took a cheap motherboard with a Phenom 2 in it and cranked the voltage to 1.45v. If the VRMs become unstable and do not regulate that voltage safely then it's highly likely you will get a spike in the power giving the CPU more voltage than you wanted to.
See also - a PSU. They are most efficient and indeed safe between 50% and 75% of their maximum output power. After that things begin to get wobbly and you can end up with a spike (or clip as it's known in the audio world) and that's what will cause damage to your components. I just wish more people understood that when building a PC. Ampage and wattage? you can NEVER have enough. It's all about what you can afford.
When upping the voltage on a CPU you are indeed causing more heat. This goes without saying. However, it's not always that heat that causes the instability. On your CPU around the cores are what are known as edge transistors. These feed data from the core to the operating system. When over driven (putting it simply) they can write a 1 instead of a 0 which is what will cause your lockup, BSOD, artefacts, ETC.
It's likely that this will happen long before you cause any actual damage, which is a blatant visual display that you have gone too far. Then you simply work backward until it stops, with the philosophy that if you didn't damage it at those settings then you're likely not to damage it at lesser settings.
Again, the damage will be caused by an instability in the VRMs or capacitors spiking the power to the CPU.
A last note is this. Why do you think MSI make their GPU boards with 'Military class' components? Hey, they also do it with their motherboards. They then give you free reign to crank the absolute balls off of a 480 GTX core and warranty it. The reason they do that is because they know the cause of failure. And it ain't the core
When Nvidia, Intel, AMD, ATI and so on all design these cores they overclock them themselves. Cores are not designed to be ran at an exact speed, it is all suck em and see once the core comes off the line. They then get it to what they consider to be a solid stable speed. Example. The GTX 580 is nothing but a rebranded 480 with more stuff unlocked and a higher clock speed. A higher clock speed they put on it
