I'd also worry about the over clocking being the cause of his problems. A chip designer explained to me once why over clocking is risky. I'm not a chip designer, and I'm reporting what he told me from memory, so I may be slightly off, but here goes:
Sometimes it is completely safe. For instance, suppose a chip maker sells a chip in three speeds: 500 mhz, 750 mhz, and 1000 mhz. You have a 500 mhz chip. It might actually have been manufactured to be a 750 mhz, and passed all the tests at 750 mhz, but the manufacturer had a shortage of 500 mhz chips and an excess of 750 mhz chips, and so marked yours as 500 mhz and sold it as such. Congratulations! This is a chip that is safe to at least 750 mhz.
Sometimes, however, your 500 mhz chip may be one that was meant to be 750 or 1000 mhz, but failed their tests at those speed. It was fine at 500 mhz, so they sold it as such. This is a chip that is risky to run at 750.
How can you tell whether you have one of the safe chips or one of the unsafe chips? You might think you can simply try it. Over clock it, put it under the heaviest load you've got, and see if it appears fine. If it seems to work, you might assume it is one of the safe ones.
The problem with this is that failure can be too subtle to notice. It might be that at 750, your chip loses the lower bit of integer multiply if one of the operands has a run of alternating bits longer than 18 bits, or something really obscure like that, something that your testing is not likely to hit. (Look how long it took the Pentium FDIV bug to be discovered).
When the chip manufacturer is doing their test to see if the chip should be a 500, 750, or 1000, they know where the weak points are and how to stress them. They are also testing before the chip is put in its case, and so can access test points on the chip and see electrically that the signals are fine at the danger points.