Well, my impression has been that WinXP is not an overclocker's paradise. But I wouldn't know from first-hand experience. I only surmise from the number of posts I've seen in various places about problems that overclockers have with Windows 2000 and XP. Win2K and WinXP are heavily centered around the concept of monitoring and controlling individual process and overal system performance. I suspect that overclocking runs afoul of OS design features that are there to serve those functions. My personal recommendation for a first step would be to stop overclocking until AFTER you've sorted everything out. In other words, even if ceasing to overclock does not cause the system to revert to proper behavior, I'd still suggest running at stock speed until you do have the system running properly. Once you have everything else sorted out, then try overclocking.
One thing I can tell you for certain is that, in all of my destructive testing experiments with stock machinery and speeds, the only things other than deliberate sabotage that I've found that are capable of clobbering WinXP are bad devices / bad drivers. When you stick a poorly designed kernel mode device driver onto this system you are guaranteed immediate and truly adverse results. And I have seen even approved drivers (yes, signed drivers) do this! I guess this must be due to an interaction amongst various devices and drivers on a given system. I haven't seen anything to suggest that a given signed driver could cause system failure all by itself. I've had NVidia drivers that performed perfectly on one system absolutely flatten another system with a different device and driver complement. The failed systems have always contained an unsigned driver somewhere. You might want to check your system over for unsigned drivers as at least one step in the diagnostic process. (Have you run the sigverif.exe utility, or are you aware of any unsigned drivers on the system?) Anyway, WinXP warns you by default if you are installing such a driver. That warning can be turned off. I suggest leaving it on and heeding it. When installation of an unsigned driver is indicated by need, it's certainly best to test the machine thoroughly after the driver has been installed.
One thing that is particularly problematic is that there are software packages that may not be suspected of installing "drivers" on the system which do, indeed, install kernel mode drivers. Anti-virus, software firewall and CD writing software have been particularly problematic in this regard, and I have direct experience and evidence with examples of each kind.
Good luck!
- Collin
