quote:
Originally posted by jordanspringer
in a while it just auto shuts off, without any warning, without any blue screen. occasionally when the computer turns back on it sometimes gives the "your computer has recoverd from a serious error" message
If "shuts off" = turns off and stays off until the power button is pressed to turn it on again, then the first suspect should be the power supply. It would be unusual for any other component to cause that without the power supply already being pushed to the edge.
You can first try to blow out all the dust with a can of compressed air and visually confirm the power supply fan is spinning. However, you will probably end up replacing it and I would do so before wasting too much time trying other things.
On the other hand, if "shuts off" = rebooting, there are other possibilities.
In virtually every case where I've ran into random reboots without ever getting a blue screen, the ultimate culprit has been CPU cache failure (sometimes a temporary condition due to the CPU overheating). Bad RAM more often seems to result in a mixture of blue screens and "application has performed an illegal instruction" type soft errors. However, if the computer is running Windows XP, make sure it isn't set up to restart automatically on blue screens (read
this).
As far as CPU overheating on a computer that used to work fine -- it should be sufficient to blow out all the dust and inspect the computer running with the cover removed, to make sure the CPU fan and any motherboard chipset fans it might have are spinning and not stuck, no heatsinks have fallen off, and so forth. There really are no invisible cooling failures -- other than improper installation (that would've manifested itself right away when the computer was built).
If you still wish to measure the temperature, it is best to download a model-specific temperature monitor program from your motherboard manufacturer's website, but you can also
try this. Remember that the CPU temperature can raise as much as +20C under load, so you should run something intensive while checking it (any 3D game would work well).
How hot is too hot? There's no simple answer (manufacturers only specify permanent failure point, not "your software starts to crash" point) and it depends on your CPU type, but anything over 60C under load would concern me. Older Intel processors (e.g. Northwood core P4) often stay under 45C with good cooling; for those even 60C could be considered high.
You can test both RAM and L1/L2 caches (by selectively disabling them for memory tests) with
memtest86. Just burn the ISO and boot from the resulting CD.