Hi Tim,

I fully agree with all previous ideas to put the cpu into idle mode
wait window, sleep API, timer, scheduler should help.

But if your computer actually has to work for long stretches of time,
there also is a slight risk that the high temperature will result in
(undetected) calculating errors and not lock-ups or crashes.

Since I somtetimes have to do some long running data-mining,
I usually chicken and get me some hardware leeway:
I use the "overclocking possibilities" to underclock
the computer: about 10% underclock will give you a nice
safety margin (the risk will rise exponential,
but only very near the failing temp).
I also have most machines in a separate room (less noise)
and if they are running maxed out I open the cases - better
cooling always helps. If security rules allow it, I sometimes
put the machines with open cases in the cellar: nice, coool surroundings.

Perhaps I "care" to much about the machines, but OTOH returning
correct data translated into more money...

my 0.02 EUR

thomas