the best way is to have a temp probe monitor core temps that's set to trigger a normally open relay hooked to the power good signal going from the mobo to the psu. This is what i've heard is very common and effective... Eheims are more reliable than most. It's just safer that way. your probably not gonna fry ur chip in a sec if the pump dies. probably not in 5 mins. if u know what your max running temp is when every thing is all good and well and u add an offset to accomodate for fluctuations in ambient temp and set ur software monitor to that value then most likely it will be very capable of shutting ur machine down b4 it gets too hot. I don't have any numbers on the growth rate of the temp of a core with a good copper block on it so u are just gonna have to think about it urself and maybe run a few tests.
But always to be on the safe side have something to monitor it via hw. Be it ur bios or a DD5 sort of thing.
It's far cheaper than replacing a proc that went critical.
Cyph3r
oh btw... this is only for those with non pelts...
software could never keep up with the temp growth rate of a pelt if either the pelt died or the pump