I always run OCCT's GPU tests to monitor my voltages, temps and stability on GFX and CPU OC's. I like this utility for several reasons.
1. It has an auto shutdown feature at temps you can set
2. It has an audible alarm if any criteria is exceeded so I can walk away from screen w/o worrying.
3. While "torture testing" the component, it does not require the use of any additional 3rd party utilities for monitoring temps, voltage, OC, load, etc.
4. You can run infinite test (till you stop) of you cans set specific time duration for test (i.e. 60 minutes).
5. When test is done, it graphs the results over time.http://www.ocbase.com/index.php/download
Using the above you can pinpoint what's going on when the failure occurs.