find the ideal settings for each card individually, test it crash proof for at least 24h.
then put it in a rig along side other cards, maintaining space for proper cooling. ensure cooling is sufficient to surpass the environmental conditions during your 24h test scenario.
if you still crash it's likely a software issue like drivers or windows trickery.
if you are crashing or hanging while running the card individually for testing ideal configuration:
increase voltage which increases heat but makes things stable at higher clocks
decrease voltage which decreases heat but makes things unstable at higher than factory clocks due to sensitivity to quality of power signal delivered by gpu
decrease intensity which makes the card run cooler and cuts the workload slightly
decrease the clocks, gpu vs memory depend on the quality of your cooling solutions
clean and repaste your GPU
i do believe most of this has been said before..
Thanks for the tips. I pulled the suspect card from the rig and running another one of same kind for awhile to see if it is specific with this card. i think the card might be bad :-/