My first, and thus far, last catastrophic failure was when I had my Biostar TF720 A2+, a cheap AM2+ motherboard, paired with a Phenom 9500. The board only officially supported up to 95w processors, and the Phenom 9500 was already 95w out of the box. Not knowing this, I attempted to overclock that processor. I don't remember what speed or what voltage, but whatever it was, it amounted to drawing around 125w. One night I ran an OCCT Linpack test on the new settings, and I noticed the computer shut off. Tired, I slept through it and figured I would mess with it the next day.
The next day, I couldn't get the machine to come back on. After some investigation I noticed the area around the VRMs was burned out. After getting an RMA on the board, along with a new processor (simply assuming the 9500 was dead without testing it, another fail), I noticed the failure also took my 9800GT, the fastest graphics card I ever owned at the time, with it. It took a while to recover from that one, not having the money I do now to throw at hardware.
I tell you one thing though, any overclocks I did with that board I religiously checked out with the overclocker's TDP formula and made sure whatever I was running didn't go over 95w. I also had those little blue Zalman ramsinks on the VRMs as a little extra help.