Killed two 890fxa-gd70s, my fault or bad luck?

Question

Not sure where to start but I had an MSI 890fxa-gd70 (amd) board for 6 months in which it was overclocked since nearly day one, all good till recently. This past friday the board failed. Durring prime the system shut down, fans and all, and a burning smell came from the northbridge area. Once out of the case it seemed like the choke below the 8pin power connector, or at least VERY close to that area was the source of the smell, NO visual signs at all on either side of the board of anything at all. The top of the choke had maybe 2-3 dots from heat im guessing. Overclock wasnt even "extreme". Just a new setting I wanted to try.
Well I got my rma number monday and picked up the replacement board tuesday and it failed durring prime the SAME way as the last one. This time absolutly no signs, the chokes had no dots. Originally started the computer, basic function tested, rebooted and applied mild OC settings and prime tested, died in less than an hour of total use and I was only maybe 10mins into prime. Im not sure the replacement was even remanufactured but maybe a "couldnt duplicate issue" and thrown into a white box to be given out. There was a capacitor right above the 1st pcie16 slot than was scratched with a slight dent and a scatch near the bottom of the north bridge heat since from someone removing their video card i guess.
So what the heck, is this bad luck or am I causing the issue? I am suspicious that I have too many fans on the mobo. My fan arrangement changed recently about the same time all this started. I have 4 120mm fans for radiators(2 kuhler 120 and 2 corsair 120), kuhler 920 pump, and the corsair mem twin fans driven off the mobo. Previously when I was overclocking I only had 2 120mm fans, pump, and mem fans on for a while but recently off, and back on when I added the other 2 fans for the radiator. Could this be partly to blame? Regardless Im planning on removing all fans from the board next time except the 2 kuhler ones being controlled by the pump.
I know overclocking is running outside rated settings but no one buys a board like this to run at stock settings, thats why they advertise all the OC features and even have a clear cmos button on the i/o pannel.
Keep in mind i have done gaming at 4255mhz cpu and 2900ish nb, but wasnt prime stable very long, while the board failed settings where as followed:
overclock that both failed on was 4140mhz (230x18)
nb was aprox 2900mhz 1st failure and 2760mhz (aprox) 2nd time
1.40v cpu vdd
1.25v cpu-nb on 1st failure and 1.21ish on 2nd failure
1st failure had a few other voltage settings up a bit, 2nd failure all others where auto except memory at 1.655v memory spec
power supply is a corsair gs800 watt:
 http://www.newegg.com/Product/Produc...orsair%20gs800
mother board:
 http://www.newegg.com/Product/Produc...82E16813130274
setup:
rosewill blackhawk case
Msi 890fxa-gd70 mobo
amd 1100t x6
antec kuhler 920 modded for 2 120mm radiators and resivout
corsair 8GB dominator gt memory 1866mhz 4dimms
msi gtx560ti twin frozr ii
1 corsair f120ssd and 1 7200rpm seagate 1TB drive
corsair GS800 psu
8 120mm fans and twin corsair memory fan that comes with dominator gt memory
So am I causing the issue or a case of bad luck and not to worry? Im really wondering if its the fans, thus why the asus crossfire has so many extra power connections on the board. I just figured the board had all those fan connections it must have been able to handle it, all but one was being used, maybe they didnt plan on 120mm fans being used in most of them.
I do wanna say Im glad msi has been friendly and am so glad they allow me to drop off and pick up since I live/work nearby. Today I dropped off the bad rma unit and they said they would give me a "new" one, probably a new refurb, but they told me to pick it up tomorrow because they are gonna burn test it overnight so I dont have any issues with the next one. Them stress testing it was nice. I would discuss my setup with them so we could both learn from it but the people I dealt with didnt seem to care how it went out.
Thanks for input guys, dont wanna waste my time or theirs smoking another board. Hope the next one does well or may need to look into a crossfire v. I have always liked MSI and also use their gtx560 ti.

Answer 1

manual settings. I wish I wrote my settings down other than just a saved profile, guess I should have also had my bios backed up too
I have played with a bunch of different settings but 4025mhz (230x17.5) was prime stable for one hour so i used that as my stable oc with only 1.40v on cpu vdd, think it even did an hour on 1.38v too if I remember but left it at 1.40 for good measure. Which if not mistaken is the max normal range the cpu is rated for and used durring normal operation when it ramps to 3.7ghz on stock settings.
I have done gaming (crysis2 w/dx11 patch and cod blackops)  on 4255mhz with 1.43v. Would easily play till I got tired after an hour of gaming, but would fail prime within 7-8mins...Would shut down and have fans all running and needed to cycle psu. I have always wondered if that was a psu issue since both games use all 6 cores but same time prime IS A KILLER so figured was just the cpu
When the board failed settings where as followed:
overclock that both failed on was 4140mhz (230x18)
nb was aprox 2900mhz 1st failure and 2760mhz (aprox) 2nd time
1.40v cpu vdd on both times
1.25v cpu-nb on 1st failure and 1.21ish on 2nd failure
memory run in the 1840mhz range on 4 dimms
1st failure had a few other voltage settings up a bit (some where most 2 ticks above auto but sb 1.22v)
2nd failure all others where auto except memory at 1.655v memory spec and like mentioned cpuvdd 1.40, even nb was auto