Originally Posted by The EX1
Did you check for memory errors after altering fhe memory timings and clocks? Just because the card ran, doesn't mean the memory wasn't throwing tons of errors out. Use HWINFO to check for memory errors in case you haven't. If you ran the card 24/7 for weeks with memory that was tossing up millions of errors, I can see that causing degredation.
Yeap, I did check it. None of my GPUs run with errors, some of them will get like 6 to 20 in 15min or so. So that wasn't it. I'm running the stock bios fine for couple days now and it seems stable. The old bios is a no go, memory clocks that go high will immediately fail the card.