Might be worth trying older versions of fglrx and see if they cause the same problem. Another remote possibility is something faulty with a card. Try another card with the same chip/stepping to see if it is reproducible.
Wait, so if AMD doesn't know what's wrong there, then how does fglrx manage to function right? There must have been some code in it that manages those cases...
If it really is a hardware bug, it might be fixed in newer cards. Jerome said he was mainly running tests on r600 and r700 cards, so who knows if it's still on issue on the newer ones.
Last edited by smitty3268; 07-14-2012 at 05:22 PM.
He harms nobody if he tells where to find the erratum list.
And if this is the results then I have not harmed anyone.
One the other side there is a popular movement who wants to force people in there way to communicate.
AMD Catalyst driver is more than 30 milion lines of code....... only an insane guy could think AMD really knows what they're doing inside that code after years of "put the brick on top of the previous brick and make a prayer that everything goes well." .