OCCT passes, AMD Driver Fails... how?

[zapp22]

friends I thought I was fairly bright/skilled on PC 'issues' but this one has me completely beaten.
I had a sapphire radeon 5570 1gb low power card for my combo win7 ultimate x64 / core-duo hp HTPC/Lab-mule system and it was just dandy. for 2 years , no issues. so I got smart and replaced with a well-reviewed MSI Radeon 6570 1GB not knowing that they are essentially the same/identical.

however, after half dozen removals of ALL video drivers, reinstalling time after time the 'new' 6570, I still have driver crashes, bsods, artifacting, etc etc that I've never ever seen on this machine. so I thought the card was bad. Read the sevenforums tutorial on OCCT and downloaded, ran it for hours. no errors in Directx 9 mode, no errors in directx 11 mode, no errors period. until I ran something simple like google chrome or IE 10... then BSOD, as before.

these errors, bsods, seem to have no pattern. it can be something or nothing. sometimes it is in flipping some web page to fullscreen. sometimes it is a video [flash/shockwave, jw, whatever]. sometimes its simply executing anything - windows explorer, a blank-page IE10 startup [i have it set to just open a tab, no webpage].

Re: OCCT - I have a question: if I run CPU [not GPU] tests, there's this cool fractal image that runs... it moves and does what fractals do. If I run GPU tests with error checking enabled etc, I get the pane that the animation is supposed to run in, but the image is static: its not updating. is that normal? seems pointless to me to open that second window.

at a loss here. a beaten man.
somebody school me...

 

[Faladu]

If you put the 5570 back in, problem solved or continues?

 

[zapp22]

the card is now sold/shipped, but I would not have chanced it. not after vaporizing all known drivers and starting over from scratch. it wouldn't have told me anything.

I've now run OCCT for many hours, both directx9 and 11, no errors found at all.
I can make it bsod, however, by randomly opening challenging things in browsers while OCCT is running.
it leaves little trail behind. no events in the event log since the event kills the event-watcher :sarc:

i spoke with a friend yesterday, far smarter than most of us, a EE IT director that still writes a lot of code for fun and profit. he says he sees this on other systems and has not found a cure. in other words, a stable running system; upgrade the vid card from card a to card b in the same family [his example was also radeon but much higher-end cards] - remove registry entries, remove drivers, update directx, reinstall drivers, etc. one card will create errors while another may not. He thinks the issue is with GPU-latching browser features [hdwe acceleration].

its an unsatisfying answer.