Troubleshooting: XFX Factory OC'ed HD 4890 artifacts unless underclocked to reference

[RevOne]

Just joined this great forum today (long time lurker, first time poster ) and I could use a little feedback if anyone feels kind enough to help.
Fair Warning: Bit of a wall of text coming, but I want to ensure I provide all the pertinent details I know so far to rule some things out. So, bear with me

Here's the card in question: XFX HD 4890 XT("Extreme Edition") http://www.hardocp.com/article/2009/10/05/xfx_radeon_hd_4890_extreme_review/

A little background:

• Bought it secondhand off ebay back in (iirc) June 2011.
  

The Problem:
Shortly after updating to the Catalyst 12.2 drivers back in March 2012, I started getting GPU driver crashes/restarts in Windows, heavy artifacts, BSOD's, and screen lock-ups; sometimes when simply idle or doing light web browsing, but mostly whenever I fired up any games; opened GPU-intensive programs like Photoshop, Flash, or Maya; or initiated any type of streaming video content (i.e. Youtube, Hule, Netflix). BSOD's consisted of an error citing "atikmpag.sys".

Troubleshooting done so far:

• Dismantled card. Inspected & cleaned cooler and PCB carefully. No burnt wires/pieces, swollen caps, or otherwise broken parts. Card visibly appears to be in perfect shape. Removed old thermal paste, applied Arctic Silver 5 to memory chips and GPU, and reassembled.
• Conducted a completely fresh install of Windows 7 x64 thinking maybe there was a corrupted driver installation that Driver Sweeper/Driver Cleaner Pro and software like that couldn't fix.
• After the new re-install of the OS and a roll back to Catalyst 11.12 drivers... everything seemed perfectly fine again. Ran cooler because of the new (better) thermal paste. Even ran and passed Furmark testing with no errors. Games, videos, programs - all working fine. No crashes, lock-ups, etc. For a while, at least.
  ... Then the artifacting started happening again about a few days or so later.

The (Temporary) Solution:
After trying everything I could think of, the only thing that seemed to work was underclocking the card to reference HD 4890 frequencies in the Catalyst Control Center. I personally did not overclock this card (can't speak for the guy before me) because I didn't want to push a secondhand unit that didn't have a warranty on it. And I was perfectly fine with playing games with it at it's "out-of-the-box" settings. So I left it at stock (factory overclocked) setting. Temps idle are 50° - 55°C and never peaked over 75°C in my case under full load (Furmark benching/testing included); so I don't really think it's heat damaged.

And XFX only put a very mild 25MHz/100MHz OC in place over reference, so instead of running at the reference 850MHz Core/950MHz Memory, it was at 875MHz Core/975MHz Memory.

In Summary:
I found that as long as I leave it at reference clocks (850/950), the card will run fine all day long with absolutely no issues whatsoever. Runs everything I throw at it, including grueling Furmark bench tests. The moment I set it back to it's default clocks (875/975), however - the instability, artifacts, BSOD's, etc. returns in full force.

Finally, I decided to download the card's stock bios from TPU, here: http://www.techpowerup.com/vgabios/...FX&model=HD+4890&interface=PCI-E&memSize=1024 ... to see if maybe a reflash would clear things up (thinking maybe a corrupt bios). Tried it, but it didn't change a thing. Still have to run it at 850/950 to avoid instability.

My Question To The Members:

Badly binned part or card on it's way out? Anyone ever experience anything like this with one of their GPU's before?

I wouldn't think a measly 25MHz factory-set OC could yield such a major difference in stability.

Also (for self-indulgent, educational purposes ):
1) Can a card's bios become corrupted and cause such issues?
2) Can bad drivers corrupt a card's vram?
3) Can overclocking other parts of your system adversely affect your GPU (HT Link; RAM timings, etc.)?

Next, I'm going to try increasing the voltage just a tad in the bios from it's stock 1.312 to perhaps no higher than 1.35 and see if maybe that makes the stock clocks more stable. Don't get me wrong - it works perfectly with the clocks set to reference, but it'd be even better if I could get it back to where it was working at manufacturer's specifications.

Any help would be sincerely appreciated. Thanks in advance!

 

[Law-II]

Hi

"Badly binned part?" this may be the case, however you do not have a full history of the card so it is hard to say, most components degrade over time.

"card on it's way out?" the card was launched in 2009, it would have had a limited lifetime parts and labour warranty from xfx - however most reference models only had 2 to 3 year warrantys, the card is getting on a bit and it may; that I am aware of at this time,other members of the TPU community may have a better idea of these

"Anyone ever experience anything like this with one of their GPU's before?" Yes; on occasions these issues arrise

"Also (for self-indulgent, educational purposes ):"
"1) Can a card's bios become corrupted and cause such issues?" Yes this is my understanding at this time,other members of the TPU community may have other answers to support this

"2) Can bad drivers corrupt a card's vram?" this link may shed some light on the matter
Look here - http://www.playtool.com/pages/artifacts/artifacts.html

"3) Can overclocking other parts of your system adversely affect your GPU (HT Link; RAM timings, etc.)?" if you raised the PCIe interface that may have a negative effect. Parts may degrade faster, however should not effect the vga card directly; more indirectly as the system will produce more heat which effects all components.

"Next, I'm going to try increasing the voltage just a tad in the bios from it's stock 1.312 to perhaps no higher than 1.35 and see if maybe that makes the stock clocks more stable." depending on the artifacting experienced this may or maynot have a positive result, having access to the GPU voltage may correct an issue with GPU however as the VTT is locked your options for memory related issues are limited to underclocking.

nb: the 12.xx serise of drivers appear to have issues

atb (all the best)

Law-II

 

[erixx]

Considering it's age, and that 25Mhz of overclock is not worth making the wall of text (hehe) even higher, I suggest running it at standard speeds until its EOL. Electronics degradation is not rocket science or exact science to be exact... If there is a (non visible, like bad caps) degradation, how are going to know and ...even if, then, what to do?

 

[RevOne]

Thank you so much for stepping in to offer some advice on this one folks Really appreciate it!

@ Law-II Good info there. Many thanks! Helps me to understand how all this stuff works even better (still new to custom building).

@ erixx Thanks Yeah, I figured it was worth a shot to see if perhaps some more experienced individuals could perhaps lead me down the path to anything I may have overlooked in trying to figure this one out. The reason I was trying to salvage it was because I was hoping I could maybe turn around and sell the thing later for about 30 or so bucks to put towards a better modern card. Oh well...

I'll probably just keep it running in this system as it is (since I wasn't planning on OC'ing the card anyway and that whole 25MHz difference isn't anything to fret over). Still does what it needs to without a problem. Once I make the move to my next build, I'll likely keep this system as a back up or gift it to a family member. The card can always be replaced later on if need be; but running at reference speeds isn't the worst outcome for an aged unit (could've been much worse - as in dead; so a good outcome altogether, considering).

 

[RevOne]

Sweet XFX RMA'd the card and found it faulty, so they're sending me back an HD 6850.

Thanks everyone for all your help