Tell me more x
, there is a new speed test available. Give it a try, leave feedback!
dslreports logo
    All Forums Hot Topics Gallery


Search Topic:
share rss forum feed


Lynnwood, WA
·Frontier Communi..

intermittently defective video card

Click for full size
Any idea on what may be wrong with this video card based on the attached image? Memory? GPU? Drivers? Cap(s)? Something else?

It's a XFX HD-677X-Z5F3, but I cannot just return it for exchange since I bought the last one.

When I initially installed the card (no drivers) it worked fine, albeit very slowly with default Win7 video driver. Installed the latest driver from AMD's site, and all hell broke loose, as per attached image.

So I uninstalled the drivers, reinstalled it from the install CD that came with the card, but that didn't make any difference. That was last Tuesday or Thursday. I left the PC powered ON, although sleeping. Yesterday I wanted to do more troubleshooting, but to much of my surprise, the card is working fine now....
(I didn't do anything that could've made any change)

Any idea on WTH may be going on here?

Card is currently in a Dell Precision 380, but once I get a Molex to 6 pin adapter, I'll move it into a Dell 745 that can take only single slot cards.
Wacky Races 2012!

Mountain View, CA
My 1st guess would be a hardware (GPU-level) issue, possibly something wrong with the RAM on the card itself or something worse (GPU core failure, things like this), or GPU overheating. Given how shoddy video cards are today (in my experience) and how many I've seen fail at the hardware level, this is my first choice.

My 2nd guess would be a corrupted driver or a driver that isn't properly working with your exact model of card. One such possibility for driver corruption could be unreadable LBAs/sectors on a hard disk (where the OS chooses to ignore the I/O error, which AFAIK Windows does not do) -- SMART attributes would prove this -- or alternately bit rot (undetectable without a checksumming filesystem). If you've uninstalled and reinstalled the drivers -- or better yet, done a full format/clean install of the entire OS (I strongly recommend this) -- then that can be ruled out.

My 3rd guess would be a broken/buggy video driver, specifically if the system is being put into sleep or standby or "hibnerate" states. Most issues relating to sleep/standby/hibernate happen after a resume, and these are all driver-level bugs.

Regarding #1: Windows Vista onward makes determination difficult, since many pieces of the desktop and windows are hardware-accelerated (i.e. 3D, since the entire desktop/UI is now DirectX), which means hardware/driver-level features are heavily used in comparison to XP (where most everything is GDI+ and only a few things are offloaded onto the hardware). This is why on Vista onward you'll see video card memory being heavily used (140MB+ in some cases) in the desktop with no games running, while on XP you'll see maybe at most 20MBytes in the exact same conditions.

You've got almost (?) perfectly diagonal lines showing up; since GPUs today focus primarily on 3D, the way you get a rectangle (e.g. a window) is to align 2 triangles then map a single texture across both of those rectangles and draw onto the texture (i.e. the texture is your 2D "screen" or "window" or "surface"). Possibly different parts of the window (menu bar, etc.) use separate polygon pairs -- I do not know.

What's shown above seems to indicate one of the triangles' texture mapping may be broken.

I cannot explain the "weird jagged lines" especially in the upper right of the picture -- if you moved a window around or jiggled it and it left "trails" of its contents, then that may explain it. Combine that with my above description of combined triangles and it might explain the rest.

The tools I know of for testing video cards (RAM and other features) are all Windows-based tools, which means if your GUI isn't usable/readable these utilities are essentially worthless. The tools I've used are OCCT, Artefact Tester, and Video Memory Test. The latter tool requires very specific checkbox/options to be set else you can get false positives.
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer.


Florence, KY
reply to aurgathor
It seems likely to be a cracked solder joint that is temporarily conducting again but may intermittently fail again at any moment in the future.

Personally I would return the card for a refund if they have no stock now, unless you feel like keeping ownership and filing a warranty claim with XFX and waiting for a replacement card, BUT if it isn't malfunctioning when they receive it, they could just return the same damaged or defective card to you.

Also if they had a marginal manufacturing process it could be that an entire batch or more of the same model are likely to have the same problem, so when I can I try to avoid getting a product replaced with the exact same thing


Lynnwood, WA
·Frontier Communi..

1 edit
reply to aurgathor
I ran all tests mentioned by koitsu -- aside from a an few non-supported video modes everything ran fine, although during OCCT sometimes the GPU was near 80C. (idle was 41C)

As for the cracked solder joint -- that's an interesting theory. The card is very unique that it is a single slot card with 5 mini DP connectors, and if I return it for a refund it would be very difficult to get another comparable card, at least until DP becomes more commonplace. I think I'll probably put it into the freezer for a while to see if that can help to repro the issue, and try to get a replacement from XFX if it does repro. Also, the orientation may matter a lot with a cracked joint, so I can try varying that, too.

The card spent about 2 hours in the freezer (in a plastic bag) -- and then worked fine afterwards.

So I guess I'll just have to rely on the XFX warranty if it does it again. Maybe I can figure the conditions that can cause the messed up display. Or maybe it never happens again, but I won't hold my breath on that.
Wacky Races 2012!