adding gpu block ruins everything

[andy497]

I finally got my gpu block in and started assembling last weekend. During the install I routed hose through a pci slot, and this area was difficult to seal from leakage. Nothing was plugged in or powered during this time, but a lot dripped along the back pane into the bottom of the case until I got this right. This I had to wipe up/hair dry and eventually wait for overnight to dry. As far as I can tell, the mobo and vid card never got wet, mainly just the case.

Now when I make it to windows, I have normal 2d, but anything 3d is garbage. Launching atitool is fine, but clicking the 3d window freezes the system for seconds at a time (cursor won't even move). The framerate reads under 1 fps (should be about 300), and the picture is garbage.

So here's where it gets interesting. I discovered as I heat the gpu up it gets back to normal. I can launch rivatuner (card is a nvidia 6800nu) and open the temp graph. Temps are very low. 30ish idle/load. Anyway, I declock my cpu, turn off my water pump, and watch the temps rise. This actually takes several minutes. I've got 3d windows open, e.g. rthldld(sp?) and atitool. They start as garbage which freezes my entire system. When the gpu creeps into the 50s however, they start literally thawing and the artifacts start going away. By the 60s (air cooled it ran at 75 C load), the framerate is coming back up and the artifacts are all but gone. Atitool artifact detect will run for seconds at a time without picking one up. Of course by this time my cpu is getting too hot and I'm approaching bluescreen.

At any time, if I switch my pump back on, like clockwork the artifacts come back and eventually the screen fills with garbage and freezes up.

So I'm hoping the dripping didn't corrode some trace on my vid card or mobo. Maybe it's just the thermal expansion between two bad contacts that's causing this behavior? I could try re-mounting the gpu block (swiftech mcw50), but I think the contact there is good given the temps.

I've pulled everything out several times, feeling for dampness, looking for corrosion or burns, and I can't find anything. I'm pretty sick of tearing it all apart. Tonight I'll try another vid card to at least see if it's that or the mobo.

[bigben2k]

Really hard to tell anything from this vantage point.

I'd suspect some kind of short somewhere on the memory bus. If it was me, I'd give this card a thorough wash in alcohol (rubbing alcohol), which should expel any remaining water out. Then a good dry, but without heating it up too much; even though it's not powered, it's still sensitive to high temps (but not as much).

[Wang]

Yeh, it doesn't seem normal... probably moisture related... I got my 6800U a wee wet (no video signal) Blow it DRY. I gave my 6800 -20c airflow during the winter...... it never showed any difference... wether it was 18c core or 70c core temp.
(if it happens when the pump is powered-up .... pump could be causing... leak, interference.) or
The 12v rail of your powersupply could be lacking, I had to replace my PS because of the 6800Ultra crapping out during 3D.

[maxSaleen]

This is a strange one. I doubt it is a short on the card. I can't think of any shorts that could be "assisted" by thermal expansion, and I've never seen a case of it (granted some chips don't behave well below zero). Could you be more specific about your setup so we can have an idea as to how everything is mounted; what you are running in your system (PSU, MB, CPU, RAM, etc) would also help in the diagnosis. In all probability it is the PSU. The addition of the pump probably hurt the 12v rail which would feed into your video card if your are running an AGP system.

[andy497]

Originally Posted by maxSaleen

This is a strange one. I doubt it is a short on the card. I can't think of any shorts that could be "assisted" by thermal expansion, and I've never seen a case of it (granted some chips don't behave well below zero). Could you be more specific about your setup so we can have an idea as to how everything is mounted; what you are running in your system (PSU, MB, CPU, RAM, etc) would also help in the diagnosis. In all probability it is the PSU. The addition of the pump probably hurt the 12v rail which would feed into your video card if your are running an AGP system.

update: I swapped my video card with one from my wifes computer (9800pro) and it worked. sort of. Actually it behaved like it did before, which is why I replaced it with the 6800. 2d looks ok, games look mainly ok with some glitter, atitool shows out of this world artifacting. e.g. error of 1,000,000 in 30,000 pixels etc. I can make this go away by clocking the gpu and memory each down 100 mhz or so, but that's a little silly.

So for fun I put this back in my wifes machine and ran atitool there. no artifacts, no glitter, no nothing. So I'm starting to think it's the mobo. I got my agp slot wet a couple months ago but was very careful about getting it dry before powering on. Could be that I guess.

I thought the psu could have been a problem before, but I put in a new ocz modstream 520. probably not that.

The 12v pump is also not powered by psu. I could get the same effect by running my chiller in reverse (heating the water in my loop to 45 C) with the pump still on.

setup is:
msi k8n neo2
a64 3200+ (stable at 2.5, stock clocks while testing)
1GB geil pc4000 (stable at 270 2.5 2 5 2 1 1:1, what a find!)
evga 6800nu 128 mb
some other stuff...

Tonight I'll try putting the 6800 in my wifes machine and see if it works there. I think it will. It's now hosed in mine. Garbage on the screen even during post, bios screen, etc.

If it works there, I'll rma the mobo I guess. I wonder what that involves...