Put GPU's under water, now can't launch 3d Apps...

[snedie]

So,

I installed my GPU loop (Dual ati 5970's), and after 4hours of leak testing I have booted the rig. All seems fine and dandy, temps are around 26c across all sensors on the card, highest being 31c on one of the gpu cores (not sure if dodgy tim application, or if that is the core running my display).

However, as soon as I launch a full screen application that so much as puts load on the gpu(s) the system freezes. This happens with heaven bench, furmark, Skyrim, Stalker SOC.

Any ideas what the problem might be?

Notes:

- Fresh install of Win7 + Latest ATI drivers = Same problem
- All temps are good across the board
- Launching Heaven in "Windowed" mode allows it to run just fine.

 

[BEn_2600+]

Possible shorting somewhere on the PCB, do they have a backplate?

 

[steve1873]

A power issue perhaps? Did they both work together prior to being water cooled?

 

[snedie]

Possible shorting somewhere on the PCB, do they have a backplate?

Yup they have a backplate, hard to see how it could short though as the backplate is at least 1/2"mm from the PCB (It has thermal pads for the backside RAM).

A power issue perhaps? Did they both work together prior to being water cooled?

Yup, both working good for a year+, and on a Corsair 950watt for the past months (I replaced a 750watt for the 950 when this happened before*).

*I have had this problem a few times in the past for no apparent reason, it's always something different.

 

[mickyflinn]

Try them individually to see if one is broken?

 

[snedie]

I can't without draining the entire loop, I know it's the best thing to do but it's such a pain in the ass I'd like to leave that at the back of the "to do" list.

 

[Carl_Bennetts]

i would think if one is shorting then you wouldn't get in to windows at all. Can you not disable one card in device manager? Not sure if that works?

Does ccc give you the option to disable one card? never had crossfire so i don't know!

 

[snedie]

i would think if one is shorting then you wouldn't get in to windows at all. Can you not disable one card in device manager? Not sure if that works?

Does ccc give you the option to disable one card? never had crossfire so i don't know!

I can disable crossfire in CCC, I'll give it a shot.

Also I'm failry certain that it is a software issue as the GPU's can physically compute code just fine. I'm running an MD5 hash cracker (yer, let's not discuss aye) with a full char set for a 20 character password of mine (just to make sure it doesn't just "find it" given the fact its doing 7.9 billion passwords a second).

 

[chief barker]

Looks like the cold bug to me.

Flash the BIOS.?

Edit running heaven windowed disables crossfire.

So it's an issue with one of the chips being too cold. I know it sounds crazy but it happens.

 

[Final8y]

Is it an XFI 5970 ?

 

[snedie]

One is an MSI, the other HSI.

Ahh yes the cold bug, that rings a bell.

Well I would try warming them up with the hash cracker first, but because I have done a bodge job with cable ties instead of hose clamps, and scissors instead of cutters, I just had a leak once the cards reached 50c (100% load on all 4 cores, not bad if you ask me): The tubing becomes quite soft at warm temp's it would seem, thus loosening the grip on the barbs.

I should have the hose clamps and cutters on Monday, and I will rebuild the loop then. It's also a chance to reseat the cards as well.

 

[Durzel]

Looks like the cold bug to me.

+1

Had an AMD card that would freeze on launching a 3D app if it was below 30 degrees C, and would fairly constantly freeze & recover in Windows. As soon as it went over 30C it was as solid as a rock.

What you're experiencing sounds identical.

 

[twist3d0n3]

Defo cold boot; below 30deg they act funny. There's a sapphire BIOS which fixes it

 

[snedie]

Defo cold boot; below 30deg they act funny. There's a sapphire BIOS which fixes it

Can this bios be applied to any brand of 5970? As far as I am aware both the MSI and HSI are reference design boards. If so, can you provide direct links for the BIOS, as well as any tools I'll need to do the flash please

 

[snedie]

+1

Had an AMD card that would freeze on launching a 3D app if it was below 30 degrees C, and would fairly constantly freeze & recover in Windows. As soon as it went over 30C it was as solid as a rock.

What you're experiencing sounds identical.

Does doesn't it. The GPU cores sit at around 28c-30c, I'll try warming them up a little with the has cracker and then try running heaven or something.

I'll post back if it works or not.

 

[snedie]

Well what do you know, warming the cards up with the hash cracker first allowed me to complete a run on heaven.

If someone could post links up for that BIOS it would be much appreciated.

 

[chief barker]

http://forums.bit-tech.net/showthread.php?t=185918

May help.

 

[snedie]

New BIOS installed onto both cards and Skyrim launched first time. Pretty glad I didn't drain the loop just to reseat the cards now

Thanks for the help guys.