5970: Did I kill the master GPU?

[oxident]

One of my two 5970 is driving me nuts: A week ago, CGMiner started to crash randomly. I figured out that this was caused by "GPU0" which is the master GPU of the 1st 5970 (ATI Reference card). I got the entire system working again by lowering the freq to 600/300 and setting the intensity to 5. Everything just on this core, of course. The remaining (slave on this card and both on the 2nd card) were still mining at 830/300, intensity 8. Temperatures were all in range (~80C on the first and ~65C on the second card which uses an Arctic cooler).

So far, so good but after a few days, the fan on the first card started to produce a heavy noise and I instantly saw a decrease in fan speed and of course an increase of temperatures. The system crashed a few seconds later. I was able to get the fan working again ... but now the first GPU of the first card disappeared in the device manager (Win7 x64). The 2nd GPU and the entire 2nd card are working perfectly. Even the two red LEDs light up for a second when starting the system.

Do you think there's anything I can do?

I thought about flashing a modified BIOS which will decrease the clock just at POST because I think, the first GPU is failing at this point. But RBE doesn't recognize the first GPU ... just the slave one :-(

Maybe someone has an idea...

[dooferorg]

Can you try the card in another system and make sure it's not some software issue? Certainly sounds like the hardware has failed in some way though

Is the card the primary video card? I'm amazed that you still get a display if it is.

I would certainly try and see if you can restore the factory bios for the card.

[oxident]

At the moment, I'm still searching for a system which can handle such a card ;-)

I've already tried swapping both cards (or running with just the defective one) but the effect remains the same: Only one GPU in the device manager (and GPU-Z) and the monitor only displays a picture if it is connected to the 2nd DVI port, even at booting stage!

I will try an alternate operating system as soon as possible but at the moment, I also doubt it is some kind of BIOS problem. A long time ago, I flashed the card (both GPUs with the according master/slave BIOS) in order to get better OC possibilities. Maybe there went something wrong ... but the card's function wasn't effected in any way.

There is one thing which sounds a little bit curious to me:

I can't control the fan speed of this card and I guess this is because the fan is strictly controlled by the master GPU (which isn't visible to me). But the automatic fan speed control is still working. I can clearly hear the fan spinning up when there is high load on the GPU.

I mean, as far as I know, both GPUs are connected to an internal PCIe-Bridge which is then connected to the PCIe-bus of my system. So the 5970's internal bridge would enumerate both GPUs at boot time, just as my system does with regular single chip GPUs, right? So maybe the 1st GPU just isn't responding fast enough (or with the right devid).

But how could I reset or reflash the BIOS if I can't "see" the card?

[The-Real-Link]

  In Afterburner and Vision Engine, it seems that fan control lets you manage the speed on one of the two displayed cards (1 for each GPU's core), that's normal.

  I have one of my own 5970s that mines fine on I think, the secondary core near the exhaust.  If I start my mienr with the first core, the display driver instantly crashes and I can't guarantee the system would even remain stable. 

  I'd imagine it's just possible that one of your cores died.

[oxident]

Yes, I doubt that but I always thought that the entire card will die if the master GPU dies... So maybe I should try to flash a master BIOS to the slave GPU in order to get at least the fan control back?

So I would effectively convert my 5970 to a 5850 

[Remember remember the 5th of November]

Your card's fan needs to be re-greased.

[oxident]

Yes, already did this at the very first time it started making these noise. The fan is working again ... but the first GPU is dead and therefore I have no chance of controlling the fan (because it seems that it can only be controlled by the first GPU).