Weird Problem With 7950s

[daserpent1]

I have 3 x Msi 7950s that I use for mining LTC. These cards are only 3 months old. Actually the whole rig is only 3 months old.

2 of my cards are in the 4x slot and 1 in the 16x slot (which is the primary card for desktop display). If i am mining with my 4x cards, and then I turn ON the 3rd card, the pc just freezes.

The kill-a-watt shows the pc drawing 250W or so, down from 700W.

My PSU is a Seasonic 1050W Gold rated.

When the PC freezes, the display goes out. But the fans and LED on the mobo keep running.

I don't know what i could do to fix this.

[mufa23]

Pretty sure it's the PSU. I'm actually having that problem right now with one my of rigs and had to RMA the PSU yesterday.

[BitcoinEXpress]

I have 3 x Msi 7950s that I use for mining LTC. These cards are only 3 months old. Actually the whole rig is only 3 months old.

2 of my cards are in the 4x slot and 1 in the 16x slot (which is the primary card for desktop display). If i am mining with my 4x cards, and then I turn ON the 3rd card, the pc just freezes.

The kill-a-watt shows the pc drawing 250W or so, down from 700W.

My PSU is a Seasonic 1050W Gold rated.

When the PC freezes, the display goes out. But the fans and LED on the mobo keep running.

I don't know what i could do to fix this.

Sounds more like driver crash to me.

Are they all the same model?

Do you have a mix of 7950 Ref Models (two six pin power connectors) mixed with 7950 that use the 7970 ref design (1 eight pin, 1 six pin)?

Post your startup script and model numbers.


~BCX~

Added: there is nothing wrong with mixing the models, just takes some tweaking on the startup that's all.

[daserpent1]

Yes all are the exact same models. And the rig worked fine with all 3 for 2 months, its only now that its giving this problem.

I found one way to run 2 cards every stably. I plug in all 3 cards but i don't power the first one. I power only 2nd and 3rd. That way it keeps running for days with no problems.

But 1 card is being wasted this way.

Its something to do with the voltage i suppose because when the card in slot 1 boosts to 960Mhz core clock (draws more voltage).. it gets shut down.

[lonesoul]

Something to try would be to rotate the cards, that way you can find out if its a bad card or not - from what you said though it does sound like when i have a driver crash. for some reason when my driver crashes it ends up screwing the wireless card, which means manual reboot to get it working again. happens occasionally so nothing really to worry about.

I would rotate the cards (card 1 move to slot 2, card 2 move to slot 3, card 3 move to slot 1) and confirm that the new card 1 and 2 work, if they both work fine then you know its not the cards that are playing up, if it blows up time to work out which card is causing it and fix/replace.

sorry i cant be more help.

[daserpent1]

Tried swapping the cards around and still i get the same problem.

Either the driver crashes (which it shouldnt) or the PSU screws up. Idk

[BitcoinEXpress]

Did you upgrade, change or reinstall any drivers, cgminer versions or add any new hardware?

Windows or Linux?

~BCX~

[Oldminer]

Power or mobo issue for sure. Considering the general reliability of the PSU I'd be looking at the mobo. Latest chipset drivers, BIOS? PCI-E slots configured correctly in the BIOS for max performance?

[daserpent1]

I have not really changed the drivers since day one. It worked perfectly for 2 months. So never felt like changing anything.

I am sure its not just cgminer, because sometimes it happens when i start playing call of duty too.

I have had the mobo replaced already. Got a new one from them and that too the latest rev version of it. So mobo is not at fault here.

It is not over-heating either. I have monitored temps each day and nothing abnormal.

it happens ONLY when the core clock changes to 960Mhz or 1000.. etc from 300Mhz.

If there was a way to keep 960Mhz as default core clock, then i think it could work fine.

[Oldminer]

it happens ONLY when the core clock changes to 960Mhz or 1000.. etc from 300Mhz.

If there was a way to keep 960Mhz as default core clock, then i think it could work fine.

Have you checked to see if the speed setting on the PCI-e slots has changed? (ie are they set to 2x, 4x)

Also, does it matter what card you lastly start mining with that causes the crash? Have you tried swapping the cards around to see if you can isolate the issue to a vid card or a PCI-e slot?