Topic: One GPU always goes SICK no matter how many cards in rig [SOLVED] (Read 4694 times)

Thank you.
Thank you.
Thank you.
Thank you.
 

I was running 3X R9 290X's in my rig for days without a problem. Then, all of the sudden, one start saying it is sick and getting no work. The memory was 150 and I couldn't change it. I plugged in the monitor and boom. I'm back in action.

Can't explain why it just stopped working like that, but I'm thankful for a solution.

I have 4 Sapphaires and similar motherboard (last digits are 55 not 80). I saw the problem when the monitor was actually connected. Then I disconnected the monitor and restarted it and the problem went away. Then an hour later (this is all in the past few hours, I just set the system up) I had an accidental power disconnect and bam one card is hashing at 80 again, the rest are 729. Pretty frustrating. I think this is power related. I get a driver error (Load MM Driver has stopped working) before this all happens with the screen freezing for a second and flashing. Cant find the issue. Did a few cold reboots and problem is there. Swapped cables on PSU and swapped the risers and no change.

I have three cards on risers and one directly on MB. I dont know which one is 0-3 in cgminer but the one failing is always 1 and sometimes it says SICK in the hashrate part.

I also run logmein to remote desktop and its Win 8.1 machine with Catalyst 13.11

I just went to control panel and attempted to disable GPU 1 and now its gone from the list. Not sure how to bring it back.

PERFECT -  I had the same exact issue as you and plugging in my monitor the the CPU0 & it FIXED My Sick - DEAD GPU. Ughhhh took me too long to find this post to figure it out... I must have re-configed every possible setting before I found this post.

Thanks for writing up your fix!

As it turns out, there wasn't anything wrong with any of my cards.  As I added cards to the MB, booted up and tested I never swapped out the video cable so that it always is attached to the card that would be "GPU 0"

If the monitor is not attached to the first card on the board it always goes sick...but if the monitor is attached to the first card everything works perfectly.  I now have 4 cards on the MB mining at 610+ kh/s with no issues.

If this is something that is obvious to others then I apologize for my ignorance...While I've built many systems before (and some with dual GPUs) I've always attached monitors to both cards and I've never really used them for this type of activity (mining) so I didn't realize that there was any importance to where the monitor should be attached.

Thanks for the responses.

Also, grue, I would agree with you as so many often point out that Intensity 20 is pretty darn high, but in this case I didn't think it was a root cause because I've been mining on all these cards (on two separate MBs) using Intensity 20 for weeks...so that couldn't be it.

As for the clock speeds being so low, yes they were...but my attempts to change the settings were being ignored by the card....

Anyway, so at least in this case the solution was just to attach the monitor to the first card (GPU 0)

20 intensity = too high

also, your memory clock is way too low for mining litecoins.

update:

I think the main cause of the SICK problem is that I am running at Intensity 20 and the card that goes sick has settings that I can't seem to modify:

low voltage: 0.950
Engine Clock 500
MemClock : 150

When I back off to I:18 or below the SICK card does run but at a fraction of it's potential kh/s rate.  So it seems like the solution would be to up the clock and voltage settings...

The problem is that when I try to do it through CGMINER it says that the driver reports ok, but doesn't actually change the settings

Any suggestions?

update: If I leave just 2 cards one still goes sick immediately upon cgminer start.  Was not happening until today.

I have 3 Sapphire 7950s in a MSI 990FXA-GD80 and one always goes SICK.  I have 4 cards total and have tried swapping out the SICK one for a different card.  I've also tried swapping out the same card to a different PCI-e slot...same results...

I tried running two cards on extender cables and no matter what one always goes sick.  I've swapped out cables and ports...different one goes SICK, but one always does.

Also, the one that goes sick always starts out either mining 0kh/s or <80kh/s before it eventually goes SICK.  When I look at the GPU settings the Engine and Mem Clock speeds are way low (150/500) and no matter what I do I can't change them.

CGMiner reports that the driver accepted my new settings, but no change in clock speeds.


Lastly, this rig was running for 15 days straight with two cards cranking out over 620kh/s.  I only powered it down to add another card and blamo!  Any suggestions?

Mike


solution on my last post