SICK/DEAD GPU within a few hours of mining LTC - need some opinions

[mxc0bbn]

Hardware:
MSI 990FXA-GD80, 8GM RAM, AMD Athlon II X2 3.4GHz, (2x) Sapphire Radeon 7950, Seasonic 1000W PSU

Software:
Windows 7 Pro-x64
CGMiner 3.1.0


CGMiner CMD Line:
cgminer.exe --scrypt -o <pool url> -u <user> -p <password> --shaders 1792 --gpu engine 1020 --gpu-memclock 1575 --temp-target 75 -I 20 -w 256 --lookup-gap 2 --thread-concurrency 21712 -g 1

From the beginning I have not seen greater than 610 kh/s per GPU with an average around 575 kh/s.  I thought I would let it run for at least a day or two in order to establish a stability baseline then start playing with some of the settings to try to get it higher; however, After only a few (6-8) hours GPU1 goes SICK and then eventually DEAD.  Only happens with GPU1.

Current WU: is ~1100
HW: 0 on both cards (although didn't notice if there was anything during the DEAD period)



The temperatures never climb above 80C and are usually closer to the target of 75C (+-3C)
GPU0 has a monitor attached
GPU1 doesn't.


I have a total of 4 Sapphire 7950s that I bought to put in the MB, but I wanted to try with two first and then ramp my way up.  the actual card identified as GPU0 never stops mining at a 575+ rate; however, the card identified as GPU1 always goes sick and dead within a few hours.  Up to now I had been using the same two PCI-E slots, but this last time I took the same card that went DEAD the last time and put it in a different PCI-E slot.  I'm currently waiting on the results, but I figured I'd ask because maybe someone out there can see something in my settings that is causing it regardless of which PCI-E slot it's in.

Any ideas?

on edit:  I have swapped out the card identified as SICK/DEAD with the other two cards I bought to make sure it wasn't a bad card, but have gotten same result so far

[biged214]

Have you tried lowering your memory clock speeds or core clock speeds? I had a similar issue with one of my gigabyte 7950's and I had to lower the clock speeds unit it stopped doing it. I know it's not as fast but at least I have non stop mining going on. If it happens again even after you made your changes on the hardware side try running it at default clock settings to see what happens.

Also I switch to GUIminer -scrypt version and I am getting the same results as directly with cgminer but it seems more stable. GUIminer also has a auto start feature that I like.

[nightengale]

Kept having issue with 6870 that would go dead, am trying reverting the clocks back to factory and we'll see what happens... Am also wondering if intensity has an effect on this issue?

[phluxx]

I use CGWatcher to monitor my CGMiner and restart it periodically/if one of the cards dies.

http://manotechnology.blogspot.ca/p/cgwatcher.html

[Cryptocables]

Any ideas?

You could potentially have a failing PCI-E slot. It happens. I would try them on another system just to make sure that it definitely isn't the cards, and this way you can see if you have a faulty PCI-E slot or not on your Mobo.

[nightengale]

After reverting back to the default clock speeds, my DEAD gpu is running again.

(scrambling to find some wood to knock on)

[bclcjunkie]

can the card be seen under windows device manager? i have a similar situation but my card is no longer visible under device manager... haven't troubleshooted it yet but probably failed..

Hardware:
MSI 990FXA-GD80, 8GM RAM, AMD Athlon II X2 3.4GHz, (2x) Sapphire Radeon 7950, Seasonic 1000W PSU

Software:
Windows 7 Pro-x64
CGMiner 3.1.0


CGMiner CMD Line:
cgminer.exe --scrypt -o <pool url> -u <user> -p <password> --shaders 1792 --gpu engine 1020 --gpu-memclock 1575 --temp-target 75 -I 20 -w 256 --lookup-gap 2 --thread-concurrency 21712 -g 1

From the beginning I have not seen greater than 610 kh/s per GPU with an average around 575 kh/s.  I thought I would let it run for at least a day or two in order to establish a stability baseline then start playing with some of the settings to try to get it higher; however, After only a few (6-8) hours GPU1 goes SICK and then eventually DEAD.  Only happens with GPU1.

Current WU: is ~1100
HW: 0 on both cards (although didn't notice if there was anything during the DEAD period)



The temperatures never climb above 80C and are usually closer to the target of 75C (+-3C)
GPU0 has a monitor attached
GPU1 doesn't.


I have a total of 4 Sapphire 7950s that I bought to put in the MB, but I wanted to try with two first and then ramp my way up.  the actual card identified as GPU0 never stops mining at a 575+ rate; however, the card identified as GPU1 always goes sick and dead within a few hours.  Up to now I had been using the same two PCI-E slots, but this last time I took the same card that went DEAD the last time and put it in a different PCI-E slot.  I'm currently waiting on the results, but I figured I'd ask because maybe someone out there can see something in my settings that is causing it regardless of which PCI-E slot it's in.

Any ideas?

on edit:  I have swapped out the card identified as SICK/DEAD with the other two cards I bought to make sure it wasn't a bad card, but have gotten same result so far

[BBQKorv]

If you are using 1x riser try with a 16x and vice-versa. Updating BIOS might help and PCIe latency helps sometimes.

[GSnak]

Your mem clock is too high. Stick with 1500. Also, try TC 22400.

edit: Side point, if you're using the dual fan Sapphire you can undervolt to ~1080 with most which will save you 70w per card. I'm running mine 1000/1500/1080v. One I have to run at 1100mv or it gets sick and dies.

[nightengale]

I went back to stock clocks, and the card has been running fine since. It's nice to get the extra hashrate, but not at the expense of stability or prematurely killing the card.

[alxs]

I agree with the mem clock being too high.  1500-1550 might be upper limit for 7950, also the intensity could be too high, try 15 or 16 rather than 20.