Xorg 99% CPU Help

[smracer]

I have a couple of machines that will mine for a day or so then the cgminer screen will lock up.

I ssh into the machine and cgminer was killed and Xorg is using 99% of the cpu.

I have seen this happen on multiple machines.

All running 11.04 Ubuntu from USB stick.  Cgminer.

Anyone had a similar problem and found a solution?

in win 1200w power supply
890fxa-gd70
sempron 145e
2gb ram dimm
5X5870's
11.04
AMD 2.4 SDK

mouse is not plugged in, screensaver off.

[smracer]

I should add that all my miners have run rock solid for 2 months straight before I saw this Xorg problem and it only started after I downclocked my 5870's to 800/150/.959

I am putting them up to 825/300/1.000 and will see if that fixes the problem.

It must be it since that is the only change I made besides switching them all to 220V which would make no difference.

[wknight]

When your at Xorg 99% that means a GPU Hung.. If you downclock and undervolt it will take a bit of work to find the sweet spot.

[smracer]

Is there a way to tell which GPU hung?  I looked through DMESG and couldn't tell which one it was.

I would rather just change the settings on 1 GPU then change all 5.

[wknight]

You could do a screen -r and you will see cgminer running.. usually you will see the gpu that has no hashes submitted or speed. It will look dead or be dead. Maybe even sick.

[smracer]

Yeah I had one machine that GPU 1 and 3 reported SICK so I was able to just raise those clocks a little.

[wknight]

Yup thats all it takes But the only way to get rid of that 99% most of the time is a coldreboot

[smracer]

Yeah i tried killing the pid.  doesn't work. 

So I get to restart and re-write my /etc/network/interfaces file everytime.

Is there any way to get my USB stick to write that file?  sudo sync doesn't write it to the usb stick.