The first thing you have to determine is whether the driver has flaked out or the miner has died. If the latter, then you simply have to restart the miner.
Now, due to the way that video drivers are handled in Linux (they are kernel modules loaded during the boot process) in order to recover this you will have to stop X, unload the driver (rmmod fglrx) and hope it actually unloads. If it's stuck and doesn't unload, then you have to reboot. If it unloaded, then reload the driver (modprobe fglrx), restart X and all your miners. It's probably faster and easier to reboot.
Now, for your particular issue, you didn't say how hot is "pretty hot" and you didn't demonstrate any awareness of how your GPUs throttle themselves when they get hot. So maybe it isn't actually getting too hot. Provide some more details and maybe we can actually help you.
Topic: Fighting heat; tricking the ati driver to continue mining on crash? (Read 2391 times)
July 12, 2011, 08:24:26 AM
How hot does your middle card get? If it locks its maybe >100°C, not good, especially not 24/7.
What about some extra fans / riser cables?
What about some extra fans / riser cables?
July 12, 2011, 07:15:23 AM
OK so with 3 cards it gets pretty hot. Sometimes the card in the middle will throttle/lock/reset or whatever the driver does. On Windows you get "Ati driver has recovered", on Linux you just get a card that sits at low utilisation or Xorg locked.
What i've noticed is that it falls back to the "Level 0" profile clocks.
Sometimes it mines, sometimes it doesnt, i havent really looked into it that much (need confirmation on what happens when it crashes/locks etc). Sometimes i get home from work and the card in the middle has been near idle for nearly all day. Obviously this is not ideal.
Instead of writing scripts to detect GPU load, heat etc and to restart your mining program/machine, has anyone raised the "Level 0" clocks to the card's default speeds? If the card dies for whatever reason, it should fallback to the set clocks which it can continue mining.
Quote
Adapter 0 - ATI Radeon HD 5800 Series
Core (MHz) Memory (MHz)
Current Clocks : 775 299
Current Peak : 775 299
Configurable Peak Range : [550-900] [299-1250]
GPU load : 98%
Adapter 1 - ATI Radeon HD 5800 Series
Core (MHz) Memory (MHz)
Current Clocks : 157 299
Current Peak : 775 299
Configurable Peak Range : [550-900] [299-1250]
GPU load : 0%
Adapter 2 - ATI Radeon HD 5800 Series
Core (MHz) Memory (MHz)
Current Clocks : 775 299
Current Peak : 775 299
Configurable Peak Range : [550-900] [299-1250]
GPU load : 98%
Core (MHz) Memory (MHz)
Current Clocks : 775 299
Current Peak : 775 299
Configurable Peak Range : [550-900] [299-1250]
GPU load : 98%
Adapter 1 - ATI Radeon HD 5800 Series
Core (MHz) Memory (MHz)
Current Clocks : 157 299
Current Peak : 775 299
Configurable Peak Range : [550-900] [299-1250]
GPU load : 0%
Adapter 2 - ATI Radeon HD 5800 Series
Core (MHz) Memory (MHz)
Current Clocks : 775 299
Current Peak : 775 299
Configurable Peak Range : [550-900] [299-1250]
GPU load : 98%
What i've noticed is that it falls back to the "Level 0" profile clocks.
Quote
Sometimes it mines, sometimes it doesnt, i havent really looked into it that much (need confirmation on what happens when it crashes/locks etc). Sometimes i get home from work and the card in the middle has been near idle for nearly all day. Obviously this is not ideal.
Instead of writing scripts to detect GPU load, heat etc and to restart your mining program/machine, has anyone raised the "Level 0" clocks to the card's default speeds? If the card dies for whatever reason, it should fallback to the set clocks which it can continue mining.
Jump to: