I got this problem twice; In both cases when searching for maximum stable mem freq I can push the 1070 gpus.
Currently running stable at:
Core: -200 (minus 200)
Mem: 765
Power Limit: 59
Hashing: 32.770 MH/s
To tune, I am using nvidia Inspector as it allows me to easily tune individual gpus.
I have this line "call nvidia-tune.bat" in start.bat before the PhoenixMiner.exe line:
::nvidia-tune.bat file content: 7 nvidia cards 0..7 -
:: Note1: card 3 - it is stable at max 625 mem clock - any atampt to go higher will produce errors in minutes after the mining starts
:: Note2: The fans are set to run at constant speed - I have found these values (different for each fan) that keep my cards at around 50-55oC, depending on the ambient temperature
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:0,0,-200 -setMemoryClockOffset:0,0,765 -setPowerTarget:0,59 -setTempTarget:0,1,65 -setFanSpeed:0,45
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:1,0,-200 -setMemoryClockOffset:1,0,765 -setPowerTarget:1,59 -setTempTarget:1,1,65 -setFanSpeed:1,60
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:2,0,-200 -setMemoryClockOffset:2,0,765 -setPowerTarget:2,59 -setTempTarget:2,1,65 -setFanSpeed:2,45
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:3,0,-200 -setMemoryClockOffset:3,0,625 -setPowerTarget:3,55 -setTempTarget:3,1,65 -setFanSpeed:3,40
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:4,0,-200 -setMemoryClockOffset:4,0,765 -setPowerTarget:4,59 -setTempTarget:4,1,65 -setFanSpeed:4,45
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:5,0,-200 -setMemoryClockOffset:5,0,765 -setPowerTarget:5,59 -setTempTarget:5,1,65 -setFanSpeed:5,60
start C:\dev\Tools\Guru3D.com\nvidiaInspector.exe -setBaseClockOffset:6,0,-200 -setMemoryClockOffset:6,0,765 -setPowerTarget:6,59 -setTempTarget:6,1,65 -setFanSpeed:6,45
In my experience the nvidia is not gracefully handling errors when it comes to errors from OC. In best scenario (when slightly overclocked), PhoenixMiner will restart as expected. In worst case, when OC is too high, nvidia driver/configuration gets corrupted and the error as above ends any attempt to mine with the card. The reboot, restart nor cold restart will not make it go away. Only the driver reinstall fixes the problem.
Thank you. I will play with that