Quote from: Jamsan on April 03, 2018, 11:35:51 AM

Quote from: xunl on April 03, 2018, 02:48:17 AM

Quote from: Jamsan on April 03, 2018, 02:40:03 AM

Quote from: xunl on April 03, 2018, 01:37:37 AM

FYI:
I have mod/flash some setting into bios so I can automate the startup process. But I found it is very inconvenient because 1) everytime you want change the setting, you have to mod it again. 2) down time is too long if you have lots of GPUs on a rig and many rigs. command line argument (software way vs. hardware way) is far more easier and quicker.

If you're tweaking memory clocks / power settings, stick with Overdriventool. You can create a profile for each unique card, apply the appropriate overclock / undervolt settings, and then use the overdriventool command line + appropriate parameters to run in the same batch file as your Claymore startup file. This will apply the appropriate overclocks, and saves the headache of modifying the BIOS every time you have a simple change you want to test.

Thanks a lot
I heard Overdriventool but haven't tried it. Many people mention that you need to uninstall WattMan (AMD setting) for Overdriventool to work, otherwise, things will reset, is that true?

I've never used Wattman, so can't speak to it (but I believe it's installed by default with drivers?), but I've never had a problem with overdriventool running and sticking. Even if it doesn't persist through a reboot, it'll be applied again before mining starts, so it'll stick one way or another.

I think WattMan is installed by default with AMD setting and drivers. You can choose customize installation and uncheck AMD setting to not install WattMan.
What you suggest is covering 95~99% of time, with some exception. I have RX 580 setting revert to default. I am assuming either WattMan (90% confidence) or (10% afterburn 10% confidence) overide the claymore OC command arguements. The system keep running w/o error (no reboot) Claymore runs (no crash/restart). Whatever is in the script doesn't get triggered and thing gone bad!

Today I found another unexpected:
One of my GPU has been finding incorrect shares for 6 hours. This GPU never had any incorrect share for 1 year. Out of 157 share found, all of them are incorrect. This never happened before. 6 hour of electricity is wasted for nothing. This is worse than crash or hang. If the miner crash/restart during the 6 hour spin, the problem would be fixed.

Eventually, when enough people confirm 18.3 is more stable and people can use claymore OC setting arguement to control RX 580. I will swiitch. Thanks