Optional GPU initialisation delay

[Strunczez]

When building DAG for ZIL OVC do not kicks in - it was fixed by last update, but on FLUX it still crashes rig on building lockup tables.

Last message from miner before PSU OVC kick in:

Another restart, one PSU is off, rig continues on 3 GPU:

Auxilairy GPU OVC protection turns it off and 3GPU's connected to it

@rigelminer Can you fix it ASAP? Please consider delaying tasks between GPU's = 100ms

[rigelminer]

I can't fix it ASAP but I think you can circumvent this issue by locking the core clock of your cards instead of setting a power limit. There's apparently a short delay between the GPU hitting 250W+ and the driver reducing the clock frequency to bring it back to 250W - by locking the core clock, which most people prefer to setting a power limit, you should be able to prevent the GPU from ever going above 250W in the first place.
Run a single GPU with --pl 250 --lock-mclock 5000, note down what its core clock is, and then use --lock-cclock <frequency you noted> --lock-mclock 5000. You can leave --pl 250 just in case. Let me know if that helps.

[Strunczez]

@rigelminer It helps, no more OVC crashes (tried 20 x miner restart), but it is not solution for me.
Locking clock gives me 50 Mh/s LESS Hashrate on whole RIG and consumes same wattage as cclock +200.

On current energy prices it is not for me :(
I buy energy on market. When energy is cheap or PV is working well I mine, when it is expensive my miners start only for ZIL then they shut down.
When auxiliary PSU crashes, rig stuck and I need to restart it manually, when main PSU crashes whole rig restarts.

[rigelminer]

Locking clock gives me 50 Mh/s LESS Hashrate on whole RIG and consumes same wattage as cclock +200.

add --cclock 200 then, it will consume less watts allowing you to increase the --lock-cclock value to get to 250W with higher hashrate

[Strunczez]

@rigelminer I did not know that I can Lock clock and Add core clock offset.
Now it works as it should :)