Moreover i also use the --eth_config parameter into the TeamRedMiner extra config, in particular this: --eth_config=B640 basically this fix the GPU Tuning allowing TRM to skip this step.
The B mode is known to lower the power consumption, maybe this can help you a little bit with your rig in general
Thanks, but with B640 all my cards do around 0.5-1 MH/s less than without this parameter… i think i am in the maximum for this cards (2 of them cant handle with 900MHz memory). Now I am trying to get lower voltages and consumption.
I have some issues with invalid shares. I tried to run with default voltage, but i still having some invalid shares. Can you confirm that invalid shares means “too high memory clocks”?
In some Telegram groups many people say yes, others say no… so I could try 895MHz, loose 0,6MH/s but stable rig without invalid shares?? is that recommended?
I thought that the maximum about memory clock is when you get GPU DEAD error, not invalid shares.
So, finally, is invalid shares related with too high memclk or too much undervolt?
That’s partially correct.
For sure if you get gpu dead THAT is its limit, but dude, you GET PAID from your accepted share which means, if you run stable a gpu without invalid you get paid more than if you have lot of invalid.
Remember if you get 1 invalid every 100 shares you are paid for 99 share, which means you lose 1% of gain.
Move this stuff to gpu hashrate, if you have your gpu mining at 55,14 and you get paid for 99% it’s exact the same if you mine at 54,58.
This is true for 1% invalid, you have 19, so i’m assuming that gpu sent 1900 shares, on that hashrate it should be about 2 days long mining right?
TLDR:
Believe it or not, 19 invalids in 2 days are A LOT.
I know, and I focus on non-invalid shares rig.
Only for curiosity, hiveon pool don’t recognize this invalid shares, i dont know if the pool is taking that invalid as stales
So, in your opinion, should I set lower memory frequency to avoid invalid shares? or the problem come from CoreCLK or undervolting?
Typically the second one is the HW on miner log, which is the red one you can see in the gpu row in the hiveos table, those with the not-allowed mark,
This kind of share is invalid BEFORE being sent to the pool, this is the reason why the pool doesn’t see them.
The rejected are those shares ALREADY sent to the pool but maybe it’s too late, maybe they are technically invalid so the pool REJECTS them, and you can see them in the top of the worker page like so, and in the miner log too (R):
you should see them right there (i’ve 0 of them so i painted)
About your problem specifically, I paste here my oc of my 5700xt which is samsung as well, but Sapphire Nitro +, take a look to compare, maybe this helps you, it could be the core clock, the memory voltage as well, maybe just silicon lottery.
If it’s stable at 895 i’d say set mem at 895 and save some power
Everyone from Turkey,
The settings I have been using for several months do not cause problems.
You need to get the bios to get good results.
I wanted to share with you.
can you share your memtweak?)