Re: PhoenixMiner 5.4c: fastest Ethereum/Ethash miner with lowest devfee (Win/Linux)

3060 Ti here - works well, unless...

- I set max overclock at start (only incorrect shares are generated - I need to start lower and increase)
- Epoch changes (same result as above)

Please fix.

happens to me as well (RTX3070), I thought it was a bug on the driver or the bios, I was updating/flashing for 2 hours before figured what you just said lol

Interesting.. thanks for reporting this.. did you try any other miner see if the same issue persists?
so it is not a driver issue, it is a general rtx 3000 series issue

If I start on +1200 mem clock I can survive but I still get bunch of invalid shares , if i start under +1000mem clock and go up I get no invalid shares even up to 1500+for now I start them at +900 and go up to +1200 since you never know when the Epoch change happens and the whole thing goes down if you have it set above +1250 or so memory clock

I have an MSI and a Zotac 3060 ti , Zotac runs hotter and faster , but more prone to invalid shares, same story as above applies to both

I get 59-60 MH on the 900+ mem
and can get close to 64 MH if I go up to +1500 and give it some power like 66% to 68%

Here's what I did,
reset every oc settings to stock
start the miner
wait it to load the dag file and everything
let it run for 2-3 minutes
crank the memory oc +1500 in one go

never got invalid shares until it loads the new dag then the issue comes back
hope this help

that is not a solution..
that is exactly what we are all saying.

you never know the epoch change time and don't want to be by the miner to check it for that.
Perhaps a new parameter to temporarily change the mem clock before the epoch change in phonexminer would fix this.

I didn't know this method has been posted before, I kinda figured all this out on my own lol.
I'm not saying this is the solution by any means, but at least I got it running pretty stable for quite a while, which is better than just start the miner and it just pukes invalid shares everywhere.
like you said, we will have to wait for either the dev or the driver to be fixed or something.