Quote from: KillrBee on Today at 02:42:31 PM

Quote from: ayman_abm on Today at 02:25:56 PM

@KillrBee

I have a rig of 3090s (6 cards) running on HiveOS. I couldn't reach 11326 mem clock (+3000 in Hive) max could reach was +2900 and after several minutes miner crashes. Cards cooled up using AC direct cold air to them.

I think you are using Windows ?!

one more thing, I used +1200 on Core, didn't try +1600 since Ethhash a memory hungry algorithm

Yeah, Linux and Windows are different beasts when it comes to the GPU clocks. +1200 would probably be just fine -- essentially I'm running at .75 - .80 Volts. I was very lucky with how well my card overclocks, I can get that memory up to 11400 for a while, but eventually the card throws an error (even though temps are well controlled).
Try pushing the intensity up and keeping the memory overclock to +2600 (I'm running XInt = 3108 right now and I can keep the memory clock down to a level that the miner doesn't drop or gpu hang for 24hr+ {which is 11326ish for my card}). It's taken A LOT of experimenting and testing to find a stable and fast hashrate/share rate, and it continues to be refined every day and every release. I am probably spending way too much time tweaking vs. doing my day job, but I'm a performance freak and always have to find the useable limits of everything. :-)

What happens if you use --lock-cclock [[1200,1200] ? for me its same hash rate but way less power used as 1500/1600, and temps less for sure since its less power consuming