I posted in another topic about this “glitch”. I was able to get the glitch once (using NBMiner 37.3 @ ERGO) with moderate OC settings (2800 memory / 115w) and it was very stable for a few days. I was using a combination of a 3060Ti with a 3060 on that rig, but I had to remove the 3060Ti so I had to turn it off and thus start all over again.
I could not get the glitch again with somewhat “normal” OC settings, so I started pumping the memory up to 3100. Still no glitch. Ultimately I used 3200 in order to achieve the glitch. Of course, this is very high memory OC and I knew that it would be highly unstable upon switching to ETH.
As expected, DAG verification failed and it was crashing every 3-10 mins randomly. So, I started experimenting with GMiner 2.54 and the various options, mostly regarding dag, kernel and intensity.
I have found a good combination that seems to run pretty stable -so far. I used these settings in GMiner:
–safe_dag 2
-k 4
-i 96
“k” sets the kernel to #4 and “i” sets the intensity to 96%. So, it reduces the load on the card and of course reduces the hashrate (from 51.3 to 49.09 Mh/s). But, it seems really stable! I will report again in a few days, but in the meantime I would like some more people to try experimenting with the kernel and the intensity options in GMiner.
This is my OC settings atm (running stable for over an hour now) :