i've been tweaking it and I noticed 8193 thread concurrency is critical! 8192 boom you are down to 600+, 8194 the same. so stay at 8193
then I tweak the engine for every settings
of the 3 41bios cards I have, one is "good" and two are "bad", the good one can run at 1072 GPU with 735Kh/s and the 2 runs at 1068 GPU with 726Kh/s
one of the bad card somehow keeps dropping Kh/s and I woke up this morning at 620 region
So now I have powertune the bad card to +20 and seems like its more stable, shall wait another few hours and see any difference
My only issue now is that the two "bad" cards seem to run in tandem (i.e. if one improves at the expense of the other) I have managed to tune both up to 728 and 730 Kh/s) but both cannot run at the optimal speed at the same time. One will have to drop to 722Kh/s if I tune up the other. Anyone has this kind of experience?