@Kaan3000,
Give it a try using M=1 and 2 workers. I am using Titan X (pascal) from 2016 and get steady 755-760 h/s. I find pascal cards do better with 2 workers and maxwell with 1 worker.
Thanks Yenbus! I started with 2 workers. It wasn't much different than 1 worker -> both gave ~600 to 650. Then I applied 4 workers, it increased. Then I applied 6 workers and it crashed. This way I lowered to 5 workers per card and that seemed to be the highest in my configuration.
Did you use DSTM? What do you get with your sweet Titan X?

(I get solid 700h/s per card with DSTM)