I have been testing all ports on the MB. PCI-E ports 2 and 3 are broken. When I using different carts on these ports using less watts and not stable. When I use GPU cards on PCI-E ports 1, 4, 5 and 6 i have stable hashing rate and using 100 watt. You can see here
https://i.imgur.com/cbNRs5O.jpgAre you using -mvddc for underclock your memory?
Hmm, I wouldn't be so sure that a PCI-E port is broken. Especially if it's your first 16x one. Try not using different cards, but only one for which you are sure it is performing well on another PCI-E port. Then start narrowing down the possible reasons - Riser, Riser cable, don't overclock the card -> test in 2 different ports with the same Riser.
I guess you meant memory overclock, not underclock. After trying a number of different tools, at the end Claymore -mvddc isn't working for me so I'm using Afterburner. For the moment core underclock and memory overclock is being auto applied properly on every reboot.