@dimafren, so I followed your advise to under-clock using Bios and flash that into the GPU and it worked. It is still about 5-8 watt more than Windows but it is now more comparable. For those that are having same issues as my rig, you will still need to set the Core/Memory Clock & Core State in the web application.

Now I am testing if I can get it stable with just 850 mV. I will post updates once I get some hours into the rig.