[spencerjford]

ok, so yes to all of these?

 - Running windows 10 with latest fall creators upate
 - AMD graphics card with latest driver 17.11.4
 - desktop resolution is 3840x2160
 - lav filters is set to copyback
-  madvr settings are stock but with direct 3D 11 ticked and nothing more.
- using external LAV filters in MPC
 - upsclaing everything in MADVR not using any 1080p display modes, sending a 4k image to TV.
 - not using FES
 - tried both h264 and HEVC mkv rips.
 - not using reclock
 - using built in sanear

any chance of a screen shot of your CONTROL J screen?   

 NO, you actually didn't get most of that right.  
but yes on h264 and h265.   using DX11 copyback on lav filters.  
i am using 1080p screen, I don't have a 4k screen.
I can use any combination of risers upscalers i could before creators update.
I do use refresh rate changing (the madvr built in on)
I do NOT use reclock or need it. 

I can go with Sanear or with out it.

https://imgur.com/a/hN6nD  

 then you dont have a similar setup to me or others with this problem, thanks for you input but it doesnt really help. I'll try to clarify again.

Is there anyone out there using KODI DS in 4k with and ATI RX card that doesn't have any frame drops when not using reclock and using the current set of drivers / filters / MADVR and WIn 10 Fall creators update? 

I do find it funny on on doom9 on madvr forum you say this is well documented, as in many have the problem, but in truth, it's only your setup and your system doing it.  No one else has confirmed the problems you have.   but you act like everyone has it, since no one says there setup is fine.  But the opposite maybe true, it could just be your setup and no one feels like going through the trouble to post contrary evidence? 

Now it could also be, that with the update, someone else changed (like task manager did behind the scene) for you and now your system runs a bit hotter/draws more power/  could be a power supply consistency problem.   

I'm not trying to say to you that you are not experiencing these issues,  but i am saying that you can't say it's well documented when no one else is reporting the same thing.