A simple example is shadows in FSO. Even a card as recent as an RX480 has trouble achieving good framerates at particularly busy scenes (Icarus, for example).
Now, granted, FSO isn't exactly super-optimized, but it's not like we're doing anything particularly weird when setting up our frames and issuing them to the driver. Especially since we cleaned up the interface to use OGL core instead of OGL compatibility, we should be able to run better on AMD, but we very clearly are not.
Take a look at this:
This is Icarus on a GTX 1070 at 3440x1440. The framerate stays at or close to 100 FPS at all times (except when a frame is late because a new asset had to be loaded in).
Now compare to this:
This is the same cutscene (albeit with an experimental build with autogenerated lensflares, please ignore those), running at 1920x1080 on an RX 480. Its framerate flucuates wildly; dipping far below 30 frames per second.
Now, granted, the 1070 is a much more powerful card. But even that power differential can't explain the performance differential seen here. Since the 480 is an otherwise very capable card and definitely no slouch when it comes to DX12 or Vulkan-based games, I think we can safely blame AMD's implementation of OpenGL for the performance issues the cards have.