Here are a couple of snippets of my box running the Beta 27 DX11 version to show the effects of the AMD Global PowerEfficiency setting in the Crimson 16.6.1 drivers on the GPU performance of my FuryX. Note: Both instances were ran between the same 2 airfields using the same air route and alt to make as close of an apples-to-apples comparison:
1st snippet is w\ PE enabled..........
2nd snippet is w\ PE disabled...............
As you can see, w\ PE enabled running Beta 27 Dx11 version the GPU clocks are being clocked down pretty heavy by PowerTune 2.0 thru AMD OverDrive due to driver reading the Beta graphical calls as low level (what PowerTune uses to calculate graphics load to then determine GPU power\clock speeds needed to match graphics draw call load = performance per watt) which is causing the low FPS, double loping graphics motion I posted on earlier. Note the GPU temps are low....due to PowerTune cutting power based on the driver read graphics draw calls sent to it from the Beta software.
Once PE is disabled the driver instructs the GPU to run at full clocks and power setting as set in driver (AMD OverDrive w\ PowerTune 2.0 bypassed) on the same Beta graphics draw calls as sent in the 1st snippet, the GPU performance is pretty much pegging the GPU clocks at 1050 and the FPS is pretty much pegged at 80 FPS (as controlled by AMD FRTC using AMD FreeSynch) and graphics motion is butter smooth.
When Beta 21 Dx11 version was being used the graph looked like the 1st snippet BUT the FPS was like the 2nd snippet as this snippet shows:
This is why I keep saying that something has changed in the Beta software (most likely not the issue) OR in the version of Dx11 being used since Beta 21 (my hunch is riding here) that is causing the AMD driver's PowerTune 2.0 to heavily underpower the GPU starting w\ Beta 22 Dx11 forward. I have tested this across 3 Crimson drivers (16.4.1, 16.5.3, 16.6.1) and verified that the issue remained across all 3 drivers before I thought to disable PE in the AMD Crimson 16.6.1 drivers.
Continuing testing...................