So with shader model 6.x in play, do you think that will resolve the GPU efficiency issues we've been having with DX12 ports?
If SM6.x is used, that will make the game truly native DX12.
It's important to note that we don't forget to perform optimizations at the high level such as writing your shaders geared towards the API usage or optimizing the data layout first before we talk about doing low level optimizations such as going into the specifics of how intermediate bytecode format like DXIL shaders (SM6.x) interact with the native GPU ISA in this case. These low level optimizations cannot be matched without the high level optimizations as an
example shown by Alex and Fiona for advanced Metal shader optimizations. I believe just stock DX12 can outperform DX11 when we consider newly exposed concepts at the high level such as D3D12 multi-engine and optimizing the barrier usage by removing redundant sync points in how resource transitions are handled ...
The only requirement for a game to have a "truly native" DX12 back end is to use the D3D12 runtime. The only thing unique I imagine you mean is the *implementation* behind each game which depends on the usage of the features unique to DX12 compared to DX11 ...
There does exist opportunity for faster running shader programs in using DXIL shaders (new HLSL bytecode format with the dxc compiler) since it better maps to modern GPU ISAs over DXBC shaders (old HLSL bytecode format with the fxc compiler) along with the newly exposed hardware/rendering features at the shader level such as wave ops, view instancing, access to barycentric coordinates along with the new raytracing specific HLSL built-ins ...
DX12 has a multi-pronged approach in succeeding to DX11 as a high performance gfx API as it continues to mature from the kernel (WDDM 2.x vs WDDM 1.x), API interface (ID3D11 vs ID3D12), display/adapter infrastructure (DXGI 2 vs DXGI 1.x), exclusive hardware/rendering features (Raytracing) to the new shader bytecode format (SM6.x vs SM5) and it continues to replace many more legacy components that I mentioned previously ... (it's this cohesive combination of improvements to the entire graphics environment that will lead DX12 to outpace DX11 as it's new features continue to get more adoption, DX12 maybe pretty even now with DX11 but high production value developers won't have any reasons not to use these new exclusive features in the new gfx API to advance their technology since Windows 7 won't receive anymore security updates beyond 2020 )