Agreed it’s not a magic bullet and requires proper implementation, but DX12 does scale across multiple CPU cores more effectively with better distribution of load. It can also reduce single thread bottlenecking caused by abstraction present in higher level APIs like earlier DX versions.
That’s not to say that DX11 can’t be written in a way that also utilises lots of cores quite well and a good DX11 implementation will almost certainly beat a poor DX12 one - star citizen would be an excellent case in point where it is massively multithreaded with good distribution and currently running on DX11 as Vulcan work is ongoing. But even so you simply cannot argue that a well written DX12 implementation won’t be more performant on higher multi core CPUs than an equally well written DX11 implementation and is less likely to see a single thread bottlenecking.
Potentially for CPU utilisation this game could also probably really benefit from the reduction in drawcalls due to the sheer amount of objects on screen in higher settings... DX12 is ridiculously more performant in this regard.