I think for LOD, it’s more about RAM/VRAM/Disk speed & size than CPU speed.
You got r^2 more data (LOD2 to 4 is like 4 to 16) to load through I/O operations and into RAM for terrain data & into VRAM for texture data from you SSD. I wonder if it’s all random read.
I feel DX12 could be a minor improvement but smart access from GPU could be a big improvement. Potentially Gen4 SSD with high random read could be another improvement
It is a fact that the coders will have much more ‘power’ and control of the whole AP interface
under DX12. This and the capability to address the GPU on more than one core will - if
well implemented - provide us all with better use of our CPU’s.