Does DXR take full advantage of the Turing architecture?

According to this topic, OptiX 5.1 doesn’t utilize the RT cores, which got me wondering if other APIs - primarily DirectX 12 ray tracing - currently do. If so, what determines what code runs on them? Is it driver-dependent or can certain “shader programs” (or perhaps Cuda programs?) be targeted?

For some additional context, I’m looking to implement octree ray tracing using DXR (like GigaVoxels or ESVO) and want to have full control of the acceleration structure while still fully utilizing the hardware available. Unfortunately there seems to be little information out there on the subject.

Any info is appreciated!