Difference between bump offset and parallax occlusion node

Found that there was a bump offset and parallax occlusion node. They both look like they give me the same results, but why are there two? Where are the differences if some are there and which should I use?

I suggest downloading the content examples map, they explained the difference there very well.

Their operation principle is roughly the same. The results are not. Bump offset simply offsets texture coordinates based on view angle to the surface to create illusion of depth.
Parallax occlusion mapping also offsets texture coordinates, but it calculates them by checking the heightmap several times to find accurate intersection of view ray and surface. In addition it allows to calculate self-shadowing within the texture. Parallax Occlusion Mapping gives incomparably better visual fidelity than simple bump offset, but also at much higher performance cost.

Ah thank you.