The standard method of ray tracing takes every object, as well as every light source and calculates reflections and shadows ...this takes processing time... films from Toy Story to Transformers and beyond need computational power to render a frame. Every frame can take hours to render to achieve a more realistic view... RTX is not taking every object in a scene and doing that in an instant... it is more geared to key components. That is to say - the complete frame is not ray traced, just partial bits of it... mixed with the standard raster method and their AI smoothing could eventually lead to more realism... Until benchmarks appear, this is seemingly a gimmick for the gaming community.
Edit: You don't go from 1 frame an hour to 60 frames per second overnight