In pretty much all current games, the 3D scene information is rendered as triangles with textures applied to them. Rendering for this is, of-course, super fast if you have a nice NVIDIA graphics card. However, this scene reprentation is limited in the amount of small details that can be represented. Furthermore, ray-casting is very slow in this representation, which makes it hard to do some advanced lighting in real-time.
A cool alternative for triangles are little cubes, basically pixels in 3D. These can very well represent a surface that isn't smooth, and are also very fast to ray-cast against. The disadvantage is that they take terabytes of space when not compressed to get sufficient detail. In this project, I compressed the memory so much that a whole navigable scene can fit into GPU memory.