Combining Displacement Map and Normal Maps and Why
...using the normal map and displacement map combined, the displacement map does the heavy lifting of adding in mid level details and the normal map does add in the real fine details, that otherwise would add retarded amounts of render time with the displacement map alone?
What I mean to say is that anything that doesn't effect the overall silhouette of the model should be stuffed onto the normal map (as its great for this), and any major shape changing of the model should be left to the displacement map. This way you have the ideal situation.
Displacement for the "big" details and Normals for the "small" Detai
Normal maps fake actual geometry and appear to be affected by light
Displacement maps influence the position of existing vertices to cause physical change
I often don't even bother with displacement, but rather ensure the topology reflects the physical changes, but for very complex models, displacement can take far less time and give better results. You can bake normal and displacement maps from the Texture Baking Tool in the Paint room under the Textures menu. As you mentioned, you need to ensure enough geometry to support your displacements via the Retopo room or by subdividing your mesh. The latter method is likely going to cost you less time, but more polys.
One benefit of displacement maps over bump maps is the way your normals face can really affect how well your normal maps work. Displacement, is simply...well, displacement. You either have enough polys in the right areas or you don't.
Edited June 26, 2014 by alvordr
...for landscapes, I would definitely use displacement, especially if you're going to do any kind of close-up work or walking around on that terrain, etc. As to what the difference between bump and normal maps are, think of it this way:
Bump maps are usually used in a similar way to displacement, and concentrates on large details. They are gray-scale images, where the lighter areas are raised and the darker areas are recessed.
Normal maps are used to fake details in a way that light actually affects the way they appear, but no actual geometry is deformed. They are RGB images, which is why you sometimes have to flip a the green or red channel, etc., depending on what you're bringing your normal maps into.
Sometimes, 3D software calls bump maps, normal maps, or vice-versa. It's not technically interchangeable, but there you go.
So, in the example of a landscape, you could use a displacement or bump map to dictate how major deformities in the land actually occur, where you might then use a normal map to indicate rocky terrain vs smooth terrain.
The workflow in 3D Coat is often the tick of a simple check box or two to decide whether or not you want to generate a displacement map or not. When using displacement, you'll be asked to smooth the mesh or subdivide it to support the deformations that are to occur. This means your poly count goes up for displacement.
