I think the "chunked" terrain system dosent really help in rendering the terrain, because we forgot about texture sizes that the terrain uses, or workarounds using a multitexture shader. I remember my friends telling me about the Playstation 1 that it could handle many thousands of polygons on screen at once, but this is when those polygons are untextured goraud shaded objects. Using texture maps on the models(or any 3d object) greatly reduces the on screen polygon amount. So a 6000 polygon terrain with a 256 texture will run faster than a 6000 polygon terrain with a 1024 texture. And I dont know what chunked terrain does in reguards to splitting up the texture, or even if it does anything besides hiding polygons.

About the terrains versus models, dont terrains discard the x and y value of each vertecy? Isnt that why we can only move a terrain's vertex in the z value? (up and down). I would guess terrains would run a bit faster than models if both had the same large amount of polygons, since models have much more information for each vertex (I might be wrong).

But then again with models, you can have a seamless transition into underground caves and holes in the floor, you can increase the polygon density of special places, over hanging geometry etc. Say for example you made a hill with a terrain. The hill's silhouette is only as smooth as the whole terrain's polygon density per distance, so if you made a steep high hill, you would see the facets of the polygons trying to approximate the curve. If you wanted to increase the smoothness of that hill silhouette, you would have to start over and increase the WHOLE terrain's polygon amount. With a model ground howver, you can bump up the polygon amount for your hills, and lower the polygon amount for areas with no silhouette (small dips in the floor, holes).