Just browsing and came across this:
Could the 3D features in TF be speeded up with this method?
http://www.okino.com/conv/imp_dem.htm"Features of the DEM Converter:
DEM datasets typically contain 60000 or more quadrilateral polygons, or 1200000 triangles (for a 258x258 resolution sample; the maximum DEM dataset size if 2050x2050 which would result in 4.2 million quadrilaterals or 8.4 million triangles). This is an enormous number of polygons for most 3d rendering programs so this DEM converter incorporates two unique options to overcome this problem:
The converter can skip over samples in the dataset so that only every n-th sample is used. Rather than importing 258x258 samples, the converter imports 51x51 samples (for a skip factor of 5) which results in only 2601 quadrilateral polygons. [don't think this one would be that useful. LR]
Rather than store the entire DEM dataset in single object, the DEM converter breaks up the data into multiple smaller objects with a common parent. This has shown to be an effective method to speeding up the wireframe redraws of the DEM data (by a factor of 2 or 3), and makes interactive user movement of a 3d camera much faster since each sub-object is only a few hundred polygons. In addition, certain rendering programs (such as Okino's NuGraf renderer) use much less memory when many smaller objects are used rather than one large object with many polygons. By default each sub-object stores a maximum of 900 polygons; contrast this with other converters which lump all 120,000 polygons into a single object - few renderers will be able to render such a large object.
A default 3d camera is added to the scene which views the DEM data from a pleasing angle.
u/v texture coordinates are added to the imported data so that a bitmap image can be easily draped over the DEM data.
The converter creates smoothed vertex normals for the DEM data so that it will appear to be smooth when rendered. "
