This demo drew attention a lot in the game industry. Also, there is no mention to it in the nvidia paper but another noteworthy voxel terrain demo was the cave demo (voxlap engine) from Ken Silverman (3D realms / Duke Nukem 3D 'Build' editor) [1].<p>[1] <a href="http://advsys.net/ken/voxlap/voxlap03.htm" rel="nofollow">http://advsys.net/ken/voxlap/voxlap03.htm</a>
its also possible to achieve similar effects using whittaker iteration as a 'sloppy but fast' alternative to the sphere tracing/distance field approach<p>there is not much good reference on it though (which is why i am compelled to self link)<p><a href="http://software.intel.com/sites/billboard/article/star-chart-developer-diary-part-2" rel="nofollow">http://software.intel.com/sites/billboard/article/star-chart...</a>
<a href="http://jheriko-rtw.blogspot.co.uk/2009/05/whittakers-method-and-brief-history-of.html" rel="nofollow">http://jheriko-rtw.blogspot.co.uk/2009/05/whittakers-method-...</a><p>its less well known by far... Steven Wittens came across it, and I shamelessly nicked it, back when we were doing AVS presets for Winamp. Speaking of Winamp, back then Geiss (the author of this article) created Monkey - it used D3D and hardware acceleration iirc, but also rendered an isosurface similar to the method described in this article.<p>It was an interesting period of actual innovation in those days...
Another take on the marching cubes algorithm is here:<p><a href="http://www.cs.unc.edu/~marc/tutorial/node130.html" rel="nofollow">http://www.cs.unc.edu/~marc/tutorial/node130.html</a><p>It more clearly states that marching cubes is a method of converting volumetric data into a bounding surface.
See also Eric Lengyel's take on this which he calls the Transvoxel Algorithm:<p><a href="http://www.terathon.com/voxels/" rel="nofollow">http://www.terathon.com/voxels/</a>
Very nice, but is this practical for games? What about collision detection, AI players etc? Or would you use it mainly for backgrounds, skies and such?