A comparison between the most basic boxy voxel rendering and the more advanced smooth voxel rendering technique can be seen on the following screenshots.
The smooth voxel rendering can be parametrized to produce a better picture while maintaining a relatively little video-memory usage.
There is almost no performance penalty for using smooth voxel rendering.
The generators are capable of producing volumetric data with a high level of supersampling.
This leads to a temporary higher demand on (main) memory and it usually takes much longer to create the near the same content.
Supersampling might be useful for the production of milestone and release quality data.
Click on the image to view its full-size version.
(opens in a new window or tab)
The scene on the left was generated with no supersampling and drawn by the boxy voxel renderer.
The scene on the right was generated with 2^3 supersampling and drawn by our smooth voxel renderer.
The difference in picture quality becomes significant during camera motion as the usually very noticeable aliasing of boxy voxels becomes negligible.
The improvement is also well visible on closeups of objects in static images (see the zoomed-in flowerpot).
Some artifacts are still visible on the right images due to a low level of supersampling and overdone optimization of some rendering algorithms.
The scene was rendered with 2xFSAA.