Volume data doesn't have to be limited to the maximum size of 3D texture memory. The visualization technique can be virtualized by dividing the data volume into a set of smaller ``bricks''. Each brick is loaded into texture memory, then data slices are textured and blended from the brick as usual. The processing of bricks themselves is ordered from back to front relative to the viewer. The process is repeated with each brick in the volume until the entire volume has been processed.
To avoid sampling errors at the edges, data slice texture coordinates should be adjusted so they don't use the surface texels of any brick. The bricks themselves are oriented so that they overlap by one volume texel with their immediate neighbors. This allows the results of rendering each brick to combine seamlessly. For more information on paging textures, see Section 6.8.