This paper overviews some aspects of using different levels of accuracy and complexity for the visualization of large data sets. Current status of the volume of data sets that can be generated is presented, together with some of the inherent problems due to such large data volumes, visualization requirements, and display limitations of existing hardware graphics. Some methods for selecting, generating and implementing different levels of detail are presented.