Three-dimensional (3D) raster data (also named voxel) is important sources for 3D geo-information applications, which have long been used for modelling continuous phenomena such as geological and medical objects. Our world can be represented in voxels by gridding the 3D space and specifying what each grid represents by attaching every voxel to a real-world object. Nature-triggered disasters can also be modelled in volumetric representation. Unlike point cloud, it is still a lack of wide research on how to efficiently store and manage such semantic 3D raster data. In this work, we would like to investigate four different data layouts for voxel management in open-source (spatial) DBMS - PostgreSQL/PostGIS, which is suitable for efficiently retrieving and quick querying. Besides, a benchmark has been developed to compare various voxel data management solutions concerning functionality and performance. The main test dataset is the groups of buildings of UNSW Kensington Campus, with 10cm resolution. The obtained storage and query results suggest that the presented approach can be successfully used to handle voxel management, semantic and range queries on large voxel dataset.