We have developed a visualization system, named Atomsviewer , to render a billion atoms from the results of a molecular dynamics simulation. This system uses a hierarchical view frustum culling algorithm based on the octree data structure to efficiently remove atoms that are outside of the field of view. A novel occlusion culling algorithm, using a probability function, then selects atoms with a high probability of being visible. These selected atoms are further tested with a traditional occlusion culling algorithm before being rendered as spheres at varying levels of detail. To achieve scalability, Atomsviewer is distributed over a cluster of PCs that execute a parallelized version of the hierarchical view frustum culling and the probabilistic occlusion culling, and a graphics workstation that renders the atoms. We have used Atomsviewer to render a billion-atom data set on a dual processor SGI Onyx2 with an InfiniteReality2 graphics pipeline connected to a four-node PC cluster.