Volume Visualisation on a Hybrid Parallel Architecture

In this paper we present the design of a new hybrid system based on commodity components to gain supercomputer power at low cost. The architecture is built around a MIMD PC-cluster linked by a high-speed network and massively parallel SIMD processor boards connected to each node. We present the design of a volume rendering algorithm in order to derive an efficient mapping onto the new architecture. This results in a volume visualisation implementation, which is faster than all previously published implementations on massively parallel architectures