Abstract-The complexity of mobile communication systems is permanently increasing. This is due to the introduction and refinement of sophisticated communication mechanisms such as Multiple Input Multiple Output (MIMO), hybrid ARQ, channelaware scheduling, or node cooperation. Many of these mechanisms require detailed multi-cell simulation models for their performance evaluation. This involves computationally extensive models of the wireless multi-path fading channel, which quickly becomes a bottleneck with respect to simulation time. It is therefore of great interest to reduce the amount of computation time spent in the channel calculation. In this paper, we present an attractive approach that offloads the channel computation to a massively parallel but inexpensive NVIDIA graphics card. We discuss the parallel architecture of the resulting simulation system, and study the involved synchronization and communication aspects. We show that the developed system achieves a speed-up factor of about 30 compared to an implementation on regular PC hardware.