Massively multithreaded parallel architectures such as Cray's MTA and XMT are proving to be highly effective for graph analysis algorithms. By providing uniform memory access times for data with much irregularity and little locality, these machines have demonstrated excellent scalability for a wide range of graph-based algorithms. However, no apples-to-apples comparisons between these architectures and our traditional distributed memory architectures had been performed using realistic input data. In the PageRank Derby, we implemented Google's web-page ranking system using both distributed memory and massively multithreaded paradigms, and compared performance on the Cray MTA and XMT with distributed-memory Linux clusters and RedStorm. In this talk, I will describe the PageRank algorithm and its implementation on both massively multithreaded and distributed memory architectures. And I will present the results of the PageRank Derby.