This paper proposes a dynamic traffic data visualization system with OpenStreetMap (OSM). The system connects server database to client web browser by a Transmission Control Protocol/Internet Protocol (TCP/IP) to access remote traffic data. In order to reduce the computation consumption of the server, the traffic estimation and prediction from large datasets is analyzed in the client. We implement a Graphic Processing Unit (GPU) programming technology to implement the data mining process in parallel for real-time approach. To provide an intuitive interface to the users, the system renders the data mining results with the OSM mid-ware, which provides geographic data of the world.