This paper presents a parallel algorithm for rendering ultrasonic volume data based on image space subdivision, along with its implementation and a performance analysis for a Parsytec CC distributed memory platform. The parallel rendering performance has been evaluated with special focus from the application's point of view which demands an "interactive" frame rate (about 2 rendered images per second). The algorithm distrubutes single rows of the image to be rendered on single computing nodes based on a farmer/worker paradigm. Once distributed, the processing nodes perform concurrent ray casting on their local copyof the replicated volume data. The generated subimage is then sent back to the farming process for compositing. There is no communication among the rendering processes necessary. The assignment of tasks to workers is done dynamically, because of the data dependent rendering time of a single task. Results are given in order to demonstrate the feasibility of the proposed algori thm in a medical application environment with its specific requirements.