Modern Real Time Ethernet technologies rely on well-described hardware acceleration techniques to ensure that hard real-time deadlines are met. Real Time Ethernet is also often used in situations where appreciable non real-time bandwidth is used and where throughput is important. In this paper, using the example of PROFINET, we examine the performance of various communication controller architectures in non-real-time data handling. Finding the performance not up to industry requirements we propose a multiprocessor architecture and describe an implementation and measurement results. This architecture allows better partitioning of communication and application processing whilst supporting maximum bandwidth utilization for non-real-time traffic.