Abstract : This paper reports on the design, implementation and benchmarking of a Java version of the Nas Parallel Benchmarks. We first briefly describe the implementation and the performance pitfalls. We then compare the overall performance of the Fortran MPI (PGI) version with a Java implementation using the ProActive middleware for distribution. All Java experiments were conducted on virtual machines with different vendors and versions. We show that the performance varies with the type of computation but also with the Java Virtual Machine, no single one providing the best performance in all experiments. We also show that the performance of the Java version is close to the Fortran one on computational intensive benchmarks. However, on some communications intensive benchmarks, the Java version exhibits scalability issues, even when using a high performance socket implementation (JFS).