Do the Java benchmarks include the start-up time for the JVM? If so, I wouldn't say it's a completely fair comparison. (At least for the times that are small).