And the fastest implementations aren't the fastest either, I was able to knock about 10% off one of the Java examples from 2010. I suspect the C# folks were a lot more motivated to "prove" that C# was no longer slow.

As others have pointed out, if you want to test C#'s ability to call C-functions in an unsafe way, then fair enough. Otherwise, with memory safety enabled, they are roughly comparable.

That said, I did port the C# version to Java and it added about 20% extra time, even with some of the cheats (algorithm optimizations) it has implemented.