Parallel performance is not as good as expected

Possible explanations are: