Can anyone explain why there is such a huge difference in some of the benchmarks: Poll, Forking, CPU Cache, Semaphores, Socket Activity, Context Switching (all Stress-NG). Can we really trust these tests?