Performance analysis of CLR languages

Very interesting article in CodeProject - 'A statistical analysis of the performance variations of assorted managed and unmanaged languages'. Interesting title with a good attempt. Particularly note the followups from Bill McCarthy regarding configuration settings. I took a cursory look at it - my concern is with the instrumentation itself, i.e. use of QueryPerformanceCounter for measurements. QPC makes several I/O port calls and this value then needs to be normalized. All this may mar the logging -  like Heisenberg's Uncertainty Principle. I will have to look more in depth into this experiment to feel more conclusive.