Why is there a difference in the speed of calculating the Fibonacci number?

Question

Mono

Straight to the problem.
I'm trying to understand the reasons for the different execution times of the calculation of the Fibonacci number, which was chosen just for the test.
The test was carried out on 2 identical servers with a xeon E5-2650v2 processor and windows 7 and windows 2012 r2 systems, a workstation with an i7 3930K with win7 and a computer with an AMD phenom ii processor (there is no information on it) with linux.
The program was also compiled in .net and in MONO.
Results:
xeon win7:: .net=0.043, mono=0.020
xeon sin server:: .net=0.069, mono=0.046
i7 win7:: .net=0.035, mono=0.021
amd linux:: .net=***, mono=0.023
Questions:
1. Why is an application compiled in mono faster than in .net.
2. Why on 2 identical servers in xeon, windows7 works almost 2 times faster than windows server 2012 r2.
3. Why i7 and amd are faster than xeon.
On the 3rd question, I was told that the speed depends on the frequency of each core, so i7 won here.
I hope I have explained the problem clearly.

Answer 1

For performance testing, there is a standard LINPACK test.
What exactly needs to be tested? Platform performance, or C# performance on that platform?