Performance of matlab function build and called by C# application

Dear all, I wrote the following matlab function (R2011b):

function y = makesquare(x)
  y = magic(x);
end

Now I'm doing some tests with Builder NE to check performances using ML code in C# 2010 with .NET Framework 4.0.

in C# I have written the interface:

public interface IClasse1CS
{
    void makesquare(int x, out double[,] y);
}

and the code (snipped) to use the ML code via C# Interface (or native classes)

using MathWorks.MATLAB.NET.Arrays;
using MathWorks.MATLAB.NET.Utility;
using MyNameSpace;

IClasse1CS c = new MechanismIClasse1CS();

for (int i = 0; i < 10000; i++)
{
   c.makesquare(1, out output);
}

the same loop (10000 times) takes about 0.24 secs to execute on MatLab, while takes about 2.6 seconds to run on C#. Is this by design or should it be optimized? If these are the performance I'll can't use NE to integrate ML code in my C# code, performance must be similar between C# and ML.

thanks