C#.NET pseudorandom number generation: C# system.random() class vs. Std::rand() in C and C++

Now, I'm working on the code in both C++ and C# that implements an AI genetic algorithm. As you've probably might know, the initial population of chromosomes in this case is randomly generated by using PRNGs. Specifically, for the genetic algorithm initialization purposes it's strongly recommended to use normal random distribution created by using for example Box-Muller transformation or suppose MersenneTwister. When I normally use std::rand() function in C++ which is very simple, I obviously get the desired quality of random number generation.

Another case is libraries and packages for C#.NET such as built-in System.Random() or MathNet.Numnerics library. In spite of that those libraries for C# allow to either perform plain random numbers generation or sampling to get various distributions including normal distribution, all those libraries for C# really don't do what they're all supposed to.

To be more specific, I normally get very different results on the code that implements a genetic algorithm in the either C++ or C#, at the most part because of the random number generator used. When I run the code written in C++ based on using std::rand() function everything is going well, and not when running C# code unless otherwise.

Thanks for guidance. I really have not done any stats tests yet at now. As for more sophisticated STL C++ PRNGs I know about them, but I would have liked to stress that at least simple std::rand() normally provides better results than those similar functions in C#.NET.

Anyway, I'd kindly appreciate for any ideas how to make using PRNGs in C# better.