Why are multi-dimensional arrays in .NET slower than normal arrays?

Single dimensional arrays with a lower bound of 0 are a different type to either multi-dimensional or non-0 lower bound arrays within IL (vector vs array IIRC). vector is simpler to work with - to get to element x, you just do pointer + size * x. For an array, you have to do pointer + size * (x-lower bound) for a single dimensional array, and yet more arithmetic for each dimension you add.

Basically the CLR is optimised for the vastly more common case.

Array bounds checking?

The single-dimension array has a length member that you access directly - when compiled this is just a memory read.

The multidimensional array requires a GetLength(int dimension) method call that processes the argument to get the relevant length for that dimension. That doesn't compile down to a memory read, so you get a method call, etc.

In addition that GetLength(int dimension) will do a bounds check on the parameter.