This article will show how to increase the performance of a rather simple algorithm up to 80%, by showing how we can accept a loss in absolutely accuracy, and also taking advantage of how processors work.
The only things I can think of with the CLR would also be one-time costs. Some APIs cache things underneath, but I don't think that applies to any of this simple code. Not sure about comparing to C and/or unmanaged/unsafe code--if you need unmanaged code, you need it, so let's make it the best it can be! :)
Anecdotally, I've found that .Net can be just as performant (at the levels I care about) as unmanaged--but you do have to care about different things.