TSQL versus CLR Performance: Which is faster, why and when?

I am beginning to work on a series to determine which is faster TSQL or CLR, why, and when to use which. I am going to be working in parallel with some other experts here to make sure that the TSQL and the CLR are both working optimally and doing similar operations to make sure that this testing is valid and covers all of the bases. This is being sparked by the following post on the MSDN forums, and will be posted on SQLCLR.net.