I don't have specific benchmarks but if you were going to benchmark the CVT library (assuming you're using the default read and write polymorphic VIs) you're going to have to take into account how often you expect the tag name string input will change. If the string name is the same as the previous iteration the read/write functions will essentially just be a string compare and then an FGV access with either an index array or replace array subset operation.
If the tag name input changes you will add a couple of comparisons and variant lookup in order to get the tag's index which will be slower (by how much I don't know).
Matt J | National Instruments | CLA