Synthetic benchmarks are made for this. However, it should be noted that they often lie about what is really going on.
The best way to measure performance changes across systems is with your own workload. (Hopefully it is something that can be timed). You will, however, have to go through each component to get a good idea of how much speed varies as a result of swapping component A with component B.
As a result, review sites like Anandtech rarely check differences between multiple components. They also do very targeted tests, running applications that should stress whatever component is being switched.
Performance analysis is pretty tricky as there are literally 1,000,000's of factors that come into play when you are asking the question "Is part A better than part B?"