This looks useful:
I agree with Jim Salter that the best benchmark is your workload.
However, you might be in a situation where you're testing basic systems with different components, and getting your app running would take more time than a simple shell utility. If you can correlate your application's behavior with a "just a number" then you can more quickly predict the advantages of different configurations
.... if you know that faster CPU is desired, then having "just a number" to work with can be a lot faster for experimentation than a full workload test.