[jacobr1]

The thing about benchmarking and profiling is that you need a target scenario. Even if you can simulate a generic representation of a given workload, you might have different workload, or your specific environment might differ from the norm. One could imagine an armed force that actually was well attuned to fight a near-peer power in a border conflict, but wasn't well setup to fight battle a continent away against guerrilla forces. This is a peril all benchmarking has, even when correct, it is still limited to its own assumptions and context.