A processor isn't even a single thing. There are clusters of processing cores, which may be homogeneous (all equal) or heterogeneous (some different designs from others), and how well the system firmware manages the allocation of threads to these is more important than the flat out speed in simplified tests. And then there is the GPU cluster, which is different again, and may be more important for some tasks. In short, there need not even be a single "most powerful", it will depend on the task.
And then the real question: what actually matters in real use? Because unless you are really sad you don't spend your time running synthetic benchmarks, and there are things they don't tell you. To pick one example where these things may matter, someone who likes high end games might want "the fastest" CPU and GPU. But if they plan on playing for more than a couple of minutes they should care about the thermal envelope, i.e. how long can you play before the phone gets hot and starts to throttle? And what sort of throttling is used (different strategies can result in more consistent or more jerky performance), and the same SoC (system on a chip, the correct terms for these processors) in different bodies will produce different results. And then there's the question of how the manufacturer has optimised their software.
In short, there isn't a simple, one size fits all answer to this.