For years we saw posts saying everything above 1600 was a waste because 3 benchmarks said so ... but as reviewers went beyond the surface and widened the testing, we saw that most games were GPU, limited, others were CPU limited and some (i.e. F1) were bandwidth limited. THG showed that 2400 was in fact 11% faster with 2400 in F1 as compared to 1600. And as it cost just $20 more for 2400 at the time, it was certainly an option worth considering.
So while I appreciate the effort here, as it does give a general idea of what might be expected, until I see a bit wider range of testing, including 2x/3x/4x SLI and minimum frame rates ... and including games known to to be impacted by RAM like F1... , as before, I don't think I'm ready to write off faster RAM as being totally useless.