What's the actual impact on latency and 0.1%/1% lows in games?

[matru]

I was looking at this chart and it seem that we are playing with such a narrow window for optimization (mostly driver level) with all of these tweaks (even the ones who are actually not bullshit):

Have you done tests, both on latency and performance? I have went through tons of resources on this, including valleyofdoom's guides, just out of curiosity, but there are 0 benchmarks that I could've found covering the non-bullshit part, at least in a well controlled manner.

Overall I am confused, seems like we are majoring on the minor and there is no one benchmarking the results of it. This is not a critic, but I am genuinely looking for opinions/proofs and since you and valleyofdoom are probably one of the best resources on the topic on the internet, which is not bullshit (following MS recommendations and beyond), I figured out that it's great place to open this discussion.