There's a few things I wonder about with this. They tested an AMD GPU, which is great as it shows off how awesome ACO and the RADV projects are. The Mesa devs alongside the AMD folks contributing to the RADV driver, along with all of Valve's work, showcases how awesome FOSS can be at optimizing. I watched RADV go from underdog to top dog in performance, and it's no longer arguable that AMD hardware works better under Linux than Windows thanks to the stellar work done on those projects.
Meanwhile we have nVidia who clings selfishly to their proprietary blobs, and I can't help but wonder how great it could be if they opened that up and let the community in. Is it already hyper optimized to the point that the community wouldn't be able to improve on it? Idk, but from an infrastructure standpoint, nVidia users would benefit from it immensely, if not from performance.
I recall when vkd3d-proton performance was severely impacted under nVidia, to the tune of a 40% delta, and that's improved significantly, but I still wonder how this would look on an nVidia GPU to compare Windows vs Linux performance.