You didn't understand. For benchmarks at 1080p, Radeon 6900XT is much faster than 3090. The benchmarker wants to remove the possibility of GPU bottleneck at that resolution. Why at 1080p? Because it is the best resolution to compare CPU's. This article is not about GPU's. Hopefully you understood now.
Thanks for the pontificating, but it is still besides the point.
Pairing an amd cpu and gpu together may (or may not) have certain advantages that are not realized if a Nvidia gpu or Intel cpu is used instead. Testing with Nvidia hardware will confirm or deny that possibility.
Second, the conclusion that the 5800x is better by x% on the 6900xt is a best case scenario, but most people don't have a 6900xt, and in fact, Nvidia is the more popular brand. So what is the performance differential between the two cpus on Nvidia hardware?
There are more layers to this than "what's the fastest 1080 gpu." Especially given the exorbitant prices of the 6900xt. Microcenter just quoted me $1800 for a red devil 6900xt the other day, and they're even more expensive on Amazon, eBay, stockx, etc. I'd argue that most people with the deep pockets to purchase a 6900xt aren't gaming at 1080p anyway. So if you want to have a complete discussion on 1080p, why not measure the performance difference on hardware that most folks who game at 1080p will actually be interested in purchasing?
Just because there's a x% difference in framerate between the two cpus on the 6900xt doesn't mean the same is true on a 3070 or even 6700xt. If there isn't any appreciable difference at all on those other GPUs then that should be pointed out. Rather than the unsupported conclusion that "because the 5800x is faster at 1080p on a 6900xt, the same is necessarily true on a 6700xt or 3060 or 3070."
The only way to know for sure is to test the other hardware. And we know these guys have the other hardware. That's why I left the comment in the first place.