3090 also costs 50% more at MSRP (not going to pretend to know the current state of post-ethereum GPU prices).
But ultimately, people who buy halo tier cards generally don't have a budget set, which means PSU, cooling options, and electricity bill aren't an issue. If you buy either a 6900xt or 3090 and can't afford an adequate PSU and electricity bill, you're probably better off with a completely different tier of GPU, so I get why "raw performance" is the crown that gets chased. MSRP of 3090 is roughly double that of 3080, but obviously it's not double the speeds. There's too many things to consider when you consider "best' GPU so "highest performing" is easiest to refine down to.
I don't fully agree, but I get it. And I opted for the 6900xt so you don't need to convince me of the benefits (for me, Linux open source drivers were on the list as well). But I'm not going to pretend like there is a super readily approachable alternative.
userbenchmark.com notoriously changes their benchmark ratings every time AMD releases a CPU because "AMD bad, Intel good". The core i3 9100f is rated higher than the Threadripper 2990wx [0]. And I don't just mean "on the subjective aspects". They put the i3 as rank #167 in all CPUs and the threadripper at #231. Yes, they claim the i3 is faster than the threadripper, because they don't count ANY scores over "8 cores", and Im pretty sure that it means "8 threads" and not 8 cores. Because "no workload needs more than 8 threads, so we're just going to ignore that." They also release a "review" of new AMD products every time they come out claiming that AMD is garbage for made up reasons.
All this just to say that the second any subjectivity is introduced into the mix, you'll wind up with people skewing tables to show the "real fair metrics" that favor their team. I'd rather just say "yeah, nvidia holds the crown and is technically 3% faster, but 3% isn't enough for me to care". I'm not going to start a debate over "Well actually, Nvidia is better because FPS is the only metric that matters!"
[0] https://www.tomshardware.com/features/geforce-rtx-3090-vs-ra...