It's a two way street. Having dedicated units means having idle die space for anything that does not use it. Having all in one units means the different tasks can 'fight' each other when all of them are enabled.
But there is a reason we went from having separate vertex and pixel shaders to unified shaders. In the long term, I think a unified compute unit is more versatile, although harder to optimize.
REALLY?? Why did it take you like 4 times before you answered that question? And tell me, which card was that? Or are you going to google now to find out which cards were the ones that did have DX12 while nVidia didn't?
Running? We have no ray tracing benchmarks. There is no basis to say that nVidia's RT is better. So don't pretend you're being rational. To be fair, I do expect AMD to do worse here, but by how much and whether it really matters is another story. Also... Competing =/= winning. The 5700XT was a viable competitor to the RTX 2070(S) (in spite of your dissatisfaction, I'm sure), despite it not winning. YOu don't always have to win to be competitive. Look at 1st gen and 2nd gen Ryzen. Technically they weren't really winning against Intel, and yet they competed well enough that many people bought Ryzen.
As for DLSS, I'll simply quote what I already said;
It ultimately is. It is the fault of people spending blindly, that the RTX 3090 is $1500, and that the 6900XT is $999.
If you claim to be objective and consistent, your past choices should be in line with your current claims. If they're not, you're full of
.
And just like that, we're back to dismissing AMD's technological advantages at any given point in time...