StingingVelvet: Skyrim is a good example of what I was saying, because the detail level and such is about as good as I'd never it to be, but the draw distance and number of characters typically on screen at once are not. In other words I'd rather have Skyrim's level of graphical fidelity with super far draw distances and a hundred NPCs on screen than have "better looking" games with the same old problems.
I generally agree that immersion goes far beyond eye-candy, but for some of that stuff the real "bottleneck" to accomplishing it has never really been GPU limited. Eg, take number of NPC's on screen at once in Skyrim. In the
city of Whiterun, you have say 80x named NPC's plus a couple of dozen guards for a total of 100 or so. Out of those, perhaps around 20 have unique voices / motion-cap body movements / character artists, and the devs will design them so the ones closest to each will be different whilst the ones that get reused will be further apart so the cut & paste "Greetings, Citizen!" is not too noticeable.
But if you squeeze in say 1,000x people in the same space, you'll end up with either:-
1. An army of clones where 10x more people will share the same voice actor / mo-cap actor / clothing and all start talking / moving / looking the same way at once like a clone army, massively breaking immersion
2. They massively scale up spending on 10x the character writers, voice & mo-cap actors and character artists making each one unique to avoid 1, but then end up cutting out something to compensate (eg, one less city or faction or a shorter story) due to development time / money constraints, or
3. The devs cheap out by adding a few extra mo-cap actors & artists but then make 90% of the 10x larger population mute & non-interactable "placeholders", which to some will also break immersion given the Elder Scrolls has been about being able to talk to anyone.
Likewise with making cities 10x physically larger, it isn't GPU draw-distance limitations that's the big issue but the fact the devs are going to have to make the equivalent work of designing 60x Skyrim sized cities for 6x ten times bigger Super-cities, making 10x the buildings (and doing so without breaking immersion through too much cut & pasting of interiors or "cheapening out" again by making 9 out of 10 buildings non-interactable art (think of some early FPS's where you see a street but only 1 out of every 10 buildings have real interiors to fake 'scale').
^ Stuff like this is why I only buy mid-range GPU's these days and have long ditched the naive teenage belief I once had of
"all we have to do is buy 2x more powerful GPU's and devs will start making games twice as better!" ;-)