After Witcher 2 came out and showed people what could be done with DX9...
Hmm... Here I disagree. Many games already demonstrated what can be done with DX9. What Witcher 2 benefited was from a godly art team. Technically however the game introduces nothing substantially new. But if you marry great art with DX9 you are bound to have a beautiful game, like so many others before it.
There's however Uber Sampling. Here I concede some. But I'm actually a critic of this technique. At least until we have mainstream cards that can make any use of it. Witcher 2 with Uber Sampling is no different than Crysis when it shipped out -- A performance hog that is in no way justifiable due to the hardware requirements it imposes in comparison to what little benefit it brings. And here, Witcher 2 gets the same amount of criticism I gave Crysis. What's worse: Uber Sampling can be easily replaced with good AA and AF for a fraction of the performance hit. I say, scratch the Uber Sampling technique entirely. We don't need that nonsense.
Also the game -- although I'm unsure if there's a correlation -- payed the price of being too heavily centered in art direction and too little in design. It has one of the worst character interfaces of a modern triple-A game I can think of. The main character animation is nothing short of horrendous. You cannot, for instance, start an action without ending the previous one and your character doesn't seem to use (or make very bad use) of angular keyframe blending. It's for all purposes linear blending only.
these measly attempts with DX11 are not exactly inspiring.
I agree fully. But personally I'm more interested at the numerous tessellation demos, than any game making use of it... for the next couple of years, I would say. Hair demos you can see on youtube, skin, cloth and similar fabrics, etc, all reveal in a more direct manner what exactly tessellation is about. I have very little doubts tessellation is the next big achievement in 3D visuals.
Like you, I don't believe that DX11 is the one bringing it in, indeed. The technology isn't even matured for an API to be realistically considered on production code, in my humble opinion. I think it's important DX11 introduced it. It formalizes the technology. But it's no wonder nothing of any relevance to gamers has been actually done with it so far.
Making use of higher resolution textures with good light and bump maps will do a lot more for a game than tessellation... at this point anyway.
Absolutely. Reminds me of Prey. A game that I remember bringing bump mapping to a level never seen before (or since, I believe). It proved -- although I think most praise goes to the Doom 3 engine -- that good lighting and bump mapping can radically alter a texture and the whole look and feel of a game without increasing poly count. There's necessarily a performance it. There's always a price. But this is the case of a technology that in fact compensates with vast visual gains.
Tessellation is such a technology too. More in fact. Just not quite ready yet. I mean, I can hardly wait for it to mature. The level of detail that can be achieved is immense. It will have a large impact also on art assets (texture for instance will have to be a lot more detailed). But boy...
I think we're just going to see several generations of gimmicks first before DX11 becomes a viable option, both from a developer and a GPU architecture perspective. It's all about maturity of the platform and the hardware available of the targeted demographic, neither of which are available at the moment.
That sums up my thoughts too.