My personal belief is that while there's a lot of misinformation floating around about why Kepler and AMD cards seem to perform worse in some games, in the end Nvidia's tactics aren't consumer friendly by any means, and are very much intended to obsolete their old products and push new ones. And that's a fair strategy as a company that's out to make profit, but it isn't beneficial for us as consumers. Making things like HairWorks that are just very inefficient effects meant to hobble their old cards and AMD cards more than their new cards isn't something I support. They could support things like Purehair that does it far more efficiently. Guess why we aren't seeing a HairWorks option in ROTTR? It would probably look pretty bad for Nvidia. The reason GameWorks exists is not to provide devs and gamers the best option for said effect, but to push more Nvidia sales. And most devs don't care, they simply see Nvidia doing their job for free.
One new thing to come to my attention is the lack of DX12 titles so far. What I've heard so far about DX12 patches is, "The performance is not there yet, DX11 still performs better". This is completely against what was touted when DX12 was released. We've seen AMD has a ton to gain using DX12, but what we've also seen is one other company struggling to make DX12 work right. So as a dev making a DX12 patch for your game, you see that it's making AMD cards xx% faster than in DX11, but also y% slower when using Nvidia cards, you probably don't feel confident in releasing said patch, because most of your users would not like that very much. I'm not saying this is the reality of it since we have no real data on the matter, but it's food for thought. It might very well take at least Pascal or even Volta before Nvidia's architecture is truly fitting for DX12 development. And this is not about feature level support, but how the architecture actually performs doing said stuff.