You all miss somethibg very important regarding the tressfx vs hairworks issue.
Lets first assume cdpr ussed tressfx instead of hairworks. All good, use open source code to apply a particular effect of hair and fur and grass on your game, who wouldnt want that?
Well, after applying it, they have to check the performance of the game with that code being used on the biggest range of hardware withib reason. For AMD, they just have to go to their tech papers regarding the uarch, its functions, and use the debuggers, sdk or other complementary troubleshooting software. This is no problem, since AMD, altough sometimes late, its totally open of exposing their uarch to ghe devs and have all those things I named available for using. Ok, now Intel and Nvidia. Well, here is where they are boned, as neither will supply neither tech papers, debuggers or sdk to the devs regarding their uarchs. If they leave the code untouched, it is highly propable their game will run like crap on Intel and Nvidia hardware when tressfx is activated. And here is when it comes the interesting part: those 2 vendors amass the gross share of igps ans dgpus, respeclty. So the dev knows here that if they plan on using tressfx, because of the closeness of those 2 aibs regarding their uarchs, the performance will bomb as the most used pc gaming hardware will probaly tank when enabling this particular effect. There is no conspiracy from AMD in this (no convenient abuse of compute in tressfx to tank kepler as some uninformed users try to imply).
So what is left for the dev? Risk the game success because of a mayor performance issue regarding the 2 most prominent hardware makers in the pc gaming space, or try another middleware with higher optimizations to the higher parties I named? This is where gameworks comes into play. Gameworks code is as closed as a grocery store on 3AM of a sunday morning, but it is designed to work well with the latest uarch of the most prominent dgpu vendor, this is maxwellv2 from Nvidia. We cant know if wheter there are explicit crippling functions ala ICC or they conveniently abuse certain effects such as tessellation(well, after the tweak it was posted in popular forums to limit tess on AMD card and regain almost all the perf lost from hairworks, this is fairly a strong assumption). Either way, performance sucks balls on kepler, GCN <1.2 and probably all your intel igps (but the latter mostly because of the inherent lack of power as they are all igps), but at least more people are allowed to run the game better as nv is the aib with the higher marketshare. Also this is good for cdpr as they get compensations fornusing gameworks (either gpus or manpower from nv) and can shift their lazy asses to move on solving the rest of the game code till retail.
I think the above is what happened here and what usually happens in the pc gaming space since NV, most notably, decided to strat being secretive about their uarchs. As someone already said in another post, it wont be surprising if AMD is more helping than NV when reaching the devs go optimize for Kepler. Which sounds absurd, as absurd as devs settling for using closed middlewares of companies that are not open about their hardware optimizations. Sad times indeed for pc gaming ahead.