This is not the experience I have had in Assetto. Before they fixed the AA (pre 1.7) I changed to PIXEL_PER_DISPLAY=1.5 in the occulus.ini file (located in the Asseto cfg folder). This had a noticeable difference on quality but there was still some shimmering.
Once ver 1.7 was released I set AA to 4x which has resulted in further improvements in image quality and no shimmering.
As to the comments made by ukrifter in another thread about the difference between AA and pixels-per-display setting, I have been asking myself the same question as I guess they are fundamentally doing the same job.
On a 980ti & firstname.lastname@example.orgGHz I am now running ppd @ 1.8, AA @ 4x, AF16x. I have come to these numbers after some experimenting over the weekend after noticing some stutters in game and realising my GPU was nowhere near fully loaded (using MSI afterburner). It currently seems that I am CPU bound in certain areas, which I was a little suprised about. I am not sure if this happened after updating to Win10 or if this was like it on Win7 also but that is for another thread
Performance of the 1070 compared to the 980ti seems comparable (certainly not worth upgrading from a 980ti). Benchmarks seem to show that the 1070 is faster in some games and slower in others although, from what I have seen, the 980ti seems to perform better than the 1070 in DX12 titles currently available e.g ashes & Tomb Raider. That being said you may get a better overclock out of the 1070.
If I was buying now my decision would be based on price. If the 980ti is noteably cheaper (say >£30) I would go for that option. SMP looks like it could have a significant effect on VR but how long is it going to be before we see it adopted in games.
I watched an interesting youtube video on AdoredTVs channel where he scaled the clocks on a 980ti and 1080. The conclusion he came to is there do not appear to be any significant architectural improvements in the Pascal Cuda cores over Maxwell and most of the performance gains come from the increase in clock speed afforded by the process node change.
TBH I was not very excited by the 1080 when it released. It was only once I watched a video on SMP that I changed my mind. To me, this seems a much more intelligent approach - to reduce the VR overhead rather than providing more GPU grunt. I am eagerly awaiting some real world results of VR content using this tech