Ok here we are.
![]()
Its pretty obvious which generation is the odd one out here. Remove the 1st release of Turing and you get a really consistent set of steps between generations and you get a consistent set of prices between generations.
The 2080 Ti is clearly an odd one out here, almost, I would say, to the extent that it does not belong in this comparison at all.
The super range is clearly where the 1st release should have been, and the equivalent card to the 1080 Ti is not the 2080 Ti, but the 2080 Super.
I see it differently. 780Ti -> 980Ti is around 50% increase for a lower cost. 980Ti -> 1080Ti is around 30% increase for a slightly higher cost. 1080Ti -> 2080Ti is around 25% increase for nearly double the cost.
The perf/$ of Pascal was already behind that of Maxwell and Kepler and Turing just turns that up several notches. The score should be a logarithmic scale so that even spacing shows an even % increase and on that sort of scale you will see that Kepler to Maxwell to Pascal to Turing shows a drastically shrinking performance delta from fastest card to fastest card.