Two days after the announcement of GeForce RTX 2080 Ti, 2080 and 2070 with an important emphasis on real-time ray-tracing, the company provided initial performance data showing that the GeForce RTX 2080 improves the performance of the GeForce GTX 1080 on average 35% to 125% with resolution 3840x2160.
You may notice that six of the 10 games tested include results with Deep Learning Super-Sampling enabled. DLSS is a technology under the RTX umbrella that requires developer support. It supposedly improves image quality through a neural network trained by 64 samples of a very high quality image. This capability is accelerated by the core tensors of Turing architecture and is not yet available to the public.
By disabling AA and using DLSS instead, NVidia achieves a similar image quality, but benefits greatly from hardware acceleration to improve performance. Therefore, in these six games, Nvidia demonstrates a great boost on Pascal thanks to the improvements in the Turing architecture undisclosed, and a second increase in the shutdown speed of AA and DLSS. Shadow of the Tomb Raider, for example, seems to get a 35% increase from Turing's changes, plus another 50% after switching from AA to DLSS.
In the other four games, Turing's architecture improvements are entirely responsible for earnings of between 40 percent and 60 percent. Enthusiasts who had previously speculated that Turing would not be much faster than Pascal due to its relatively lower CUDA core scares were not considering the underlying architecture. There's a lot more to it than the specification sheet suggests.
Nvidia does not list the detail settings used for each game. But Nvidia's benchmarks at least suggest that Turing-based cards greatly improve their predecessors, even in existing games.