Petersen clarified that the increase in performance is not only due to the increase in the number of CUDA processors and the increase in frequencies. In the architecture of Turing, there are other changes. In particular, the increase in the size and efficiency of the cache memory, the optimization of the execution pipeline, the increase in the rate of work with integers and the improvement of the original shader cores.
It is interesting that in some scenarios the GeForce RTX 2080 may even be faster than the current flagship, the GeForce GTX 1080 Ti, although the marketer did not bring examples. But he clarified that the best cards will reveal themselves in actual and new games with ray tracing and anti-aliasing with artificial intelligence DLSS.
By the way, the open Microsoft DirectX Raytracing API (DXR) will be used to use these chips, which will theoretically allow them to be implemented on older video cards. But the processing speed will, of course, be much lower. After all, GeForce RTX has hardware RT-cores and tensor processors, which is not present in older models.
As for the prices, the actual cost for the GeForce RTX 2080 Ti Founders Edition is 1199 dollars. Although the partners of the company have analogs from $ 999. In this case, Petersen argues that soon prices will fall.
Finally, he promised that all new models of video cards, including reference cards, can be overclocked. In particular, it is promised up to 2.1 GHz by the core. And NVIDIA Scanner and GPU Boost 4.0 technologies are being prepared to simplify overclocking, although there are no special details about them yet.