An interesting detail is that Nvidia has studied what it would cost to take workloads of raytracing to Pascal at 60 fps . This architecture was designed with DX11 and DX12 in mind, but not DXR, and the green believes it would need a GPU four times more powerful than the GTX 1080 Ti to perform such calculations. That is, it would have required a chip of 35 billion transistors and a size of 1324 mm ^ 2. That is not practical at present, since it should not be manufactured more than 800 mm ^ 2, and that is why Nvidia created Turing.

Apart from < em> raytracing in real time, the Turing architecture includes a redesigned pipeline that allows simultaneous FP and INT executions. Previously, many cards depended exclusively on FP calculations, but INTs are becoming more common, thanks in part for ray tracing algorithms. According to the game and the implementation with DXR , Nvidia data indicates that the Turing architecture, even without the RT cores, is more powerful than Pascal (what would be the GTX 1660 Ti and 1660, go) .

NVIDIA image