1

Topic: ray tracing goes to masses

https://www.youtube.com/watch?v=jkhBlmKtEAk in the official press release it is told that till the end of this year developers of games can already start to implement elements of trace of rays for obtaining of photorealistic images and visual effects of cinema quality. Also be required to them for this purpose not only the videocard with architecture Volta, but also their followers https://overclockers.ru/hardnews/show/9 … schestvuet NVIDIA not only "hides" advantages of trace of rays in closed ecosystem GameWorks, but also hints that hardware resources on adequate playback on speed of appropriate scenes suffices only at Volta and more modern videocards. At least, Volta it is equipped with the certain hardware unit which is responsible for trace of rays. https://www.youtube.com/watch?v=LXo0WdlELJk Expanding DirectX 12: Microsoft Announces DirectX Raytracing https://www.anandtech.com/show/12547/ex … raytracing rushed

2

Re: ray tracing goes to masses

Hello, xma, you wrote: xma> SEED - Project PICA PICA - Real-time Raytracing Experiment using DXR (DirectX Raytracing) xma> https://www.youtube.com/watch?v=LXo0WdlELJk this demo write that on three Titan V one more worked here, https://www.youtube.com/watch?v=81E9yVU-KB8 and it PowerVR (the mobile chip), more low: https://www.youtube.com/watch?v=Xcf35d3z890

3

Re: ray tracing goes to masses

On hearings with GDC this (upper) demo on four Titan V was twisted https://www.youtube.com/watch?v=J3ue35ago3Y https://www.youtube.com/watch?v=tjf-1BxpR9c

4

Re: ray tracing goes to masses

5

Re: ray tracing goes to masses

6

Re: ray tracing goes to masses

Hello, novitk, you wrote: N> 60 - means that through AWS it well practically all. It I to volume that 10 years ago the similar demanded computer center which was presumed only by the companies of the size Pixar/Dreamworks. Well $60 it is turbid enough, since actually there 4x Tesla V100 - as a matter of fact an equivalent 4x Titan V = 4x $3k = $12k as to the further perspectives it is possible to look at this picture from here, https://www.nvidia.com/en-us/about-nvidia/ai-computing/ accordingly, it is possible to expect 150-200 TFlops in tops to 2025' to year (x1.5) ^ 7 years = x17, GTX 1080: 9 TFlops * x17 = 153 TFlops. 24 fps now is on: 4xTeslaV100 (DGX Station $60k) = 4 x 14TFlops = 56 TFlops i.e. in 7 years - one 3 nm "GTX 1060" will be faster DGX Station (4x Tesla V100)