Real-time ray tracing is now possible. I kid you not. But you need a seriously powerful graphics card to run it. The first Volta GPU, the Titan V, has special tensor cores built into it which means that it can efficiently run deep learning applications.

Without a denoiser, which is used in the new technique, ray tracing with only 1 sample (light ray per pixel) looks terrible; the video output looks very grainy. Previously, in order to get high quality footage, we had to use many samples (light rays per pixel); we'd use hundreds of samples for cinematic quality. With the new method, we use only one sample (one light ray per pixel) but we then put the video output through a neural network to denoise it (take away the graininess). The neural network (which is a deep learning application) is input a grainy animation and outputs an animation that is cinema quality. The significant reduction in processing time means that cinematic quality ray traced video game graphics will be able to run in real time on a Volta graphics card like the Titan V.