It's more about estimating the edges, and that the underlying depth estimate is always going to be noisy and approximate. For example, the Pixel 3 does somethin

It's more about estimating the edges, and that the underlying depth estimate is always going to be noisy and approximate. For example, the Pixel 3 does something similar for Portrait mode [1]. Don't most modern real-time DOF techniques rely on some sort of G-buffer with primitive IDs? In the article it explicitly says they wanted to just work with RGBZ so that existing titles can just be post-process blurred. Finally, I think the time required is probably closer to 50ms per frame per GPU. If you assume that the content was running at about 60 fps (16ms/frame) on a single GPU, and linearly sped up via magic to 240 fps (4 and change ms), then the delta to 60 fps is 12 ms across four GPUs or about 48 ms on a single GPU. [1] https://ai.googleblog.com/2018/11/learning-to-predict-depth-on-pixel-3.html https://ai.googleblog.com/2018/11/learning-to-predict-depth-...