GeForce RTX 20-series owners can now enjoy a spot of deep learning enhanced video playback

pcgamer.com:

When Nvidia launched its Super Video Resolution technology (VSR) in February, people with GeForce RTX 20-series cards were a bit miffed to find out that their GPUs couldn't run the deep-learning powered video upscaling system. Well, it's taken eight months for the complaints to be addressed but Nvidia has finally brought them into the fold with a new driver.

Millions of us enjoy streaming short videos and movies all the time, but on PCs sporting a high resolution monitor, the visual quality isn't always great. Part of that is down to whatever compression tech the video provider has used but sometimes it's simply down to the resolution of the streamed content.

There's nothing wrong with using a browser to watch a 1080p video on a 4K monitor, but the upscaling method employed by Chrome, Edge, or whatever browser you use, is unlikely to be super advanced. Straight lines and edges in the video can be overly blurry and colour gradients can look a tad blocky.

VSR gets around these problems by scanning each frame in the original low-res video, using a neural network to predict what the frame's residual image will look like after it's been upscaled. The algorithm then uses this to tweak the upscaled frame, so that any artifacts are fixed. In theory, at least.

It does work pretty well, although depending on what quality setting you use, the GPU's power consumption during video playback can be quite high. The biggest problem with it was that there was no reason as to why RTX 20-series cards couldn't use it, as they seemed to have all of the hardware features required.

Eight months on and Turing owners can finally get it try out, as Nvidia has expanded the support to the full range of RTX GPUs in the new version 1.5 of VSR. If you