With an update of the GeForce Now app to version 2.0.37, Nvidia provides users of the game streaming service with expanded options in the event that the network connection presents a bottleneck. If you have a fast GPU, you can even upscale the stream to the display resolution with AI.

New video stream upscaling options

GeForce Now (test) users usually receive the image rendered in the game resolution set on the cloud system as a video stream in the resolution of the output display. If the Internet connection does not allow this, the streaming resolution is reduced and the image then has to be scaled to display it in the native resolution.

To date, this has been done via the proportional multiplication of the pixels with subsequent bilinear filtering, which requires hardly any computing power, according to Nvidia. That makes sense, since GeForce Now is also explicitly aimed at users who want to play current games without having the necessary hardware themselves.


With the update, this approach remains set as the standard, but alternatively the options "Improved" and "AI-Improved" are now also available via the setting in the app. "Improved" uses a Lanczos process and is intended to achieve better upscaling quality in this way, but it costs more system resources and can increase latency. There are no special system requirements.

AI upscaling for Tegra X1+ and fast desktop GPUs

"AI-Improved" uses a neural network for upscaling, so like DLSS it does not scale without reference to the content, but rather taking into account the material to be scaled - the knowledge was previously taught to the network. Roughly speaking, the AI ​​tries to deduce what the uncompressed original image must have looked like in order to result in the lower-resolution image, taking into account the compression algorithm. Nvidia calls the process "AI upscaling". In contrast to DLSS 2.0, the upscaling is done frame by frame and does not take into account a time component.

However, users who use GeForce Now on PC and notebook because their computer is older or only very weakly equipped are excluded from "AI-Improved". Somewhat surprisingly, Nvidia mentions very strong GPUs as a requirement, specifically all GeForce RTX 3000, GeForce RTX 2000, GeForce GTX 16xx, GTX 1080 and GTX 1080 Ti as well as the Titan series are mentioned. Nvidia Shield TV and Shield TV Pro with SoC Tegra X1+, which were introduced in 2019, are also supported. The two streaming consoles have been able to use AI upscaling since the beginning of 2020 in order to be able to play back lower-resolution source material in native TV resolution in the best possible way.

AI-enhanced mode requires an NVIDIA GPU with enough performance to run game content through the trained neural network in real-time. Refer to the list of system requirements above.

"AI Improved" is not offered on unsupported end devices. In the editorial department, this was also the case in a first test on a notebook with a GeForce RTX 3080 laptop GPU. Apparently there is still something wrong with the introduction of the new setting.

AI-enhanced is not offered on the notebook with RTX 3080

If you have any queries or concern regarding this article please mail me at - ask@thedroidverse.com / amru.meajy@thedroidverse.com

Post a Comment

Comment Your Thoughts About This -

Previous Post Next Post