NVIDIA RTX Video Super Resolution Now Uses 30% More Efficient AI Model

NVIDIA has just announced a new update to its RTX Video Super Resolution (VSR) video-upscaling feature, now adopting a new AI upscaling model that uses “up to 30% fewer GPU resources”. This under-the-hood improvement will require the NVIDIA App update, which should be automatically done as soon as you open the app.

NVIDIA Updates AI Model For RTX VSR

In case you missed it, RTX VSR was introduced back in early 2023 which gives all RTX GPUs the capability to upscale videos on various online platforms (YouTube, Twitch, etc) and VLC Media Player. This is done using the GPU’s onboard Tensor Cores to reduce compression artifacts and sharpen the video to the monitor’s native resolution – but this can consume significant amounts of power, making it less than ideal when the GPU is already in use, such as gaming.

According to NVIDIA, the new model will use up to 30% fewer GPU resources at the highest quality setting, which means older, less powerful GPUs should be able to use this option without causing stuttering issues. Besides that, HDR videos will now be upscaled as well, which can be coupled to the existing AI-powered HDR tone mapping from the RTX Video HDR feature.

When RTX VSR is enabled under ‘Auto’ setting, users can now choose the priority level of the upscaling workload – say for example you’re running a game whilst watching a video at the same time, this setting allows you to adjust how much resources can be allocated to upscaling or gaming. ‘Low’ gives most resources to the games currently utilizing the GPU, while ‘High’ will assign as much GPU resources as it needs to upscale the videos at best quality available.

Additionally, there’s a toggle available which shows the user if RTX VSR or RTX Video HDR are active; when enabled, both indicators will be located on the top-right of the screen to show if either of the feature are working.

Pokdepinion: Given the power draw this feature can pull, any bit of improvement is good to have.