NVIDIA’s Deep Learning Supersampling (DLSS) uses artificial intelligence to increase the resolution of the rendered frames, allocating a smaller frame and prudently increasing it. This method has been praised for the ability to render games in high resolution with much greater efficiency than rendering to the same resolution initially. Now the latest version of DLSS 2.1 includes support for VR, and can significantly improve the accuracy of VR content.
DLSS is a technology that allows you to “squeeze” more details in every frame by increasing the efficiency of rendering. It uses hardware-accelerated AI operations in GPUs Nvidia RTX to increase the resolution of each frame by analyzing and recovery of new personnel with higher resolution.
The goal is to achieve the same resolution and level of detail that rendering of a frame the same resolution and do it all more efficiently. This means that more of the capabilities of the graphics processing available for other things such as better lighting, textures, or just an increase in frame rate in General.
For example, the game with the support of DLSS can render their own frame with a resolution of 1920 × 1080 and then use the DLSS to improve picture resolution up to 3840 × 2160. In many cases, this is faster and retains almost identical level of detail compared to the original rendering of the frame with a resolution of 3840 × 2160.
Because of the need to reduce delay to a minimum, the VR already has the bar high frame rate. PC VR headset I expect that games will be consistently displayed at a speed of 80 or 90 fps. Maintaining framerate requires a compromise between visual accuracy and efficiency of the rendering because the rendering of the game at 90 frames per second has only half the time to complete each frame in comparison with a game that shows 45 fps.
NVIDIA has announced that the latest version of DLSS 2.1 now supports VR. This may mean, mainly, what VR games will look better, but with the required frame rate.
For example, the game VR headset Valve is running on the Index may initially display each frame with a resolution of 1440 × 800, and then use the DLSS to increase the picture resolution to match the resolution of the headset display at 2880 × 1600. The initial rendering of each frame with reduced resolution most of the compute power of the GPU can be spent in the other direction, for example, advanced lighting, particle effects or range rendering. In the case of Index which supports frame rates up to 144 Hz, additional efficiency can be used to improve the overall frame rate, thus further reducing the delay.
NVIDIA have not talked a lot about how DLSS 2.1 will work with VR content, and are there any reservations about technologies that are unique to VR. And since DLSS need to add for each game, yet functional examples of its use with VR.
However, DLSS received a lot of praise in the games, not related to virtual reality. In some cases, DLSS may even look better than the frame of the originally rendered with the same resolution. Employees Digital Foundry have an overview of what it means DLSS for content not related to VR.
It is not clear when will be launched DLSS 2.1, but it is likely that it will be released in tandem with the latest NVIDIA cards RTX 30 series, which will be supplied with RTX 3080 on September 17. DLSS 2.1 will also be available on maps series 20 RTX of the previous generation.