NVIDIA Promises Breakthrough: AI to Reduce VRAM Consumption in Games by Sevenfold
NVIDIA has introduced the RTX Neural Rendering system, which is set to change the approach to creating and rendering game graphics.
RTX Neural Rendering is based on AI models and aims to solve the problem of growing video memory requirements. According to NVIDIA, RTX Neural Texture Compression technology allows for a sevenfold reduction in the use of video card VRAM and system RAM.
The company also announced RTX Neural Materials — a system that uses AI to compress shader code. This provides a fivefold increase in material rendering speed and, according to NVIDIA, "allows you to get cinematic-quality objects at the frame rates used in games."
Another component is RTX Neural Radiance Cache (NRC). It uses AI to improve the performance of lighting with path tracing. NRC is already available in the RTX Global Illumination SDK and will soon appear in RTX Remix.
NVIDIA announced that developers will have access to RTX Neural Shaders and RTX Neural Compression as early as the end of January via the NVIDIA RTX Kit toolset.
For gamers, the key innovation remains texture compression, which the company has been talking about for a long time. It should significantly reduce VRAM requirements in modern games.
In the video, NVIDIA said that it is training RTX Neural Rendering on thousands of hours of video footage, including scenes with snowfall and rain, to improve the rendering of weather effects.
The company has been training new AI models DLSS M and L (DLSS 4.5) for a year to improve upscaling quality.
