The resolution is increased by 4 times and the video memory is reduced by 30%. Nvidia launched the neural texture compression algorithm NTC.

Thank you, CTOnews.com netizens. Hello, rabbit clue delivery! CTOnews.com May 6 news, Nvidia recently issued a press release showing a new texture compression algorithm Neural Texture Compression (NTC). Compared with the block compression algorithm, the video resolution can be increased by 4 times and the video memory consumption can be reduced by 30%.

CTOnews.com Note: block compression (BC) is a technique used to reduce the amount of video memory required to store color data. By storing some colors in their original size, while others use encoding mode, you can significantly reduce the amount of video memory needed to store the image. Because the hardware automatically decodes compressed data, using compressed textures does not cause performance degradation.

NTC uses tensors (3D) to represent textures, and the new method aims to compress multiple channels and mipmap (textures of different sizes) together to achieve better quality and bit rate than JPEG XL or AVIF. Unlike traditional block compression algorithms that require special blocks, the NTC algorithm uses matrix multiplication techniques performed by modern GPU itself.

Neural textures can be rendered in real time, texels is 16 times more than the BC method, and 4K rendering speed is 1.15ms (measured on RTX 4090). More information about the study will be published in the August 6 issue of SIGGRAPH 2023.

The original text of Nvidia is attached to CTOnews.com, which can be read by interested users.