What is NVIDIA's Neural Texture Compression (NTC)?

NTC is a machine learning approach that compresses textures by storing them as compact latent features, which are then reconstructed by a neural network on the GPU at runtime.

How much VRAM can NVIDIA's NTC save?

In a recent demo, NTC achieved an 85% reduction in VRAM usage, lowering consumption from 6.5GB to 970MB.

Does Neural Texture Compression cause visual artifacts?

No, unlike traditional compression methods like BCN, NTC can increase texture quality while avoiding typical compression artifacts.

Home / Technology / NVIDIA's AI Texture Tech Slashes VRAM by 85%

NVIDIA's AI Texture Tech Slashes VRAM by 85%

5 Apr

•

Summary

AI compresses textures, reducing VRAM usage significantly.
NTC tech demonstrated an 85% VRAM reduction in a demo.
It avoids compression artifacts, potentially improving visual quality.

NVIDIA's AI Texture Tech Slashes VRAM by 85%

NVIDIA is reiterating the benefits of its Neural Texture Compression (NTC) technology. Introduced nearly three years ago and available via an SDK since early 2026, NTC employs machine learning to store textures more efficiently. Instead of saving every texel, it compresses textures into learned latent features that capture essential visual information. A small neural network on the GPU reconstructs these texels at runtime.

The system offers significant advantages. In a Tuscan Villa demo, NTC achieved an 85% reduction in VRAM usage, consuming 970MB compared to 6.5GB for traditional BCN compression. This is crucial for VRAM-heavy games. Furthermore, NTC can maintain texture quality without the artifacts often seen with BCN.

Disclaimer: This story has been auto-aggregated and auto-summarised by a computer program. This story has not been edited or created by the Feedzop team.