Nvidia's neural texture compression saves vram and boosts performance

Nvidia has presented a new technique for graphical improvement that harnesses the power of AI, but without the generative aspects that sparked controversy recently.

Introducing neural texture compression (ntc)

The tech giant's latest innovation, NTC, aims to dynamically reconstruct textures in real-time using neural models based on what's displayed on the screen. This includes factors like angle, lighting, and shadows. By learning from the original textures, NTC can significantly reduce the channels needed to load and render them, resulting in substantial VRAM savings.

In a demo showcasing the tech's capabilities, Nvidia demonstrated how a GPU with 8GB of VRAM could drop from a whopping 6.5GB usage with traditional BCn (Block Compression) to just 970MB with NTC, all without compromising visual texture quality. This is achieved through