Intel has put out a demo video to showcase its newest Texture Set Neural Compression (TSNC) technology. Leveraging neural networks, this technology enables efficient texture compression. It boasts the ability to condense game texture data to a mere 1/18th of the volume of conventional solutions, all while maintaining nearly uncompromised image quality. The working principle of TSNC technology bears a resemblance to NVIDIA's NTC. It transforms textures into compact features via machine learning. Subsequently, during runtime, it reconstructs textures in real-time using a neural network on the GPU, ultimately cutting down on video memory usage. This technology is anticipated to notably shrink the size of game installation packages, enhance loading efficiency, and could potentially be implemented across a broader spectrum of graphics cards.
