Intel has showcased an updated version of Texture Set Neural Compression, an AI-powered technology for texture compression. It reduces texture size by up to 18 times with minimal quality loss, saving VRAM, memory, and disk space.
The technology supports Intel GPUs with XMX and integrated Panther Lake chips with Arc B-Series, and also has a fallback mode for other GPUs and CPUs.
TSNC compresses standard BC1 textures during installation or gameplay. The neural network stores data to restore the original and works during installation, loading, texture streaming, and per-pixel sampling. Intel notes that the technology is suitable for both saving VRAM and reducing the size of games and patches.
TSNC availability as an SDK is expected this year. Two modes are available – Variant A and Variant B, the latter being more aggressive, with a loss of only 7% visual quality.
