Intel's Texture Set Neural Compression (TSNC) technology delivers textures up to 18 times smaller than standard BCn formats while maintaining visual quality, according to a recent TechPowerUp video demonstration. Using AI-trained neural networks, TSNC processes input textures through an encoder, compresses them into space values, and decodes them via a network decoder. This results in significantly smaller game textures that load faster, consume less VRAM, and improve performance on modern GPUs.

Intel offers two variants: Variant A achieves 9x compression with minimal visual impact (5% quality drop), while Variant B provides the 18x compression at a higher 7% quality trade-off. The technology is expected to ship later this year as an alpha version, with beta and stable releases to follow. Intel benchmarked the system using the latest "Panther Lake" setup, achieving sub-200 nanosecond texture generation latency.

This advancement represents a major leap in efficient texture handling for gaming and graphics workloads.