Nvidia has unveiled impressive results from beta testing of its new RTX Neural Texture Compression (NTC) technology, demonstrating up to 96% reduction in video memory usage during 3D workloads.
Recent benchmarks conducted on an RTX 4090 graphics card revealed dramatic improvements in texture memory efficiency. The technology was tested in two modes: "NTC transcoded to BCn" and "Inference on Sample," with the latter showing particularly striking results.
At 1440p resolution with DLSS enabled, the "Inference on Sample" mode reduced texture memory usage from 272MB to just 11.37MB - a 96% decrease compared to traditional compression methods. The "NTC transcoded to BCn" mode achieved a 64% reduction.
While the memory savings are substantial, early testing shows some performance impact. The "Inference on Sample" mode saw frame rates drop from mid-1,600 FPS to mid-1,500 FPS range on the RTX 4090. However, this being beta software on previous generation hardware, performance optimization may improve with newer GPUs.
The technology represents the first major advancement in texture compression since the 1990s. According to Nvidia's documentation, RTX NTC could enable games to use textures up to four times higher resolution than currently possible.
Surprisingly, the technology appears to have broad compatibility. While primarily designed for RTX series cards, testing shows it functions on GTX 10 series GPUs, AMD Radeon RX 6000 series, and Intel Arc graphics cards, suggesting potential widespread adoption across different platforms including gaming consoles.
The release date for the final version remains unannounced as Nvidia continues development and optimization of this promising technology.