Other NVIDIA's new texture compression tech

[cm2]

Read from tweaktown:

Called 'Random-Access Neural Compression of Material Textures'

The Neural Texture Compression (NTC) aims to tackle the spiraling requirements for high-res textures, allowing them to better fit into the available RAM. We're told that it delivers 16x more texels than Block Compression or BC (standard GPU-based texture compression).

In practical terms, though, a 4K render takes 1.15ms which isn't quick enough - compared to 0.49ms as measured on an RTX 4090 - so it'll need to be a good deal faster than this for NTC to work well enough to be usable.

Wronski a researcher at Team Green who's part of the project.
admitted that:

"This requires future work, but it will only get better, and I'd argue that it is already in a practical realm."

So, given some time and refinement, this could be a key technology for achieving better textures. Or rather, fitting high-res textures into more modest VRAM loadouts.

 

[froggyboy604]

NVIDIA's new texture compression tech sounds like good tech for playing VR and 4K games at higher video resolutions.

 

[cm2]

Read this post from nvidia and videocardz:

The continuous advancement of photorealism in rendering is accompanied by a growth in texture data and, consequently, increasing storage and memory demands. To address this issue, we propose a novel neural compression technique specifically designed for material textures. We unlock two more levels of detail, i.e., 16× more texels, using low bitrate compression, with image quality that is better than advanced image compression techniques, such as AVIF and JPEG XL. At the same time, our method allows for on-demand, real-time decompression with random access similar to block texture compression on GPUs. This extends our compression benefits all the way from disk storage to memory. The key idea behind our approach is compressing multiple material textures and their mipmap chains together, and using a small neural network, that is optimized for each material, to decompress them. Finally, we use a custom training implementation to achieve practical compression speeds, whose performance surpasses that of general frameworks, like PyTorch, by an order of magnitude.

— Random-Access Neural Compression of Material Textures, NVIDIA

Unlike common BCx algorithms, which require custom hardware, this algorithm utilizes the matrix multiplication methods, which are now accelerated by modern GPUs. According to the paper, this makes the NTC algorithm more practical and more capable due to lower disk and memory constraints.