This web page was created programmatically, to learn the article in its unique location you possibly can go to the hyperlink bellow:
https://www.tomshardware.com/pc-components/gpus/nvidia-ai-tech-claims-to-slash-vram-usage-by-85-percent-with-zero-quality-loss-neural-texture-compression-demo-reveals-stunning-visual-parity-between-6-5gb-of-memory-and-970mb
and if you wish to take away this text from our website please contact us
As video games grow to be extra advanced and photorealistic, the business has more and more relied on upscaling know-how to fulfill surging {hardware} calls for. One of the largest points arising from this subpar optimization is VRAM utilization, which has risen sharply over the previous few years. To fight this, Nvidia has developed a know-how known as “Neural Texture Compression” (NTC), which was introduced up once more in at present’s GTC speak. The greatest graphics playing cards will be capable to leverage Nvidia’s NTC know-how.
Go deeper with TH Premium: GPUs
In the instance under, Nvidia ran a Tuscan Villa Scene that was consuming 6.5 GB of VRAM with customary block compression, however switching to NTC diminished that to simply 970 MB, and the picture appears equivalent. Previously, one other demo from the corporate confirmed a flight helmet with 272 MB of uncompressed textures — block compression minimize that right down to 98 MB, however NTC diminished it to simply 11.37 MB, about 24x lower than the unique.
Article continues under
The firm additionally demonstrated Neural Materials, following the identical idea: letting a neural community consider and decompress materials texture knowledge as an alternative of counting on computationally costly BRDF math. Typically, a number of texture maps are stacked for a fabric, and the GPU should calculate how mild interacts with every layer concurrently within the rendering pipeline.
Neural Materials simply asks the neural community how the sunshine will react in that situation and shades the pixel accordingly. The neural community is educated on all the feel knowledge, so it already is aware of the consequence given the sunshine and angle. As such, within the demo scene under, Nvidia achieved as much as 7.7x sooner render occasions at 1080p decision with no loss in picture high quality.
Image 1 of 2
NTC is so environment friendly as a result of it makes use of matrix acceleration engines, that are a separate {hardware} block in trendy GPUs, so base efficiency is not affected. Nvidia calls them Tensor Cores, Intel calls them XMX engines, and AMD calls them AI accelerators. This is the place upscalers like DLSS, FSR, and XeSS additionally dwell, as they reconstruct a low-res body right into a higher-resolution output, so it is a part of Nvidia’s neural rendering ambition.
The idea of neural rendering would not have the widest acclaim in the neighborhood but, and the phrase “neural network” would possibly lead you to suppose that is simply extra AI slop. It’s really the alternative, and one of many higher makes use of of AI since it is not generative in any respect. NTC will probably be educated solely on the precise set of textures it must reference throughout recreation improvement, so there is not any probability of hallucination.
Image 1 of 2
Textures by far eat probably the most VRAM in any recreation, so any method to maintain them in test is a welcome addition. That stated, it is essential to notice that this is not unique to Nvidia, as Microsoft has standardized it as “Cooperative Vectors” in DirectX. Intel has beforehand proven off its personal demo with noticeably higher textures in comparison with block compression. AMD final talked concerning the tech in 2024, however it’s possible onboard the mission as nicely.
Currently, no recreation helps Cooperative Vectors or Nvidia’s Neural Texture Compression, however we should always begin to see it applied quickly, given the business’s trajectory. AI has grow to be the reply to seemingly each age-old downside, and firms are inventing new methods to include it the place it would not belong. Innovations like NTC, nonetheless, present that it may be applied tastefully to make an precise, significant distinction.
Follow Tom’s Hardware on Google News, or add us as a preferred source, to get our newest information, evaluation, & opinions in your feeds.
This web page was created programmatically, to learn the article in its unique location you possibly can go to the hyperlink bellow:
https://www.tomshardware.com/pc-components/gpus/nvidia-ai-tech-claims-to-slash-vram-usage-by-85-percent-with-zero-quality-loss-neural-texture-compression-demo-reveals-stunning-visual-parity-between-6-5gb-of-memory-and-970mb
and if you wish to take away this text from our website please contact us
