This era of 8GB VRAM screen cards are also left over by NTC technology.
NVIDIA introduced a new technology at GTC 2026, Neural Texture Compression (NTC), designed to solve the big problems of the modern gaming industry like the overuse of VRAM, especially in games with realistic graphics and a lot of detail.
.
NTC uses the Machine Learning concept instead of traditional compression. It does not store the full texture in memory, but converts it to small data (latent features) and gives the neural network on the GPU to calculate the image details in real time. It makes much less use of VRAM to maintain the same quality. The key point is that this system is not Generative AI, but a deterministic system that produces the same results every time.
.
According to the test, the Tuscan Villa scene that originally used 6.5GB of VRAM when using NTC reduced to only 970MB, or about 85%. The image remained almost different from the original. Another example is that the 272MB texture was shortened to 11.37MB, or 24 times smaller. In addition to reducing the use of VRAM, it also allows smaller game files, faster loading, and reduced patch size.
.
NVIDIA also released Neural Materials, which uses AI to calculate reflectivity instead of traditional physics equations, resulting in image rendering at 1080p resolution up to 7.7 times faster, maintaining the same quality.
.
These technologies run on Tensor Cores or specialized processors, thus not compromising the core performance of GPUs, and even though no games are currently supported, Microsoft has already started developing standards in DirectX, while AMD and Intel have similar approaches, making them highly likely to be implemented in the near future.
.
Source: wccftech












































































































