Real time neural rendering technology evolves so that 15 out of every 16 pixels can now be generated by AI
Nvidia DLSS 4, the latest release of the suite of neural rendering technologies that use AI to boost 3D performance, will be supported in visualisation software – D5 Render, Chaos Vantage and Unreal Engine – next month.
The headline feature of DLSS 4, ‘Multi Frame Generation’, brings revolutionary performance versus traditional native rendering, according to Nvidia.
Multi Frame Generation is an evolution of Single Frame Generation, which was introduced in DLSS 3 to boost frame rates with Nvidia Ada Generation GPUs by using AI to generate a single frame between every pair of traditionally rendered frames.
In DLSS 4, Multi Frame Generation takes this one step further by using AI to generate up to three additional frames between traditionally rendered frames. The feature is available exclusively on the new Blackwell-based Nvidia RTX 50 Series GPUs, which were announced at CES today. According to Nvidia, this is because the Blackwell architecture has been built and optimised specifically for neural rendering.
Multi Frame Generation can also work in tandem with other DLSS technologies including super resolution (where AI outputs a higher-resolution frame from a lower-resolution input) and ray reconstruction (where AI generates additional pixels for intensive ray-traced scenes). When these technologies are combined, it means 15 out of every 16 pixels are generated by AI, which is much faster than rendering pixels in the traditional way.
According to Nvidia, in D5 Render enabling DLSS 4 can lead to a four-fold increase in frame rates, leading to much smoother navigation of complex scenes.
Nvidia has also enhanced its Single Frame Generation model that works with Nvidia Ada Generation and the new Nvidia Blackwell GPUs. According to Nvidia, this results in increased performance and lower VRAM usage.
Finally, Nvidia has integrated neural networks inside of programmable shaders to create what it calls neural shaders. According to Nvidia, neural shaders can be used to compress textures by up to 7X, saving massive amounts of graphics memory.
