At GTC, NVIDIA introduced DLSS 5, the next iteration of its Deep Learning Super Sampling technology, marking a shift from performance-focused upscaling to AI-driven image generation. The update is scheduled to arrive this fall.
Unlike previous versions, DLSS 5 moves beyond reconstructing pixels and frame generation. It introduces a real-time neural rendering model that enhances scenes with photorealistic lighting and material properties. The system uses a game’s color and motion vector data per frame, then applies an AI model to produce consistent, physically plausible visuals anchored to the original 3D content.
The company positions this as a transition point for graphics rendering. Since its debut in 2018, DLSS has primarily been used to improve performance by rendering at lower resolutions and reconstructing higher-resolution images using AI. It later expanded into frame generation, effectively inserting additional frames between rendered ones. NVIDIA says DLSS is now integrated into more than 750 games.
Earlier this year at CES, DLSS 4.5 pushed that concept further, with AI reportedly generating the majority of on-screen pixels. DLSS 5 builds on that trajectory, focusing less on performance gains and more on visual fidelity.
The key difference is how images are formed. Traditional rendering pipelines rely heavily on precomputed lighting models and rasterization techniques. DLSS 5 instead applies neural networks to simulate lighting and materials dynamically in real time, aiming to match the kind of realism typically associated with offline rendering used in film production.
NVIDIA says DLSS 5 can run at up to 4K resolution while maintaining interactive frame rates, though specific performance metrics were not disclosed.
Support is already lined up across major publishers and studios, including Bethesda, Capcom, NetEase, Tencent and Ubisoft. Confirmed titles include Starfield, Assassin’s Creed Shadows, Hogwarts Legacy, NARAKA: BLADEPOINT and The Elder Scrolls IV: Oblivion Remastered, among others.
Early partner feedback suggests a focus on immersion and visual consistency. Developers highlight improvements in lighting accuracy, material realism and overall scene cohesion, areas where traditional rendering techniques often require compromises between performance and quality.
From a market perspective, DLSS 5 signals NVIDIA’s continued push to define graphics pipelines around AI. Competing technologies such as AMD’s FSR and Intel’s XeSS have largely followed the same trajectory in upscaling and frame generation. Neural rendering, however, represents a deeper shift that could further widen the gap depending on hardware requirements and developer adoption.
For players, the impact will depend on implementation. If DLSS 5 delivers consistent visual improvements without introducing artifacts or latency, it could redefine expectations for real-time graphics. If not, it risks becoming another optional feature in an increasingly complex rendering stack.