DLSS 5: NVIDIA Aims for Photorealism in Video Games

NVIDIA once again raises the bar for real-time graphics by announcing DLSS 5, a new evolution of its AI-powered technology. This isn't just about improving performance: the stated goal is to bring video games closer to the visual quality typical of cinematic productions.

Even More Real and Immersive

At the heart of this innovation is a neural model capable of directly enhancing the final image, enriching lighting and materials in a way that is consistent with the original scene. Unlike traditional generative solutions, which are often unpredictable, DLSS 5 maintains deterministic behavior: each frame remains faithful to the 3D world created by developers, avoiding artifacts or inconsistencies between frames.

The system uses information such as color and motion vectors to reconstruct complex details – from skin to fabrics, up to advanced light management – all in real-time and up to 4K resolution. This is a significant step forward, considering that cinematic rendering requires enormously longer times compared to the 16 milliseconds of a game frame.

After revolutionizing the market with upscaling and frame generation, DLSS is now changing direction: less focus on the quantity of pixels, more on perceived quality. Industry support is already extensive. Among the first compatible titles are major productions such as Starfield, Assassin’s Creed Shadows, and Hogwarts Legacy, along with numerous other upcoming projects.

The debut is scheduled for autumn, with a first technical demonstration shown during this week's annual NVIDIA GPU Technology Conference (GTC) event.