NVIDIA has announced DLSS 5, its next-generation upscaling technology set for release this fall. The reveal comes shortly after the debut of DLSS 4.5, signaling a rapid pace of development. The company states this iteration uses a real-time neural rendering model to generate lighting and surface textures that approach photographic quality.

During a demonstration at GTC 2026, CEO Jensen Huang presented the technology running on titles including *Resident Evil: Requiem* and *Starfield*. The preview highlighted enhanced detail in elements like character hair and skin. However, the comparison was made against versions without any DLSS active, leaving open the question of its improvement over a fully-enabled DLSS 4.5 setup.

According to NVIDIA, the system analyzes a game's color and motion data frame-by-frame. An AI model then synthesizes lighting and material effects that are tied to the original 3D assets, maintaining consistency across the sequence. The process happens in real time and supports resolutions up to 4K.

Huang's demo utilized a dual RTX 5090 configuration, though NVIDIA says a future single-GPU implementation is planned. The broader ambition, as presented, is to achieve visual fidelity comparable to offline film rendering, but in real time on consumer hardware. It suggests a shift toward AI models that game developers can steer directly, rather than relying solely on traditional rendering pipelines.

NVIDIA has labeled DLSS 5 the most significant graphics advancement since real-time ray tracing. Yet with ray tracing still not universally adopted, the industry will be watching to see if players embrace this new AI-driven approach to pixel generation.

Source: Engadget