NVIDIA launches DLSS 5 and delivers photorealistic lighting with AI | Ukraine news
NVIDIA launches DLSS 5 and delivers photorealistic lighting with AI | Ukraine news
https://mezha.net/eng/bukvy/nvidia_launches_dlss_5/
Publish Date: 2026-03-17 03:00:00
Source Domain: mezha.net
NVIDIA showcased a neural rendering upgrade that promises cinema-quality lighting in real time, but the demos relied on very high-end hardware. Gamers and developers will watch whether DLSS 5 can deliver visible gains across titles without demanding dual top-end GPUs.
After the announcement of DLSS 4.5 at CES, NVIDIA immediately promised a new step in image upscaling technology, and now the company has introduced DLSS 5. The main emphasis is on artificial intelligence: it is claimed that DLSS 5 “infuses pixels with photorealistic lighting and materials” using a real-time neural rendering model, and a release is expected this fall.
What does this mean in practice? During a presentation at GTC 2026, NVIDIA CEO Jensen Huang demonstrated the technology with examples from Resident Evil: Requiem, Hogwarts Legacy, and Starfield. DLSS 5 adds noticeable detail to characters’ hair and skin tones, but it looks as if they were compared with the same games without DLSS features enabled. It’s unclear how big the difference is compared to DLSS 4.5 with ray tracing and all of its capabilities.
DLSS 5 takes the color and motion vector of each frame as input data and uses an artificial intelligence model to provide the scene with photorealistic lighting and materials tied to the source of 3D content and synchronized from frame to frame.
– NVIDIA
The technology operates in real time and can render images up to 4K resolution, the company says.
During the demonstration, Huang ran a system with two RTX 5090 GPUs. In the future, DLSS 5 is expected to run on a single graphics card – but as reviewers speculate, it would require nearly the same power as two 5090s. NVIDIA also describes DLSS 5 as a step toward delivering “cinematic-quality” real-time visuals without the power requirements studios previously demanded. This sounds like a combination of generative AI with direct developer control, rather than just a set of AI prompts.
