NVIDIA unveils DLSS 5; Jensen Huang hails it as a “GPT moment” for graphics

Overview

NVIDIA (英伟达) today used its GPU Technology Conference keynote to unveil DLSS 5, a generative-AI-driven real-time rendering system CEO Jensen Huang (黄仁勋) called the “GPT moment” for graphics. Huang framed the release as the biggest evolution in 25 years since programmable shaders, promising a leap toward Hollywood‑level photorealism in games while preserving artist control. NVIDIA says DLSS 5 will roll out this autumn.

How it works

At the core is a real‑time neural rendering model that ingests per‑frame color and motion vectors and then synthesizes realistic lighting, materials and subtle effects such as hair sheen and subsurface scattering. NVIDIA presented DLSS 5 as capable of maintaining frame‑to‑frame coherence — a long‑standing weakness of generative video models — and of running in real time at up to 4K. The company also emphasised developer controls for intensity, color grading and masking so artists can tune the AI’s output rather than surrender authorship.

Industry support and availability

It has been reported that major studios and publishers including Bethesda (B社), Capcom (卡普空), Ubisoft, Tencent (腾讯) and NetEase (网易) have already committed to integrate DLSS 5. Players are expected to see the feature in high‑profile titles this fall — IT Home listed games such as Starfield (《星空》), Assassin’s Creed: Shadows (《刺客信条：影》) and Resident Evil: Requiem (《生化危机：安魂曲》) among early adopters.

A global context

This technical breakthrough arrives amid a fraught geopolitical backdrop. NVIDIA is a U.S. company that in recent years has faced tighter export controls on its most advanced AI accelerators, which has complicated access to top‑end GPUs in some markets. Will real‑time generative rendering reshape game visuals worldwide, or will hardware and policy limits slow adoption in parts of the globe? Huang says the moment is here — the industry will now put that claim to the test.