NVIDIA DLSS 5登場！黃仁勳：電腦圖學的「GPT時刻」，AI神經渲染讓遊戲畫面跨入電影級紀元

Topics Artificial wisdom Autopilot network Processor 手機 exhibition activities About us 2026 / 03 / 17 12:10 Tuesday Login TOPICS ARTIFICIAL WISDOM AUTOPILOT NETWORK PROCESSOR 手機 EXHIBITION ACTIVITIES ABOUT US Discover more 相機 Tech repair services VR AR headsets This is an advertisement. Home exhibition NVIDIA DLSS 5 is here! Jensen Huang: The "GPT moment" of computer graphics, AI neural rendering brings game graphics into the cinematic era. It claims to completely bridge the gap between real-time rendering and Hollywood-level visual effects. Author: Mash Yang 2026-03-17in exhibition, Market dynamics, audio and video, Processor, observe, software A A 0 Share to Facebook Share on Twitter Share to LINE

At GTC 2026, NVIDIA CEO Jensen Huang officially announced DLSS 5, the next-generation deep learning super sampling technology. This is not merely a version number update; Huang described it as "the most significant breakthrough in computer graphics since the advent of real-time ray tracing in 2018." DLSS 5 introduces a completely new real-time neural rendering model, capable of injecting realistic lighting and materials into every pixel, claiming to completely bridge the gap between real-time rendering and Hollywood-level visual effects.

From "frame interpolation" to "light enhancement": The stunning transformation of DLSS 5

Since its introduction with the RTX 20 series graphics cards in 2018, DLSS technology has undergone multiple iterations, from the initial resolution upscaling to the later frame generation. Its core goal has always been to improve game smoothness without sacrificing too much image quality.

However, DLSS 5's ambitions are clearly much greater.

In his keynote address, Jensen Huang stated, "25 years ago, NVIDIA invented programmable shaders; today, we are reinventing computer graphics once again. DLSS 5 is the GPT moment in graphics—it merges handcrafted rendering with generative AI, achieving a leap in visual realism while retaining the control artists need for creative expression."Natural Language Processing

The core of this new technology is that it no longer passively processes pre-rendered images, but truly "understands" the scene.

The neural network model of DLSS 5 is trained end-to-end and can identify complex scene semantics by analyzing a single frame, such as characters, hair, cloth, translucent skin, and different ambient lighting conditions such as front lighting, backlighting, and cloudy days.

Based on this understanding, it can instantly generate photorealistic pixels while preserving the original scene structure and artistic style, accurately handling subsurface scattering under the skin, the delicate luster of fabrics, and the complex interaction between hair and light.

The brute-force output of computing power has reached its limit; AI is the only way out.

At the conference, NVIDIA pointed out that from the introduction of programmable shaders in the GeForce 3 in 2001, to the introduction of CUDA in the GeForce 8800 GTX in 2006, to the Turing architecture enabling real-time ray tracing technology in 2018, and the implementation of path tracing in the RTX 5090 in 2025, the computing power of graphics cards has increased dramatically by an astonishing 37.5 times in 25 years.

However, the fundamental challenge of real-time rendering lies in time. A game frame must complete all calculations within 16 milliseconds (60 FPS), or even as short as 8 milliseconds (120 FPS), while a photorealistic VFX frame in a Hollywood movie may take minutes or even hours to render. This gap in computing resources cannot be bridged simply by stacking more hardware.

NVIDIA believes that DLSS 5 is the key to solving this problem. It elevates the AI ​​model from a role of "assisting performance enhancement" to the core of "dominant image quality generation." In fact, it was unveiled at CES 2026 earlier this year...DLSS-4.5In the past, AI has already taken charge of drawing 23 out of 24 pixels on the screen. Now, DLSS 5 goes a step further, allowing AI not only to draw, but also to make the image "beautiful".

DLSS 5 as seen by developers: Unprecedented creative freedom

To ensure that this AI-driven enhancement does not deviate from the game developers' original intentions, NVIDIA provides granular controls. Developers can adjust intensity, color mapping, and masking to precisely determine the area and method of applying AI effects, maintaining the unique aesthetic style of each game.

This technology has gained the support of many of the world's leading game developers and publishers, including Bethesda, CAPCOM, NetEase, Tencent, Ubisoft, and Warner Bros. Games.

Todd Howard, director of Bethesda Game Studios, shared his initial experience of implementing DLSS 5 in Starfield: "When NVIDIA showed us DLSS 5, we put it into Starfield, and it was amazing to see the graphics come to life. We've already played it, and we can't wait for you to try it too."

CAPCOM Executive Producer Jun Takeuchi