Skip to content
IFAI

AI tools

NVIDIA debuts DLSS 3.5 with Ray Reconstruction

Published Reach 39

NVIDIA just introduced DLSS 3.5 at Gamescom, featuring a new technology called Ray Reconstruction. Instead of relying on hand-tuned denoisers, it uses an AI network trained on an NVIDIA supercomputer with 5x more data than DLSS 3 to enhance reflections and global illumination across all RTX GPUs. The upgrade will debut this fall in titles like Cyberpunk 2077: Phantom Liberty.

Image: blogs.nvidia.com

Real-time ray tracing made game worlds prettier, but it also introduced a dirty little secret: traditional hand-tuned denoisers often blur water reflections, smear fine details, and leave annoying ghost trails.

On August 22 at Gamescom, NVIDIA officially unveiled DLSS 3.5. Instead of just pushing for higher frame rates, the company is bringing in an AI overhaul to fix those messy ray-traced lighting artifacts once and for all.

Goodbye denoisers, hello Ray Reconstruction

The marquee upgrade in DLSS 3.5 is Ray Reconstruction. In ray-traced rendering, GPUs can only fire a limited number of rays per frame to maintain playable performance. Hand-tuned denoisers are usually tasked with filling in the missing gaps, but they often struggle to distinguish between genuine reflections and visual noise, scrubbing out crisp details in the process.

NVIDIA tackled this by training an AI neural network on a supercomputer using five times more data than DLSS 3. This network learns to recognize various lighting patterns—from reflections to global illumination—and makes smarter choices across spatial and temporal data.

The resulting DLSS 3.5 suite brings together Ray Reconstruction, Super Resolution, Deep Learning Anti-Aliasing (DLAA), and Frame Generation. The best part? Ray Reconstruction isn't locked behind the latest hardware. Every single GeForce RTX GPU owner can take advantage of it.

The feature is already lined up for this fall's heavy hitters, including Cyberpunk 2077: Phantom Liberty, Alan Wake 2, and Portal with RTX. 3D creators also benefit: apps like D5 Render, Chaos Vantage, and Omniverse will integrate the technology, saving artists from waiting hours for full renders.

Emotional NPCs and cloud expansions

NVIDIA's gaming showcase didn't stop at graphics cards.

For game characters, the NVIDIA Avatar Cloud Engine (ACE) introduced NeMo SteerLM. This training technique lets developers tune NPC personalities and emotional ranges, allowing in-game characters to engage in natural language dialogues that feel reactive rather than scripted.

Meanwhile, four top modding teams are actively rebuilding Valve's classic shooter under the Half-Life 2 RTX project using RTX Remix, packing full ray tracing, DLSS 3, and Reflex.

On the cloud front, GeForce NOW completed its rollout of RTX 4080 SuperPODs across North America and Europe, alongside imminent Microsoft Store integration to stream select Xbox PC Game Pass titles.

Why it matters

DLSS 3.5 marks a critical shift in how GPUs use machine learning. Up to now, gamers viewed DLSS primarily as a speed hack—rendering at lower resolutions or synthesizing frames to juice FPS counters. Ray Reconstruction proves that neural rendering can actually deliver superior image fidelity compared to decades of handcrafted graphics algorithms.

Opening Ray Reconstruction to all RTX generations is a calculated, welcome move. After facing pushback for locking DLSS 3 Frame Generation to RTX 40-series cards, NVIDIA is throwing a massive bone to older 20 and 30-series owners. It lowers the barrier for studios to adopt advanced path tracing, knowing millions of legacy RTX users will actually benefit.

Graphics rendering is no longer just about brute-force silicon; it is rapidly turning into an AI software arms race. The only catch is that NVIDIA's branding is getting dizzyingly fragmented: explaining to regular gamers which feature runs on which generation is becoming a puzzle all on its own.

Sources (1)

  1. [1]
    blogs.nvidia.com(opens in a new tab)Originalblogs.nvidia.com

Best of AI, daily