Nvidia is expanding DLSS 4.5 by bringing an improved Ray Reconstruction AI denoiser to all GeForce RTX graphics cards starting this August. Powered by a second-generation transformer model, the update delivers increased computing power and sharper ray-traced lighting without raising hardware requirements or costing additional performance.
When Nvidia rolled out DLSS 4.5 earlier in the year, the software package brought major headlines for Dynamic Multi Frame Generation, a 6X Multi Frame Generation mode, and a second-generation transformer model dedicated to Super Resolution. Yet one significant piece remained absent from that initial rollout: Ray Reconstruction, which continued relying on the first-generation transformer model introduced back in 2023.
How Transformer-Powered Ray Reconstruction Works
Traditional ray tracing and path tracing cast a limited number of rays per frame—far fewer than the total pixels on a screen—leaving the raw output full of visual noise. Conventional rendering methods rely on hand-tuned denoisers to clean up this noise before the upscaler runs, discarding vital lighting and material data from the game engine in the process. Ray Reconstruction replaces these traditional denoisers with an AI-powered model that merges denoising and upscaling into a single pass, giving the pipeline direct access to motion and geometry data.
The updated model shifts away from older convolutional neural network approaches by adopting a second-generation transformer architecture. Nvidia notes in company announcements that the upgraded denoiser maintains a similar performance cost to the previous version while building a deeper spatial awareness of in-game scenes and evaluating engine motion data more efficiently.
Broad Compatibility Across RTX Generations
Unlike other cutting-edge features in the software suite, the new Ray Reconstruction model does not require the latest hardware silicon. While multi-frame generation features remain restricted to newer graphics cards, the updated AI denoiser operates entirely on existing Tensor Cores. The feature is compatible with every GeForce RTX graphics card all the way back to the RTX 20-series.
Owners of older RTX hardware do not need to wait for individual title updates or developer patches to experience the upgrade. The feature functions as a driver-level override accessible through the Nvidia App, mirroring the flexibility of DLSS Super Resolution. Players can also manually swap out the necessary DLL files for titles that support custom overrides.
Visual Improvements in Major Path-Traced Titles
Hands-on testing with high-end hardware configurations reveals subtle yet distinct improvements across demanding path-traced environments. In Resident Evil Requiem, testing shows denser particle presentation during rainstorms, sharper ripple effects on the ground, and cleaner micro-details across character clothing and wet stonework compared to the older 3.5 version of the tech.
Other demanding technical showcase games demonstrate similar clarity gains. In Indiana Jones and the Great Circle, the updated model reduces distracting streaks in snow and particle effects. In Pragmata, laser lighting effects react faster with fewer visual artifacts, while Alan Wake 2 displays finer patterns on in-game televisions with increased stability.
Launch Lineup and Configuration
When the feature rolls out in August, it will arrive with native support across 27 titles. The initial software update includes compatibility for prominent ray-traced releases such as Cyberpunk 2077, Star Wars Outlaws, Avatar: Frontiers of Pandora, Hogwarts Legacy, Half-Life 2 RTX, and DOOM: The Dark Ages.
