NVIDIA has released DLSS 4.5 Ray Reconstruction, upgrading the denoiser to a second-generation transformer model. This new model processes 20% more parameters and delivers 35% more compute capability than its predecessor while maintaining similar performance costs. Unlike other DLSS 4.5 features such as Multi Frame Generation—which are exclusive to RTX 50-series—Ray Reconstruction supports all GeForce RTX GPUs back to the RTX 20-series.

Ray Reconstruction replaces traditional hand-tuned denoisers with an AI model that merges denoising and upscaling into a single pass, preserving engine data like ray and material information. The updated model was trained on a larger dataset, offers deeper spatial awareness, and provides developers finer control over temporal sampling. It is available immediately via the NVIDIA App as a driver-level override (Preset F) or through manual DLL replacement.

Testing across several titles shows the new model enhances ray-traced visuals with crisper, richer textures, particularly in path-traced scenarios. The upgrade requires DLSS Super Resolution to function and works with DLAA or native resolution. NVIDIA confirms no new hardware requirements were added, making this a free quality uplift for the entire RTX ecosystem.