Skip to content
Tech News
← Back to articles

Nvidia DLSS 4.5 Ray Reconstruction Is Better Almost Everywhere

read original more articles
Why This Matters

Nvidia's DLSS 4.5 Ray Reconstruction update significantly enhances ray tracing visuals by improving upscaling quality and reducing artifacts, all while maintaining high performance. This advancement benefits gamers and developers by delivering more realistic graphics with less computational overhead, pushing the boundaries of real-time rendering. The update underscores Nvidia's ongoing commitment to AI-driven graphics improvements, shaping the future of immersive gaming experiences.

Key Takeaways

Nvidia has finally updated Ray Reconstruction with the latest DLSS 4.5 technology, and we've spent the past week testing it. The new model improves upscaling quality when using Ray Reconstruction, further reduces some of the artifacts we see when ray tracing is enabled, and Nvidia claims it does all of this with virtually no impact on performance. Best of all, GeForce RTX owners can try the feature right now.

DLSS 4.5 Ray Reconstruction is the next iteration of Nvidia's AI-based denoiser for ray tracing. It's been updated to use the second-gen transformer architecture that first debuted with DLSS 4.5 Super Resolution earlier this year, making this the third iteration of Ray Reconstruction. Nvidia claims the new model is more efficient, delivering "35% more compute capability" while processing "20% more parameters" than the previous model, and it has also been trained on a larger dataset.

The DLSS 4.5 update to Ray Reconstruction is an important step forward because Ray Reconstruction isn't simply a denoiser. It effectively combines an upscaler and denoiser into a single model. When you enable Ray Reconstruction alongside DLSS upscaling, the normal DLSS upscaling pass is bypassed, and Ray Reconstruction handles both upscaling and denoising at the same time.

TL;DR: What is ray reconstruction and why is it an important upgrade?

This is a neat and efficient approach, but it has a drawback: whenever Nvidia releases a new DLSS upscaling model, the improvements in that model aren't immediately available when Ray Reconstruction is used.

This happened earlier this year when Nvidia launched DLSS 4.5 Super Resolution. You got all the benefits of the new DLSS 4.5 models when using upscaling on its own, but the moment Ray Reconstruction was enabled in a supported game, DLSS 4.5 was effectively replaced by the upscaler built into the Ray Reconstruction model. At the time, that meant falling back to DLSS 4.0-era tech using the first-gen transformer model.

That created a strange disconnect within the DLSS feature set. You could choose the new DLSS 4.5 upscaler, or choose Ray Reconstruction for denoising, but you couldn't use both at the same time.

That's been resolved with the release of DLSS 4.5 Ray Reconstruction. The upscaling component now uses the second-gen transformer architecture, so there's no longer a quality gap between Super Resolution and Ray Reconstruction.

As before, Ray Reconstruction is available across all RTX GPUs, including the RTX 20 and RTX 30 series. The main way to enable it right now is through the Nvidia App's DLSS override settings. Browse to a supported game, head to the DLSS override options, and change the Ray Reconstruction model preset to Preset F, which is the new DLSS 4.5 model. The other main option is Preset D from DLSS 4.0. There's no option to select Preset E or the original DLSS 3.5 models.

Upscaling Now Matches DLSS 4.5

... continue reading