NVIDIA GeForce RTX 50: DLSS 4.5, Path Tracing, and Everything You Need to Know

Nvidia has unveiled DLSS 4.5, a significant evolution in its deep learning super sampling technology, marking a shift toward more sophisticated AI-driven rendering that aims to redefine visual fidelity and performance. Announced at CES, the update introduces a second-generation transformer model designed to enhance image quality across the entire RTX graphics card lineup, while reserving its most aggressive performance boosts for the new GeForce RTX 50-series hardware.

For the global gaming community, the arrival of GeForce RTX 50 DLSS 4.5 represents more than just a incremental frame rate bump; This proves a fundamental change in how the GPU predicts and generates pixels. By leveraging a transformer model trained on a significantly larger dataset and utilizing five times the compute power of its predecessor, Nvidia is targeting the most persistent artifacts in AI upscaling, such as shimmering and ghosting, which have historically plagued high-motion scenes.

This update creates a two-tiered benefit system. While current RTX owners can immediately access improved image stability and anti-aliasing, those upgrading to the Blackwell-based RTX 50-series will unlock a new frontier of fluid motion through expanded frame generation capabilities. As AI continues to move from the cloud into the local silicon of the GPU, the gap between traditionally rendered frames and AI-generated imagery is narrowing.

The Second-Gen Transformer: Solving the Artifact Problem

At the core of DLSS 4.5 is the transition to a second-generation transformer model. In computer science, transformers are architectures designed to handle sequential data by weighing the importance of different parts of the input—a technology most famously used in large language models. Nvidia has applied this to temporal rendering, allowing the GPU to better understand the relationship between frames to produce a cleaner image.

These architectural changes are designed to reduce visual noise and improve the precision of anti-aliasing. The result is a more stable image in challenging environments, particularly in titles with complex lighting and high-contrast edges. According to Nvidia’s official documentation, these neural rendering technologies are backed by an AI supercomputer in the cloud, ensuring that the models can be iteratively improved and deployed to users via driver updates.

The impact is already being felt in a library of over 400 supported titles. Specific games highlighted for seeing notable improvements include Kingdom Come: Deliverance 2, Indiana Jones and the Great Circle, and The Elder Scrolls IV: Oblivion Remastered. By refining how the AI handles “challenging scenes,” Nvidia aims to make the synthetic nature of DLSS nearly invisible to the end user.

Unlocking the RTX 50-Series: 6x Multi Frame Generation

While the image quality upgrades are universal for RTX users, the most disruptive feature of DLSS 4.5 is exclusive to the GeForce RTX 50-series. Nvidia is introducing 6x Multi Frame Generation (MFG), a technology that allows the hardware to generate five intermediate frames for every single traditionally rendered frame.

This is a substantial leap from the previous 4x limit, effectively multiplying the perceived smoothness of the gameplay. To accompany this, Nvidia has introduced Dynamic Multi Frame Generation. This system allows the GPU to adjust the number of generated frames in real-time to match the refresh rate of the user’s monitor, reducing the risk of screen tearing and minimizing the perceived latency that often accompanies high-level frame interpolation.

This combination of 6x MFG and Dynamic MFG is designed to push framerates to levels that were previously unattainable without sacrificing native resolution. By shifting the heavy lifting of frame creation to the AI tensor cores, the RTX 50-series can maintain high visual fidelity while delivering an ultra-fluid experience, provided the underlying engine can handle the rapid injection of synthetic frames.

Deployment and Hardware Requirements

Accessing the new features requires a specific software path. The fidelity and stability improvements of DLSS 4.5 are bundled with the GeForce 591.74 driver, which is available for download now. However, users cannot simply install the driver to see the changes; they must opt into the Nvidia app beta via the Settings > About menu to enable the DLSS 4.5 options.

The Nvidia app now includes the ability to automatically upgrade older versions of DLSS within supported titles, streamlining the process for users who have large libraries of legacy RTX games. This automation ensures that the second-gen transformer model is applied consistently across different game engines without requiring individual developers to patch every single title manually.

DLSS 4.5 Feature Breakdown

The Performance Trade-off for Older Hardware

Despite the image quality gains, the transition to a more compute-intensive transformer model has not been without costs for older architectures. Early testing indicates a performance divergence based on the GPU generation. While RTX 40 and 50-series cards maintain strong gains over native rendering, older cards may struggle with the increased overhead of the new AI model.

Reports suggest that mid-range Blackwell GPUs, such as the RTX 5070, may see a slight performance dip compared to DLSS 4, while older Ampere and Turing models can experience more significant slowdowns. In some instances, the performance loss on these older series is reported to be between 20% and 30%, potentially leading to negative scaling where the AI overhead outweighs the rendering benefit. Users on older hardware should carefully benchmark their specific titles to determine if the image quality improvements justify the potential drop in raw FPS.

What This Means for the Future of Path Tracing

The integration of DLSS 4.5 is a critical stepping stone for the wider adoption of path tracing. Path tracing, the “holy grail” of rendering, simulates the physical behavior of light to create photorealistic environments, but it is incredibly demanding on hardware. By increasing the efficiency of frame generation to 6x, Nvidia is effectively lowering the “performance tax” of path tracing.

When a GPU can generate five frames for every one it renders, the massive computational cost of calculating every light bounce in a scene becomes manageable. This suggests that future titles will be able to enable full path tracing by default, relying on DLSS 4.5 to fill the gaps and maintain a playable frame rate. The synergy between AI-generated frames and physical light simulation is where the industry is heading, moving away from “faked” lighting toward true simulation.

The next major milestone for this technology is expected this spring, when the 6x Multi Frame Generation and Dynamic MFG features officially roll out for RTX 50-series owners. Until then, users can utilize the 591.74 driver and the Nvidia app beta to experience the image quality improvements of the second-generation transformer model.

Do you think AI-generated frames are the future of gaming, or do they introduce too much latency for competitive play? Share your thoughts in the comments below.