Nvidia DLSS 5 Gets Major Performance Boost with 2x Faster AI Model and Reduced VRAM Usage

Nvidia has announced significant performance improvements to its DLSS 5 neural rendering technology, marking a pivotal refinement just weeks after the AI-powered graphics technology officially launched in NBA 2K27. The updated AI model delivers twice the performance of the September release while substantially reducing hardware demands on RTX 50-series GPUs and upcoming RTX Spark laptops.1

1

DLSS 5 AI Model Achieves 2x Performance Gains

The October update to Nvidia DLSS 5 represents a dramatic leap in efficiency for the AI-powered graphics technology. Nvidia has reduced the neural rendering model's footprint by 24% to approximately 118 million parameters, down from roughly 155 million in the previous implementation.3

3

This streamlined faster AI model achieves comparable visual fidelity while cutting GPU overhead in half compared to the September version.4

4

According to Nvidia's Andrew Burnes, the company achieved these performance improvements through "systematic pipeline optimizations and neural model refinements."3

3

The reduced 4K VRAM usage of 20% makes the technology less demanding on graphics cards, addressing one of the primary criticisms since DLSS 5's controversial debut earlier this year.5

5

The neural rendering technology now runs 10x faster than the original version demonstrated in March 2026, when it required a pair of RTX 5090 GPUs to function.4

4

In NBA 2K27, where the optimized model debuts on October 16, frame rates improve by up to 40%.5

5

Frame Generation Performance Mode Delivers Additional Speed

Beyond the core DLSS 5 optimizations, Nvidia introduced a new Frame Generation Performance Mode called Preset C through the Nvidia app beta. This feature works on both RTX 40 Series and RTX 50-series GPUs, using a lighter algorithm that replaces the secondary AI model previously used to calculate optical flow for elements like UI or ray-traced effects.2

2

The new DLSS frame generation algorithm boosts performance by up to 35% on the RTX 5060, according to Nvidia's internal testing.1

1

In Cyberpunk 2077 running at Medium RT with DLSS at Balanced on a 9800X3D system with 64GB DDR5, frame generation set to 2x achieved 132 FPS compared to 120 FPS with the original algorithm. At 4x, performance jumped from 197 to 244 FPS, while 6x mode increased from 255 to 348 FPS.2

2

Nvidia acknowledges the faster algorithm involves "minor trade-offs in image quality," though the company maintains these effects are subtle.1

1

The streamlined approach reduces the impact on native rendering, potentially resulting in higher underlying frame rates and fewer typical frame generation artifacts.2

2

Hogwarts Legacy and Visual Refinements Address Early Criticism

When Nvidia first showcased DLSS 5 earlier this year, the neural rendering technology faced immediate backlash from gamers who felt the company had "lost the plot" by bringing generative AI into video games. The most glaring issues involved facial transformations—a 15-year-old student from Hogwarts Legacy appeared much older, while Grace from Resident Evil Requiem looked artificially enhanced.1

1

Nvidia DLSS 5 arrives in Hogwarts Legacy on November 5 with significantly toned-down visual changes. New comparison screenshots show facial alterations are now as subtle as those in NBA 2K27, with teenagers no longer resembling middle-aged adults and environmental modifications appearing more restrained.1

1

Developer Avalanche Software adjusted the integration to maintain the game's intended visual style.5

5

The technology will also come to Where Winds Meet in late November, and RTX Remix gains support for the optimized model on October 23, allowing modders to apply neural rendering to compatible remastering projects.5

5

Nvidia maintains that developers retain "detailed artistic control" over DLSS 5 implementation in their games.1

1

What These Changes Mean for Gamers

The rapid performance improvements signal Nvidia's responsiveness to criticism about DLSS 5's hefty resource requirements. The technology initially demanded powerful RTX 50-series GPUs to run smoothly, creating accessibility concerns as graphics card prices continue climbing.4

4

However, questions remain about backward compatibility. Gamers are asking whether the optimized AI model will enable official support for RTX 40 Series cards, given the dramatic efficiency gains. Nvidia has promised to work on bringing support to last-generation GPUs after optimizing RTX 5000 performance, though this likely won't materialize until early next year.4

4

The Frame Generation Performance Mode proves particularly valuable for RTX Spark laptops and users with modest GPUs like the RTX 5060, where every frame rate boost matters. While some users report increased glitching and artifacts with Preset C, experiences vary by game and individual sensitivity to visual anomalies.4

4

As DLSS 5 matures, these early optimizations represent where the largest gains occur. Future performance improvements will likely become more incremental as the neural rendering technology approaches optimization limits. Independent benchmarks will determine real-world performance gains and image quality trade-offs across different games and GPU configurations.5

5

The technology's expansion to GeForce Now and upscaling applications through RTX Remix suggests Nvidia views neural rendering as central to its graphics strategy moving forward.