NVIDIA's Deep Learning Super Sampling (DLSS) is an AI-powered upscaling technology designed to boost gaming performance while maintaining or even improving image quality. DLSS 4 represents a significant leap forward, primarily due to its adoption of advanced Transformer models for both Super Resolution and Ray Reconstruction. Unlike older Convolutional Neural Network (CNN) based models, these Transformer models offer superior image fidelity, detail retention, and temporal stability, fundamentally transforming how games are rendered at higher resolutions.
Within this cutting-edge framework, NVIDIA has introduced various presets, with Preset J and Preset K being the most discussed. These presets are essentially different configurations of the underlying AI model, each optimized for specific visual outcomes and performance characteristics. The choice between them often comes down to a user's preference for sharpness versus temporal stability and artifact reduction.
Preset J serves as the initial Transformer-based DLSS 4 model. It marks a substantial improvement over previous DLSS iterations, offering a noticeably sharper image with enhanced detail retention. For many users, Preset J provides a compelling visual experience, delivering crisp graphics and minimizing artifacts that were more prevalent in older DLSS versions.
However, despite its strengths, Preset J can exhibit some temporal instability, leading to minor ghosting or shimmering, particularly in scenes with rapid motion or intricate elements like vegetation. Some users have also noted that it can appear slightly over-sharpened, which, while providing a clear image, might introduce subtle artifacting in certain contexts. Generally, Preset J is well-regarded for its stability and high image quality, making it a reliable choice for a broad range of games.
Preset K is a direct refinement of Preset J, positioned by NVIDIA as the "latest" model for DLSS Super Resolution as of January 30, 2025. Its primary objective is to enhance temporal stability, reduce ghosting, and improve detail rendering during motion. This means that objects in fast motion or camera movements are rendered with greater clarity and fewer visual trails.
A notable improvement in Preset K is its ability to reduce vegetation shimmering—a common artifact in games when using upscaling technologies. While not entirely eliminated, Preset K handles this issue more effectively than Preset J, contributing to a more natural and stable visual experience. Users often report that Preset K, while potentially appearing slightly less sharp than Preset J, provides a smoother and more consistent image with fewer distracting artifacts like ghosting streaks or blurry trails during movement. However, user experiences can vary, with some reporting occasional blurriness or subtle banding in specific scenarios, particularly at lower resolutions.
The table below summarizes the key distinctions between DLSS 4 Preset J and Preset K, highlighting their strengths and areas of differentiation based on both official statements and user feedback:
Feature | DLSS 4 Preset J | DLSS 4 Preset K |
---|---|---|
Model Type | Baseline Transformer-based Super Resolution | Refined Transformer-based Super Resolution |
Temporal Stability | Good, but can exhibit minor ghosting/shimmering | Improved; significantly reduced ghosting and flickering |
Image Sharpness | Generally sharper, potentially slightly over-sharpened | Slightly less sharp but cleaner and more stable |
Detail in Motion | Effective, but may show some trailing artifacts | Enhanced detail, less trailing/blurring in motion |
Vegetation Shimmering | Can be noticeable in some titles | Reduced noticeably, leading to a more natural look |
Performance | Consistent, solid performance | Potentially slightly better efficiency on newer hardware, but higher compute requirements for older cards |
Ray Reconstruction | Still uses Preset J for "Latest" model | Primarily refines Super Resolution; Ray Reconstruction remains on Preset J |
Recommended Use | For users prioritizing maximum sharpness and clarity | For users prioritizing smoother motion, reduced ghosting, and better artifact handling |
The practical implications of choosing between Preset J and K often depend on the specific game, individual hardware, and personal visual preferences. While NVIDIA positions Preset K as a clear improvement, real-world testing has revealed a nuanced picture:
Some users find Preset J to offer a crisper image, which is desirable for games where fine details are crucial. Conversely, Preset K can, in some scenarios, lead to a slightly less sharp appearance or even blurriness and noticeable trails during very slow, minute movements. This suggests that the "enhanced detail in motion" of Preset K focuses more on overall fluidity during significant movement rather than absolute static sharpness.
While both presets are designed for efficient performance, some reports indicate that Preset K might offer slightly better frame rates in certain performance modes, especially on newer hardware. However, older RTX GPUs might experience a marginal performance cost with Preset K due to its increased AI compute requirements. This reinforces the idea that the "optimal" preset can be hardware-dependent.
To fully utilize Preset K and its refinements, users need to ensure they have the latest NVIDIA Game Ready Driver (version 572.16 or higher) and the NVIDIA App (version 11.0.2.312 or higher). The NVIDIA App allows users to override default DLSS settings and select the "Latest" model for Super Resolution, which defaults to Preset K. In some cases, for games not officially supported by the NVIDIA App's override function, users might need to manually replace the DLSS DLL file or use third-party tools like NVIDIA Profile Inspector to force the desired preset. It's important to note that while the Transformer model for Super Resolution (which includes J and K) is supported by all GeForce RTX GPUs, the Multi Frame Generation feature of DLSS 4 is exclusive to RTX 50 series cards.
User feedback on platforms like Reddit and Steam forums provides a valuable, albeit anecdotal, perspective. While many appreciate Preset K's improvements in temporal stability and artifact reduction, some report a preference for Preset J due to perceived consistent sharpness or unexpected visual anomalies with K. This highlights that the "best" preset can be subjective and game-dependent. Testing different presets on a per-game basis is often recommended to determine the ideal setting for an individual's gaming experience.
To further illustrate the nuanced differences between Preset J and K across various performance and image quality metrics, here's a radar chart. This chart is based on aggregated user experiences and official statements, providing a comparative overview of their perceived strengths.
This radar chart illustrates the perceived strengths of each preset. While Preset J excels in raw sharpness, Preset K demonstrates superior temporal stability and reduction of common artifacts like ghosting and vegetation shimmering, making it a stronger choice for dynamic, fast-paced games where motion clarity is paramount.
To further contextualize DLSS 4 and its presets, here's a mindmap illustrating the interconnected concepts and technologies that define this advanced AI upscaling solution.
This mindmap outlines how DLSS 4 operates, from its underlying Transformer models to the specific improvements offered by presets like J and K, and the practical considerations for users.
Understanding the technical differences between presets is one thing, but seeing their impact in real-time gaming is crucial. The following video offers a direct comparison of DLSS 4 Preset J versus Preset K across multiple games, allowing you to visually assess their impact on image quality and performance. This comparison helps illustrate the subtle yet significant improvements in temporal stability and artifact reduction that Preset K aims to deliver.
The video, titled "Preset K vs J - DLSS 4 Transformer - 5 Games Image quality comparison - RTX 3070 - 1440p," provides a direct side-by-side comparison. It's particularly relevant because it tests these presets on an RTX 3070 at 1440p, offering insights into how these advanced DLSS 4 models perform on current-generation hardware, not just the latest RTX 50 series. This practical demonstration allows viewers to observe the subtle differences in ghosting, shimmering, and overall image stability that differentiate Preset J and Preset K in dynamic gaming scenarios.
The distinction between DLSS 4 Preset J and Preset K encapsulates NVIDIA's ongoing commitment to refining AI-driven super-resolution technology. While Preset J laid the robust foundation with its significant advancements in image quality and sharpness using Transformer models, Preset K represents a targeted evolution, focusing on critical aspects of temporal stability, ghosting reduction, and motion clarity. This refinement aims to deliver an even smoother and more artifact-free gaming experience, especially in dynamic scenarios and around complex elements like vegetation. As the "Latest" preferred model, Preset K signifies a step towards increasingly fluid and visually consistent gameplay. However, individual results can vary, and users are encouraged to experiment to find the preset that best suits their specific gaming setup and visual preferences.