One of the benefits of choosing an Nvidia graphics card is DLSS or Deep Learning Super Sampling (DLSS). DLSS has revolutionised the gaming and graphics landscape, providing an innovative solution to enhance visual fidelity while maintaining optimal performance. In this comprehensive guide, we delve into the intricacies of DLSS, exploring its workings, benefits, and the impact it has had on the gaming industry.
Table of Contents
How does DLSS work
DLSS, short for Deep Learning Super Sampling, is an advanced anti-aliasing technology developed by Nvidia. At its core, DLSS employs the power of deep learning and artificial intelligence to address aliasing artefacts in rendered images. It operates by employing a two-pronged approach: supersampling and deep learning. Supersampling involves rendering images at a higher resolution than the native resolution, providing more detailed information about the scene. The deep learning aspect utilises machine learning models trained with high-resolution scans and extensive datasets. These models, known as neural networks, learn to reconstruct missing details in lower-resolution images.
A critical component of DLSS is Nvidia’s specialised hardware called Tensor cores. These cores are exclusive to Nvidia’s RTX GPUs and are designed for accelerated matrix multiplication, making them ideal for the complex computations involved in deep learning. As generations progress, such as the RTX 40-Series, advancements in Tensor cores contribute to a substantial increase in per-core performance.
One distinctive feature of DLSS is its offline processing capability. Unlike traditional methods that require rendering higher-resolution images locally on the user’s computer, DLSS performs these computations offline. This approach significantly reduces the computational load, enabling real-time application of supersampling without compromising performance.
DLSS versions
DLSS originated as a technology designed to upscale games from lower to higher resolutions, aiming to enhance frame rates while maintaining acceptable image quality. Over its evolution, DLSS has incorporated five key features: DLSS Super Resolution, DLSS Deep Learning Anti-Aliasing, DLSS Frame Generation, Nvidia Reflex, and DLSS Ray Reconstruction. Here’s a brief overview of each DLSS version:
The first version introduced DLSS Super Resolution upscaling on the RTX 20 series. However, it relied on developers training the AI-derived upscaling algorithm individually for each game, comparing thousands of in-game frames to create a unique approach. This process was cumbersome, leading to slow implementation and often resulted in a blurry image.
DLSS 2.0 improved upon Super Resolution by implementing a generic upscaling algorithm already trained on hundreds of games. This enhancement simplified the implementation process for developers and significantly improved image quality.
DLSS 3 introduced advanced Super Resolution technology from DLSS 2.0 and added Frame Generation, which operates independently of Super Resolution. This means games can incorporate AI Frame Generation without relying on DLSS 2.0 Super Resolution. However, since Frame Generation is exclusive to RTX 40 series cards, like the GeForce RTX 4090, its features may not be accessible to most RTX gamers. DLSS 3 also mandates Nvidia Reflex as part of its suite.
DLSS 3.5 encompasses all the features of previous versions and introduces Ray Reconstruction technology. This innovation enhances the image quality of ray-traced lighting and shadows with minimal impact on performance, and it is available for all GeForce RTX GPUs.
![]()
Drawbacks
DLSS does come with certain limitations. First is the possibility of CPU bottleneck, which can hinder the expected boost in framerates while adjusting resolution or graphical settings. In cases of CPU limitations, DLSS may not significantly increase framerate, since it achieves higher frames by lowering the actual resolution, resulting in an upscaled image without any frame improvement.
This issue has been resolved with DLSS 3, but it comes with a problem of its own where the AI struggles with duplicating UI elements like text and minimaps. While DLSS 1 and 2 handle this by upscaling 3D elements first and then applying the UI, DLSS 3 with frame generation uses a fully rendered frame which can cause issues with text clarity.
DLSS 3.5 has made some improvements in this regard, but occasional issues may still be noticeable. Moreover, frame generation adds significant latency, as it requires two rendered frames, with one coming after the AI-generated frame. This delays the delivery of the latest frame, resulting in a higher framerate but with unchanged latency. While the game may appear smooth, the response to button presses may not be as rapid as expected due to the increased latency.
Another drawback is game support. There are only a few hundred games that currently support DLSS, most of which are popular AAA titles that have launched in the past five years. While Nvidia is actively working with all major game developers, you may not be able to boost performance using DLSS in your favourite game.
The competition
While DLSS is a pioneering technology, it’s not the only player in the field and faces competition from AMD’s FidelityFX Super Resolution (FSR) and Intel’s Xe Super Sampling (XeSS). DLSS holds a notable advantage due to its unique ability to generate new frames, outperforming Intel XeSS and AMD FSR.
It excels in both quality and performance, but there are some challenges. Limited hardware support restricts DLSS to recent Nvidia graphics cards, with DLSS 3 requiring the latest Nvidia RTX 40-series hardware. As mentioned before, game support is also a concern, with hundreds of games supporting DLSS 2 and even fewer with DLSS 3 compatibility.
In contrast, AMD’s FSR offers broader compatibility, running on Nvidia or Intel graphics, albeit requiring developer support. With the latest FSR 3.0 update, AMD has also introduced its own frame generation technique to multiply your performance. Similarly, Intel’s XeSS is compatible with AMD and Nvidia hardware, emphasising Intel hardware for optimal performance, yet facing limitations in the number of supported games.
DLSS-supported GPUs and games
To use DLSS, you must have an Nvidia GeForce RTX 20-series or newer graphics card, as DLSS relies on the presence of specialised AI processing units (Tensor Cores) found exclusively on RTX cards. Nvidia’s GTX series is not compatible due to the absence of Tensor Cores.
DLSS support is specific to individual games, and its availability depends on the titles you play. Fortunately, the roster of DLSS-supported games is continually expanding. Nvidia offers an up-to-date list of compatible titles, outlining the supported versions and features for reference.
How to enable DLSS
Since DLSS is exclusive to specific games, it can only be enabled within a game’s settings. To activate DLSS, launch the desired game and navigate to the settings menu. Head to the in-game display or graphics settings, where you’ll find the option to toggle DLSS on or off.
Furthermore, you can choose the level of DLSS deployment, including Performance, Balanced, Quality, or Ultra Quality (applicable to DLSS 2.1 or later). Most games allow you to toggle DLSS without requiring a game restart. Additionally, depending on the DLSS version your game employs, you may encounter extra settings related to ray reconstruction, DLAA, and sharpness.
