The presentation of DLSS-5 During GTC 2026, NVIDIA has once again placed itself at the center of the debate about the future of PC graphics. The company presents this version not as just another iteration of its well-known upscaling system, but as a step towards... real-time neural rendering, with the aim of reducing the gap between what we see in movies and what video games offer today.
Although many questions remain unanswered—particularly regarding resource consumption, compatibility, and how it will affect art direction—what has been shown so far makes it clear that DLSS 5 is aiming to be a significant change of focus in the way the final image is generated. The first examples in titles such as Starfield o Resident Evil Requiem They serve more as a statement of intent than as a finished product ready for the average PC gamer in Spain or Europe.
What exactly is DLSS 5 and how does it differ from previous versions?
Until now, the acronym DLSS was primarily associated with gain performanceAI-powered upscaling, pixel reconstruction, and additional frame generation to boost FPS or play in 4K without putting as much strain on the GPU. With DLSS 5, NVIDIA shifts gears and focuses on the visual quality and cinematic finish of each scene.
The company describes DLSS 5 as a model of real-time neural rendering It's capable of adding photorealistic lighting and materials to every pixel without altering the game's original geometry or textures. Instead of simply reconstructing the image, the AI ​​interprets the scene, understands what's on screen, and decides how the light should behave and how the materials should respond.
This leap means that DLSS 5 is no longer "just" a supersampling system. The technology becomes an extra layer that sits between the graphics engine and the image the player seesreinterpreting the final result with a visual treatment that is more reminiscent of a Film color grading than a simple graphic filter.
In the words of Jensen Huang, CEO of NVIDIA, DLSS 5 is “the GPT moment of the graphs": a system that blends classic rendering with generative AI techniques to deliver a level of detail and realism that, according to the company, was not feasible with the traditional brute-force approach."
How DLSS 5 Neural Rendering Works
In practice, DLSS 5 takes as input the data generated by the game itself In each frame: color information, motion vectors, depth, and other internal engine buffers. From there, the AI ​​model analyzes each frame individually, while maintaining consistency with previous frames to avoid flickering or artifacts.
NVIDIA claims the model is end-to-end trained to understand the semantics of the sceneThat is, it not only detects shapes and edges, but also identifies specific elements: characters, skin, hair, fabrics, metallic materials, water or translucent surfaces, as well as varied lighting conditions (backlighting, diffuse lighting, cloudy skies, etc.).
Based on that interpretation, the AI ​​generates pixels with complex physical properties These techniques are typically reserved for pre-rendered film effects: subsurface scattering on skin, soft highlights on fabrics, more natural reflections on metals, and richer interactions between light and hair. All of this is applied while striving to maintain the original structure and style defined by the studio.
One important aspect, especially for the demanding PC gamer, is that DLSS 5 is designed to work in real time up to 4KThis means that all this processing must be done in about 16 ms per frame if a 60 FPS experience is to be maintained. To approach this goal, the model runs on the specialized cores of RTX GPUs and relies on years of prior work with DLSS 2, 3, and 4.5.
Creative control: what developers can adjust
One of the points that most concerns studios and players is to what extent DLSS 5 can distort the artistic intention of a game. In the first comparisons, a very marked change in lighting and materials can be seen, with more defined faces, more "perfect" skin and scenes that, in some cases, lose part of the original character in exchange for a more homogeneous and "advertising" look.
To mitigate that risk, NVIDIA insists that DLSS 5 is not an automatic effect that is simply activated, but a tool that studios must integrate and configure from the graphics engineThrough the NVIDIA Streamline framework, developers have parameters to adjust the intensity of the effect, the color grading, the specific areas where it is applied, and the order in the post-processing chain.
In practice, this means that a studio can decide, for example, that the AI ​​only interacts with certain materials (skin, fabrics, metals) but not with certain elements of the environment, or that the intensity is reduced in gameplay scenes and increased in cutscenes. The key will be how each team interprets these adjustments and the time they dedicate to fine-tuning the result.
From a technical standpoint, this integration implies additional work for the studios, something that NVIDIA has not detailed in depth: there has been no talk of production costs, implementation times nor the effort involved in adapting already released games versus titles developed from scratch with DLSS 5 in mind.
Limitations, doubts and controversies surrounding DLSS 5
Although the official discourse focuses on the visual advancement, the presentation of DLSS 5 has also generated reviews and reservations within part of the community and among some professionals. One of the most discussed points is that AI is not limited to polishing the image, but goes further to noticeably change the appearance of characters and settings.
In demos like the one featuring the protagonist Grace, you can see how the system enhances facial features, softens or modifies dark circles, and alters skin texture to give it a look more typical of an advertising campaign than an everyday person. In games with less cutting-edge graphics, such as some open-world titles, the contrast between the original material and the processed version can be quite jarring.
This transformative nature opens a debate about where is the limit? between improving image quality and rewriting some of the artwork. Some gamers fear that, if the effect is overused, many games will end up sharing a very similar look, losing aesthetic variety in favor of an "NVIDIA standard."
Furthermore, there is currently a lot of opacity about the real impact on performanceIn the initial demonstration, NVIDIA acknowledged that the scene was running with two GeForce RTX 5090 graphics cardsOne GPU is dedicated to the game, and the other to DLSS 5, with the intention of optimizing it to run on a single consumer GPU. It remains to be seen what will happen when the technology reaches commercial titles and mid-range PCs, which are common in the European market.
Hardware compatibility and technical requirements
Another big unknown at the moment is the final list of GPUs compatible with DLSS 5Officially, NVIDIA has only indicated that the technology will be available starting in the fall and will initially arrive in the family GeForce RTX 50without clarifying whether previous models — such as the RTX 40 or RTX 30, very common in PCs in Spain and the rest of Europe — will receive full, partial or no support.
The demo itself, with two RTX 5090s, and the constant references to the new cores dedicated to neural shading and path tracing suggest that the focus is clearly on the new generationHowever, the company has not yet published an official compatibility chart or minimum requirements.
Nor have any specific figures been detailed regarding VRAM consumption, CPU usage, or RAM memory requirementsNVIDIA suggests that the bulk of the work will fall on the GPU's AI cores, but until there are independent benchmarks, it will be difficult to know what sacrifices have to be made in other areas to maintain a stable FPS rate with DLSS 5 active.
For PC users considering a graphics card upgrade in the coming months, this positions DLSS 5 as a strong argument in favor of the upcoming RTX 50 series, although right now There is no visibility into pricing, availability, or behavior. of these models in the European market, where the performance/price ratio is a decisive factor.
Industry adoption: engines, studios, and games confirmed
Beyond the technical aspects, one of the areas where NVIDIA seems to have an advantage is in the support from major publishers and studiosThe company has confirmed that DLSS 5 is integrated through the framework NVIDIA Streamline, the same one already used for previous versions of DLSS and for Reflex, which facilitates its incorporation into engines like Unreal Engine and in our own developments.
Among the companies that have already announced a collaboration are names very familiar to European players: Bethesda, CAPCOM, Ubisoft, NetEase, Tencent, NCSOFT, Hotta Studio, S-GAME y Warner Bros. GamesAmong others. It's not just theoretical support; several of their upcoming releases and already available games will receive support.
The list of titles that will incorporate DLSS 5 in the coming months is extensive and combines new games with remasters of classics. Among the notable names are Starfield, Resident Evil Requiem, Hogwarts Legacy, Assassin's Creed Shadows, Delta Force y The Elder Scrolls IV: Oblivion Remastered, all of them with great appeal among the PC community in Spain.
Along with them, NVIDIA also mentions AION 2, Black State, CINDER CITY, Justice, Naraka: Bladepoint, NTE: Neverness to Everness, Phantom Blade Zero, Sea of ​​Remnants and Where Winds MeetIt is expected that this list will grow as the launch approaches and that many companies will choose to incorporate DLSS 5 via an update in titles already established in the market.
Release schedule and current development status
Regarding the schedule, NVIDIA places the arrival of DLSS 5 in autumnwithout a firm date yet. The timeframe allows studios time to adjust their implementation and for the company itself to finalize the model after internal testing and the first public demos.
It's worth remembering that DLSS 5 was announced just a few months after DLSS 4.5 and Multi Frame Generation 6xThese technologies are still in the adoption process. This overlap of versions can be somewhat confusing for the user, who sees different branches of DLSS with different functions (upscaling, frame generation, neural rendering) coexisting under the same brand.
NVIDIA defends this approach by arguing that DLSS has become an umbrella Under which it groups several AI techniques applied to rendering, and the name is maintained because gamers already identify it with graphical improvements in GeForce RTX since 2018. In return, the risk is that the label loses clarity about exactly what each version offers.
Meanwhile, the company insists that DLSS 5 is still in the development phase. pre-launch technologyWhat we've seen so far are carefully controlled demos, mainly in cinematic scenes where it's easier to showcase the effect. We still need to see how it performs in more mundane situations: open environments, intense combat, rapid camera changes, or competitive games where any visual inconsistency is instantly noticeable.
Based on everything known to date, DLSS 5 is shaping up to be an ambitious attempt to to bridge the gap between real-time graphics and the finish of a film productionThis shifts some of the rendering workload to artificial intelligence models trained over years. Many practical questions remain—the price of the necessary hardware, the impact on real-world performance, the degree of control studios will have, and how it will integrate into the daily lives of PC gamers—but the move clearly marks the direction NVIDIA wants to take in the coming years: an RTX ecosystem where AI not only helps to render more pixels, but also decides how they should look.