NVIDIA DLSS 5 is shaping up to be one of the biggest changes to AI-powered graphics in modern PC gaming. Instead of focusing only on higher frame rates or cleaner upscaling, this new technology introduces Neural Rendering, a real-time AI system designed to improve how lighting, skin, hair, fabrics, materials, and other visual details appear on screen.
The goal is simple but ambitious: make games look more photorealistic while still respecting the original art direction chosen by developers. NVIDIA describes DLSS 5 as a major step forward for real-time graphics, giving game creators new tools to enhance scenes without completely replacing the work of artists.
Although DLSS 5 has not officially launched in games yet, PC modders have already found a way to test an early version. An experimental Neural Rendering runtime appeared inside the early access PC build of NBA 2K27. Soon after, members of the RenoDX community managed to inject the feature into other games using RenoDX and ReShade, with Control becoming one of the first major test cases.
Since then, the unofficial DLSS 5 mod has been tried in several modern PC games, giving enthusiasts an early look at what NVIDIA’s next-generation Neural Rendering could bring to future titles.
DLSS 5 is different from previous DLSS versions because it is not only about reconstructing frames or improving performance. Instead, it analyzes the game’s existing image data and enhances the way surfaces and lighting behave. NVIDIA says the system uses color information and motion vectors from each frame, then feeds that data into a real-time neural model.
That AI model can recognize visual elements such as human skin, hair, clothing, translucent objects, reflective materials, and complex lighting conditions. From there, it can improve details like subsurface scattering, material shine, fabric response, and how light interacts with characters or environments.
The result can be a more realistic and grounded image, especially in close-up character shots. Faces may appear more detailed, hair can gain extra depth, and clothing may react more naturally to light. In some scenes, the difference can be subtle. In others, the effect can make a game look dramatically more cinematic.
However, DLSS 5 is not meant to be a one-size-fits-all visual filter. NVIDIA has emphasized that developers can control how strong the effect is, where it appears, how it affects color grading, and which parts of the image should be masked or protected. That level of control is important because aggressive Neural Rendering settings can change the entire mood of a scene. If pushed too far, it may make faces look unnatural, alter lighting in unwanted ways, or shift a game away from its intended style.
The unofficial DLSS 5 tests began when modders discovered the file “nvngx_dlssnr.dll,” which appears to be the Neural Rendering runtime. The RenoDX community quickly experimented with it and managed to get it working through ReShade in Control. From there, the implementation expanded into other DirectX 12 games that already support DLSS.
At first, the leaked runtime only worked on RTX 50 Series graphics cards because the included CUDA binaries were built for Blackwell GPUs. Later, a community modder managed to replace those binaries with versions compatible with Ada Lovelace hardware, allowing the experimental feature to run on RTX 40 Series GPUs as well.
That does not mean official RTX 40 Series support is confirmed. It only means the community has found a way to make this early and unofficial version work on current-generation cards. Official support will depend on NVIDIA’s final implementation and future driver or SDK updates.
RenoDX plays a major role in these experiments. It is a ReShade-based modding framework known for improving HDR, tone mapping, and color presentation in PC games. In many cases, RenoDX can deliver better HDR results than a game’s default implementation. With DLSS 5 Neural Rendering now being tested through the same ecosystem, it has become one of the most active tools for early visual experimentation.
Trying the unofficial DLSS 5 mod is not recommended for casual players. It is experimental, unstable, and not officially supported by NVIDIA or game developers. Anyone who still wants to test it should use a compatible DirectX 12 game with DLSS support and should always back up any files before replacing or modifying them.
ReShade with Full Add-On Support is usually required, and the RenoDX DLSS 5 add-on must be placed in the correct game folder. The Neural Rendering runtime file is also needed, along with any other required NVIDIA Streamline files depending on the game. Some implementations may require newer DLSS Super Resolution or Ray Reconstruction files as well.
RTX 50 Series users need the Blackwell-compatible runtime and a recent NVIDIA driver. RTX 40 Series users need the community-modified Ada-compatible version. Once everything is installed, the game can be launched, the ReShade menu can be opened, and the RenoDX Neural Rendering options can be enabled.
Users should avoid pushing every setting to the maximum. Conservative values are recommended because the effect can become too strong very quickly. In many test videos, the best results come from careful tuning rather than simply enabling the most aggressive preset.
It is also strongly recommended to avoid using these types of injection mods in multiplayer games or any game protected by anti-cheat software. Experimental graphics modifications can cause crashes, compatibility issues, false positives, or other problems.
Early community comparisons show that DLSS 5 Neural Rendering can have a major impact, especially in Control. Since Control was one of the first games used for testing, it has become the main showcase for the mod. The effect on Jesse, the game’s main character, can be striking. Skin detail appears richer, facial lighting gains more depth, hair can look more defined, and clothing materials respond more noticeably to light.
In some scenes, characters take on an almost pre-rendered quality, with more lifelike shading and stronger material definition. Darker environments can gain a more cinematic appearance, while close-up shots can look more detailed than the original presentation.
Still, the results are not always perfect. Because this is not a proper developer-made DLSS 5 implementation, some scenes can look overprocessed. Faces may appear too sharpened, lighting can become inconsistent, and color balance may shift in ways that do not match the original artistic intent. That is why the official version of DLSS 5 will likely depend heavily on developer tuning.
The most important thing to remember is that these early DLSS 5 tests are not representative of final game support. The RenoDX community has done impressive work, but they are using an unofficial runtime without access to the full developer toolset. A real DLSS 5 integration would allow game studios to fine-tune intensity, masking, color, lighting response, and scene-specific behavior.
Even with those limitations, the early results are exciting. DLSS 5 suggests a future where AI graphics technology does more than upscale an image. It could help improve the quality of lighting, character rendering, and material realism in real time, potentially making games look closer to offline-rendered cinematics while still running interactively.
For PC gamers, this could be one of the most interesting visual upgrades since ray tracing and DLSS Super Resolution first arrived. For developers, it could become another powerful tool for improving visual fidelity without manually rebuilding every asset or lighting setup.
At the same time, DLSS 5 will need careful handling. Photorealism is not always the goal of every game, and many titles rely on stylized visuals, exaggerated lighting, or specific color palettes. If Neural Rendering is applied too broadly, it could harm the look of those games rather than improve it. The best implementations will likely be the ones where developers use DLSS 5 selectively and thoughtfully.
For now, DLSS 5 remains an unofficial playground for enthusiasts rather than a finished consumer feature. The modded RenoDX version is fascinating, but it comes with expected issues such as crashes, visual glitches, heavy performance costs, VRAM problems, and broken HDR behavior in some games.
Even so, these first experiments give us a glimpse of where NVIDIA’s AI graphics technology may be heading. If the final version delivers on its promise, DLSS 5 could become much more than another performance feature. It could mark the beginning of a new era of neural real-time rendering, where AI enhances not just resolution and frame rates, but the very way games look and feel.DLSS 5 Neural Rendering tested across major PC games: stunning visual upgrades, but a steep FPS cost
Nvidia’s experimental DLSS 5 Neural Rendering is starting to appear in early game comparisons, and the results are already sparking plenty of discussion among PC gamers. Across titles like S.T.A.L.K.E.R. 2: Heart of Chornobyl, Cyberpunk 2077, The Witcher 3: Wild Hunt, Grand Theft Auto V Enhanced, Stellar Blade, Final Fantasy VII Rebirth, and Assassin’s Creed Shadows, the technology shows a clear pattern: it can make games look dramatically more detailed, more realistic, and more cinematic, but it can also change the original art direction in ways that not every player will enjoy.
The biggest takeaway from these early DLSS 5 tests is simple. Neural Rendering has enormous potential, but it needs careful developer tuning. When applied globally, it can improve faces, skin, hair, materials, lighting, and surface detail. However, it can also over-sharpen images, alter character appearances, exaggerate contrast, create haloing around objects, or make stylized games look less faithful to their intended visual identity.
The other major issue is performance. In several current comparisons, enabling DLSS 5 causes frame rates to drop by more than 50% on high-end graphics cards such as the RTX 5080 and RTX 5090. That makes it clear that this is not yet a polished, optimized version of the technology for everyday gaming.
S.T.A.L.K.E.R. 2: Heart of Chornobyl with DLSS 5
S.T.A.L.K.E.R. 2: Heart of Chornobyl is one of the most interesting DLSS 5 test cases because the game already pushes modern visuals hard with dense environments, complex materials, and advanced lighting powered by Unreal Engine 5.
With DLSS 5 enabled, the game’s presentation changes in several noticeable ways. Surfaces can appear more naturally connected to the lighting around them, material response becomes stronger, and facial details receive a more realistic Neural Rendering pass. In certain scenes, the added depth and lighting treatment can make the world feel more grounded and believable.
However, the effect is not always subtle. Some facial structures become too pronounced, and local contrast can look overly aggressive. This gives the impression that DLSS 5 is not simply enhancing the image but partially reinterpreting it. That may be exciting from a technology standpoint, but it also shows why official developer control will be essential.
The performance hit is also severe. On an RTX 5080, early testing shows that enabling DLSS 5 can cut performance by more than half. For a demanding game like S.T.A.L.K.E.R. 2, that is a major trade-off, especially for players targeting high frame rates at 1440p or 4K.
Cyberpunk 2077 with DLSS 5
Cyberpunk 2077 is already one of the best showcases for modern PC graphics, especially with path tracing enabled. Because of that, it provides a fascinating look at what DLSS 5 can and cannot improve.
DLSS 5 does not replace Cyberpunk 2077’s existing lighting system. Instead, it works on top of the rendered image and changes the way certain elements appear, including skin, hair, materials, shadows, and reflections. In the best examples, faces look more lifelike, clothing gains more texture, and some highly detailed surfaces take on a near-photographic quality.
At the same time, Night City is a carefully designed world with a very specific visual mood. Extra contrast or altered illumination can sometimes work against the atmosphere that the developers originally created. Cyberpunk 2077 is not just about realism; it is about style, color, mood, and cinematic lighting. If DLSS 5 pushes too far, it risks making some scenes look technically sharper but artistically less consistent.
This is where selective implementation could make a huge difference. If artists could decide exactly where Neural Rendering should be applied, DLSS 5 could enhance specific materials, characters, or lighting conditions without disrupting the game’s original look.
Performance remains a major concern here as well. Early RTX 5080 testing shows a frame rate drop of more than 50% after turning on the experimental DLSS 5 model.
The Witcher 3: Wild Hunt with DLSS 5
The Witcher 3: Wild Hunt is one of the most revealing examples because it shows how Neural Rendering can transform an older game without changing its actual geometry.
With the next-gen update already improving lighting, textures, and effects, DLSS 5 adds another layer of visual enhancement. Character faces show more surface detail, clothing reacts differently to light, and certain scenes appear more photographic than before. In some moments, it almost feels like a new visual upgrade has been applied on top of the existing next-gen version.
But The Witcher 3 also highlights one of the biggest challenges for AI-based rendering improvements. The game has a distinct fantasy art style. Its characters, environments, and lighting were not designed to look fully photorealistic. Making every surface more physically realistic does not automatically make the game look better.
For example, more detailed skin or stronger material response might look impressive in a screenshot, but it can also reduce the painterly, atmospheric quality that makes The Witcher 3 so recognizable. An ideal DLSS 5 implementation would need to improve image quality while preserving the game’s identity.
No reliable performance data was provided for this comparison, but based on the results seen in other games, the performance cost is likely to be significant.
Grand Theft Auto V Enhanced with DLSS 5
Grand Theft Auto V Enhanced may be one of the most dramatic DLSS 5 examples so far. Since the original game is more than a decade old, its character models, lighting foundation, and asset design naturally show their age compared to newer open-world titles.
DLSS 5 can make GTA V Enhanced look surprisingly modern in the right scenes. Skin shading becomes more detailed, faces gain stronger definition, materials react more convincingly to light, and outdoor environments can take on a much more realistic appearance. When the effect works well, the result can be genuinely impressive.
However, GTA V also shows how Neural Rendering can go too far. Some characters begin to look less like upgraded versions of Rockstar’s original designs and more like AI reinterpretations. That can be distracting, especially in a game where character personality and recognizable facial design are important.
Another issue seen in some comparisons is haloing around player characters. This creates a bright outline near character silhouettes, which can make the image look artificial and reduce immersion.
There is no confirmed performance data from these GTA V Enhanced tests, but because the game is generally easier to run than many modern AAA titles, the DLSS 5 performance hit may be more manageable on newer RTX 40 Series and RTX 50 Series graphics cards. Even so, the visual side still needs careful tuning.
Stellar Blade with DLSS 5
Stellar Blade is another important case because it features highly stylized character designs. Early screenshots show that DLSS 5 can add visible detail to faces, hair, skin, clothing, and nearby lighting. In some scenes, the extra detail looks convincing and gives the image a sharper, more premium appearance.
However, Stellar Blade also makes it clear that more detail is not always the same as better art direction. The game’s characters are designed with a specific stylized look, and DLSS 5 can sometimes push them closer to realism in a way that may not fit the original presentation. Players who prefer the intended clean, stylized aesthetic may find the Neural Rendering effect too aggressive.
Performance data from one RTX 5090 and Ryzen 7 9800X3D test gives a useful look at the current cost. In some areas, frame rates reportedly drop from around 155-250 FPS without DLSS 5 to roughly 118-125 FPS with DLSS 5 enabled. In one indoor scene, performance reportedly falls from around 270 FPS to about 132 FPS.
That is still playable on such powerful hardware, but the drop is large. The tester also reported behavior resembling a memory leak when repeatedly toggling the feature, which further reinforces that this experimental implementation should not be treated as final.
Final Fantasy VII Rebirth with DLSS 5
Final Fantasy VII Rebirth may be one of the best examples of why DLSS 5 needs different tuning profiles. The game’s visual identity blends high-end rendering with stylized character design, so an overly aggressive Neural Rendering pass can easily change the look too much.
In early comparisons, different DLSS 5 settings produce very different results. More aggressive presets can create an overly sharpened image with visible haloing and stronger contrast. While this may make screenshots look more dramatic, it can also make the game feel less natural.
A more restrained setting appears to work better. It keeps more of the original game’s presentation while still adding improved material detail and lighting depth. This suggests that the best DLSS 5 implementation may not be the one that creates the most obvious before-and-after difference. Instead, the best version may be the one that improves the image so naturally that it feels like it was always part of the game.
That lesson applies to almost every title tested so far. DLSS 5 should not simply make games look different. It should make them look better while respecting the original artistic direction.
Assassin’s Creed Shadows with DLSS 5
Assassin’s Creed Shadows is another title where DLSS 5 could become especially interesting because of its detailed environments, natural lighting, character materials, and historical setting. Early on/off comparisons suggest that Neural Rendering can affect surface detail, lighting behavior, and character presentation.
For a game like Assassin’s Creed Shadows, the value of DLSS 5 would likely depend on subtlety. Enhancing fabrics, armor, skin, foliage, and environmental materials could make the world feel richer and more realistic. But if the effect becomes too strong, it could interfere with the intended tone, color grading, and cinematic atmosphere.
As with the other examples, proper developer tuning would be the difference between a meaningful visual upgrade and an image that simply looks processed.
DLSS 5 looks powerful, but it is not ready to be judged as a finished feature
These early DLSS 5 comparisons are exciting because they show what Neural Rendering may bring to future PC games. The technology can add detail that goes beyond traditional upscaling. It can reinterpret lighting, improve perceived material quality, enhance facial detail, and make older games look far more modern.
But the current results also show the risks. DLSS 5 can change character faces, exaggerate image sharpness, alter lighting, create haloing, and reduce the impact of carefully crafted art direction. In games with stylized visuals, the danger is even greater because realism is not always the goal.
The performance cost is currently the biggest practical problem. Losing more than half of the frame rate on an RTX 5080 or RTX 5090 is a serious compromise, especially when many players buy high-end GPUs specifically for smooth 4K gaming, path tracing, and high refresh rates.
Still, it is important to remember that these are early experimental tests, not polished official implementations. A properly optimized DLSS 5 runtime, combined with developer-controlled masks and selective application, could deliver much better results. If developers can choose where and how Neural Rendering is used, DLSS 5 could become one of the most important visual technologies in PC gaming.
For now, DLSS 5 is best viewed as a preview of what may be possible. It can be stunning, strange, beautiful, and flawed all at once. The technology is clearly powerful, but its future success will depend on optimization, artistic control, and careful tuning for each individual game.DLSS 5 Neural Rendering looks impressive, but unofficial tests show why tuning matters
DLSS 5 is already becoming one of the most talked-about graphics technologies among PC gamers, especially because early community experiments suggest it can dramatically change how lighting, skin, materials, and character detail appear in modern games. While the results can be stunning, they also raise an important question: when does “more detailed” stop being better?
Recent unofficial DLSS 5 comparisons using an experimental RenoDX implementation give us a closer look at how NVIDIA’s Neural Rendering approach may affect major games such as Assassin’s Creed Shadows, The Elder Scrolls IV: Oblivion Remastered, and The Last of Us Part II Remastered. These tests are not based on final, developer-approved versions, so they should be treated as a preview rather than a final judgment. Still, they offer a fascinating glimpse at both the promise and the problems of AI-assisted rendering.
Assassin’s Creed Shadows is one of the most interesting examples because it was previously shown as part of NVIDIA’s official DLSS 5 showcase lineup. That makes it a useful comparison point for seeing how the technology behaves when applied by the community through unofficial means.
In the comparison, the difference between DLSS 5 off and DLSS 5 on is immediately visible. Naoe, one of the game’s two main playable characters, appears with more defined skin texture, altered facial shading, and a noticeably different interaction with the surrounding light. Shadows appear softer in some areas, while material definition across the scene becomes more pronounced.
At first glance, the added detail can look more realistic. Skin structure appears sharper, surfaces respond to light in a more complex way, and the overall image can feel more physically grounded. However, the reaction from viewers has been split. Some players like the extra realism, while others feel the Neural Rendering version changes the character’s appearance too much. In certain moments, the image can appear flatter, brighter, or slightly washed out compared to the original presentation.
That reaction gets to the heart of the issue. DLSS 5 is not simply a “better graphics” switch. If applied too strongly or without careful control, it can interfere with the artistic direction of a game. Ubisoft’s artists designed Naoe and the world of Assassin’s Creed Shadows with a particular visual tone in mind. Even if AI rendering adds technically impressive skin shading or material detail, that does not automatically mean it matches the intended look.
This is why developer tuning will be essential. Neural Rendering may need careful adjustment for specific characters, lighting conditions, weather, cutscenes, and gameplay environments. Without that tuning, the results can vary from impressive to distracting.
The Elder Scrolls IV: Oblivion Remastered is another important case because Bethesda has already confirmed that the game is expected to receive official DLSS 5 support. That makes community testing especially intriguing, even though the available screenshots are more limited.
Unlike the Assassin’s Creed Shadows comparison, the Oblivion Remastered screenshots do not include direct DLSS 5 on versus off images. As a result, it is harder to judge exactly how much the experimental Neural Rendering changes the original image. Still, the screenshots show that the current implementation can produce rich material detail and strong surface definition.
Armor, stone, skin, and environmental materials can appear more textured and responsive to light. In some scenes, the added rendering complexity gives the world a more modern visual quality, which could be especially appealing for a remastered version of a classic RPG.
However, the biggest concern appears to be lighting balance. The user who shared the images noted that nighttime scenes can become too bright. That is a major issue for a game like Oblivion Remastered, where atmosphere plays a huge role in exploration. Nighttime should feel dark, mysterious, and sometimes dangerous. If Neural Rendering lifts the brightness too much, it can weaken the mood and make the world feel less natural.
This is another clear example of why DLSS 5 cannot be judged only by screenshots that show extra detail. A great image is not only about sharp textures or realistic materials. It is also about contrast, tone, color grading, and the emotional feel of a scene. If an AI rendering system improves one part of the image while damaging the overall atmosphere, the final result may not be an upgrade.
Official support should allow developers to tune DLSS 5 for different scenarios. Bright outdoor areas, candlelit interiors, foggy forests, dungeons, and moonlit roads may all require different levels of intensity and masking. Without that level of control, the experience can become inconsistent.
The Last of Us Part II Remastered may be the toughest test for DLSS 5 Neural Rendering so far. Naughty Dog’s game is already known for extremely detailed faces, carefully crafted character materials, cinematic lighting, and highly controlled art direction. Because the original presentation is so polished, any AI-driven change has to be handled very carefully.
In the community comparison, enabling DLSS 5 adds noticeable structure to skin and lighting. Facial surfaces can appear more detailed, with stronger definition in pores, contours, and subtle shading. In some ways, the technology shows exactly why Neural Rendering is exciting: it can enhance areas of the image that are already complex and make them look even more layered.
But that same strength can become a weakness. When applied aggressively, the effect can make a familiar character look different from the intended design. In a game as cinematic as The Last of Us Part II Remastered, faces are not just visual assets. They are central to storytelling, emotion, and performance. Even a small shift in skin tone, facial softness, shadow placement, or material response can change how a character reads on screen.
This is where the “uncanny valley” concern becomes important. More realistic skin detail can sometimes clash with animation, expression, or movement. If the rendering becomes hyper-detailed while the performance remains authored for a different visual style, the character may look less natural rather than more lifelike.
That does not mean DLSS 5 is a bad fit for cinematic games. It means the opposite: games with high-end presentation may benefit greatly from Neural Rendering, but only when the effect is guided carefully by artists. The goal should not be to maximize detail everywhere. The goal should be to preserve the creative intent while improving the image in ways that feel natural.
Across these early experiments, one theme appears again and again: DLSS 5 has enormous potential, but it is not a magic button. Neural Rendering can enhance lighting, improve material response, and reveal finer surface detail. At the same time, it can alter faces, brighten scenes too much, flatten contrast, or change the mood of a game if it is not properly tuned.
NVIDIA has emphasized artist control as part of the technology, including options such as intensity adjustment, masking, and color-grading controls. Those tools may be crucial for making DLSS 5 work in real games. Developers will need the ability to decide where the effect should be strong, where it should be subtle, and where it should not be used at all.
Performance is another major concern. The unofficial implementation being tested by the community appears to be extremely demanding. In some reports, even very powerful GPUs lose a large portion of their frame rate when the experimental DLSS 5 effect is enabled.
That raises important questions about how Neural Rendering will fit alongside ray tracing and path tracing. DLSS 5 may improve the appearance of lighting and materials, but it does not fully replace physically based lighting techniques. Ray tracing and path tracing still provide grounded light behavior that AI enhancement alone cannot completely reproduce.
Combining Neural Rendering with heavy ray tracing or path tracing could therefore become very expensive from a performance perspective. In one community Cyberpunk 2077 test, performance reportedly dropped from 71 FPS with path tracing to 45 FPS after adding the unofficial DLSS 5 implementation. Since this is based on unreleased and unofficial technology, those numbers should not be treated as final. Still, they suggest that fully realizing DLSS 5 alongside the most demanding lighting techniques may require future graphics cards with far more processing power.
For now, these DLSS 5 comparisons are best viewed as an exciting technical preview. They show a future where AI-assisted rendering could make games look richer, more detailed, and more realistic. But they also show why artistic control matters just as much as raw technology.
If developers tune DLSS 5 carefully, it could become a major step forward for PC gaming visuals. If it is pushed too far, it could make beautifully crafted games look inconsistent or unintentionally altered. The technology is clearly powerful, but its success will depend on how well it respects the art behind the games it enhances.
