DLSS 5 Tested in Five Games: See Which Visual Upgrades Hold Up

DLSS 5 Tested in Five Games: See Which Visual Upgrades Hold Up

الحلقة التالية

DLSS 5 can make skin, cloth, foliage, and lighting look more detailed, but the early results come with a major catch. In the unofficial builds tested across five games, crashes, flickering, and strange resolution bugs often arrived before the visual gains had time to impress.

Nvidia describes DLSS 5 as a new rendering pipeline built around neural rendering. The version tested here was not a finished feature supported by developers. Modders applied leaked or unfinished files to Control, Watch Dogs, Grand Theft Auto IV, The Last of Us Part I, and The Witcher 3.

The tests also used high-end systems with GeForce RTX 5090 and RTX 5080 graphics cards. Even then, performance and stability remained poor. The current modded version may bypass safeguards that help protect hardware, so casual players should wait for official game support.

DLSS 5 Promises a New Kind of Game Rendering

Neural rendering changes textures and lighting without adding geometry

DLSS 5 does not add polygons to a game’s models. A low-detail character or object stays low-detail, even if its lighting and surface texture look better.

The neural process can change how textures appear and how light reacts across a scene. It may improve eyes, skin, shadows, cloth, foliage, bricks, and other surfaces. In the best cases, it strengthens the game’s existing art style instead of replacing it.

That difference matters. Better lighting can make a familiar character look more natural, but it cannot turn a low-poly model into a modern one. The effect works best when the original assets already give it enough information to work with.

Unofficial testing reveals crashes and image bugs

The early files were not designed for every game tested. They may also have been several months old, while modders removed limits and restrictions to make them run across different titles.

Control appeared to be the most stable example, yet it often lasted only about four minutes before crashing or showing a serious visual problem. In one test, the DLSS 5 overlay became far lower resolution than the game image beneath it.

Similar crashes appeared in the other games. A normal playthrough was not practical. This does not prove that the final DLSS 5 release will behave the same way, but it shows why unofficial compatibility mods are poor substitutes for developer integration.

Control Shows DLSS 5’s Most Convincing Enhancement Potential

Characters gain better skin, eyes, and facial lighting

Control produced the strongest result in the five-game test. The game still looked like Control, but human features appeared more refined.

Skin and eyes gained extra depth, while small shadows around the nose and beneath hair looked more natural. These changes were not dramatic, yet they helped faces respond to light in a more believable way.

This restrained approach worked better than an aggressive visual overhaul. DLSS 5 improved parts of the image without bulldozing the game’s art direction.

Better surface response cannot fix weak models

The results also showed the limits of neural rendering. DLSS 5 can improve texture response, lighting, and shadows, but it does not rebuild missing geometry or increase polygon count.

That means the final result depends on the game’s source assets. A strong model may gain convincing light and texture detail. A simple model may only become a better-lit simple model.

Even with the best-looking test, stability was the main problem. RTX 5090 and RTX 5080 hardware did not prevent crashes or overlay failures, so the current build is not ready for regular play.

Watch Dogs Shows How DLSS 5 Can Hurt Image Quality

Daytime scenes looked blurry and distorted

Watch Dogs produced the weakest daytime footage. Aiden Pearce sometimes looked almost like a soft, melted image, while the wider scene lost the clean look of the original game.

A low-resolution overlay may have caused part of the problem, but the exact reason was unclear. Whatever the cause, the result undermined a game known for its ambitious visual presentation.

The characters still looked like their original designs, only with more awkward detail. Neural enhancement cannot fix every asset, and pushing more processing onto a flawed face may make the flaw easier to see.

Night scenes and cutscenes fared better

Nighttime scenes looked much better than daytime footage. Cutscenes also avoided some of the worst problems seen during open-world play.

Darker scenes may hide image errors, while bright lights and strong contrast can make environmental changes appear more pleasing. That does not confirm why the difference occurred, but it shows how much scene lighting affects DLSS 5’s results.

The test also pointed toward a useful solution: developers could apply different strength levels to different objects. Skin could receive a lighter effect, while cloth, buildings, weather, and backgrounds could receive stronger treatment.

Grand Theft Auto IV and The Last of Us Part I Reward Controlled Intensity

Grand Theft Auto IV benefits from environments and cloth

The Grand Theft Auto IV footage used RTX Remix as well as DLSS 5. RTX Remix supplied higher-resolution textures, new lighting features, and DLSS compatibility, so the visual gains cannot be credited to DLSS 5 alone.

The combined setup still showed strong results. Backgrounds, cloth, rigid objects, and environmental lighting appeared more detailed and natural. Inanimate surfaces often handled the neural effect better than faces.

Skin improved at lower strength settings. The test found an intensity around 0.3 more pleasing than the maximum setting of 2. At higher levels, faces developed the harsh texture associated with excessive sharpening and artificial AI detail.

The Last of Us Part I gains cinematic lighting

The Last of Us Part I benefited from its moody lighting. Many scenes use dark interiors, narrow windows, silhouettes, and hard shadows across faces. Those conditions gave DLSS 5 more chances to improve the image.

Some scenes looked grittier and more cinematic. Bricks, foliage, shadow transitions, and skin reacting to light all showed promising changes.

Outdoor areas often looked better overall, though some indoor scenes with strong light direction also worked well. Locations without a clear light source looked worse. The quality depended on the lighting setup, not only on whether the scene took place indoors or outside.

The Witcher 3 Shows Why Neural Effects Cannot Replace a Remaster

Flickering offsets isolated lighting gains

The Witcher 3 showed a few appealing lighting improvements, including better-looking light on skin. The overall image, however, was flickery and inconsistent.

Temporal stability matters during gameplay. Objects need to stay coherent as the camera moves, and lighting should not shimmer or change from frame to frame. In this test, the visual gains did not outweigh those problems.

The game also provided the least convincing reason to use DLSS 5. Its effect was too unstable to feel like a meaningful upgrade.

A remaster offers a broader improvement

DLSS 5 alone cannot provide a full asset overhaul. A traditional remaster can improve textures, models, lighting, compatibility, and performance together.

The test compared The Witcher 3 with an upcoming free remaster described as a more substantial upgrade. That version was expected to offer better support without leaked files, removed limits, or risky setup work.

The game shows why older titles may need custom integration. Nvidia and developers may have to tune DLSS 5 for each game instead of offering one universal switch.

DLSS 5 Needs Masking and Game-Specific Tuning

Per-object masking could protect faces and skin

The clearest solution from these tests is object-based masking. Developers could identify faces, skin, eyes, and hands, then apply a lighter neural effect to those areas.

Cloth, foliage, bricks, rigid surfaces, weather, and distant environments could receive stronger treatment. These materials often benefited more from extra detail and lighting than human skin did.

This approach could preserve character identity while improving the surrounding world. It would also reduce the overly processed face effect seen in several tests.

Consistency matters more than maximum strength

Turning DLSS 5 up to its highest setting does not guarantee the best image. It can create harsh skin, strange lighting, flicker, and other distractions.

Finished games should tune the effect around their original art style. A subtle improvement that stays stable during movement will look better than a dramatic effect that breaks after a few minutes.

Players should wait for official patches, supported drivers, and developer-made implementations. The version linked to NBA 2K27, cited as arriving on September 3, should be judged separately once the final game and integrated build are available.

Conclusion

The five-game test showed both the promise and the limits of DLSS 5. Control offered the most natural upgrade, while Watch Dogs often looked worse during the day. Grand Theft Auto IV showed strong environmental potential with RTX Remix, and The Last of Us Part I gained from moody lighting. The Witcher 3 improved in small areas but suffered from flicker and poor consistency.

The current unofficial version is unstable, even on RTX 5090 and RTX 5080 systems. Its best use appears to be selective and conservative processing, with lower strength on skin and higher strength on suitable environmental surfaces.

DLSS 5 may improve future games, but its success will depend on stable performance, careful masking, and developer-led tuning. For now, treat the technology as a preview of what neural rendering could become, not as a finished upgrade worth installing on your games.