How to try DLSS 5: games, GPUs and useful settings
Check how to try DLSS 5 in supported games, tell it apart from Super Resolution and Frame Generation, and compare images without random mods.

On this page
This guide reflects the situation as of 17 November 2026. Things are moving fast: before changing anything on your PC, check the current notes from the game and from NVIDIA, especially if you are following a video published months ago.
Check whether DLSS 5 is available and what your game needs #
In its official update of 13 October, NVIDIA lists DLSS 5 as already available in NBA 2K27, with a new model due on 16 October. The documented local path covers GeForce RTX 50 Series cards; the announcement also includes platforms such as GeForce NOW and RTX Spark, each with its own availability conditions.
Since 5 November DLSS 5 has also been available in Hogwarts Legacy: after updating the game, you turn it on with the Neural Rendering option in the graphics settings, on GeForce RTX 50 Series cards (desktop and laptop), RTX Spark systems and GeForce NOW Ultimate. NVIDIA has also announced support in War Thunder, with the Eagle’s Shadow update in mid-November, and in Where Winds Meet from 23 November: before trying them, check that the patch is actually out.
Don’t assume this support covers every RTX GPU or every game that shows the DLSS name. The list and the requirements must be checked for the specific feature.
A three-step check #
First read the game’s official announcement, then check the installed patch, and finally look for the option in the menu. Announced future support, a beta feature and a release that is already out are three different states. If you are reading this guide well after its review date, repeat the check on the official NVIDIA and game pages.
Also note the platform you play on. A cloud session can have different conditions and hardware from your local PC, and a manufacturer demo may use a setup that is not equivalent to yours. Seeing the technology’s name in a video title is not enough to verify the environment.

If the option doesn’t appear #
Don’t go looking for a hidden setting straight away. Read the requirements for your game and GPU combination, check for updates and look at the graphics modes the game provides. If a documented requirement is missing, an absent option may be the correct behavior, not a driver bug.
To avoid buying hardware on the strength of an incomplete promise, only count the support announced for the exact feature you want to use. Support for a different DLSS component does not automatically confirm support for neural rendering.
If you are thinking of changing graphics card for this feature, start from choosing your PC components: check power supply, case clearance and system compatibility as well as the announced support for the game.
Neural rendering, reconstruction and frame generation are different things #
| Feature | Main purpose |
|---|---|
| Super Resolution | Reconstruct an output image from a lower rendering resolution |
| Ray Reconstruction | Reconstruct ray-traced effect information in the supported pipeline |
| Frame Generation and Multi Frame Generation | Generate intermediate frames in the supported configurations |
| DLSS 5: 3D-Guided Neural Rendering | Change the final look, for example lighting and materials |
The version number does not mean every feature needs the same hardware. An updated Super Resolution model or a new frame generation mode does not prove that DLSS 5 is active.
NVIDIA Research describes a model conditioned on the game’s rendering and on information from the engine. It is not simply a sharpening filter applied on top of any image on your desktop.
Reconstructing a reference versus changing the look #
A reconstruction technique tries to produce an output image from information created by the renderer. The DLSS 5 approach described by NVIDIA Research instead brings in learned knowledge about appearance to change the final presentation, while staying tied to engine information and the game’s art direction.
That is why it is wrong to treat it as a universal upgrade of a single file. The integration has to supply the required information and allow the controls the game provides. Two games can expose different options and give different results without either menu being wrong.
Displayed FPS and responsiveness #
When Frame Generation is involved, the number of frames displayed and the pace of traditional rendering are not interchangeable. A useful comparison also looks at stability and how quickly the game responds to your input. A higher counter alone does not prove that every part of the experience has improved by the same amount.
So test each feature separately. If you turn on reconstruction, generation and neural rendering together, the result may be interesting, but you can’t attribute the change to any one technology.
Set up an official test without changing several things at once #
- Check your GPU model in the system or in the NVIDIA app.
- Update the drivers from the official channel and note the version installed.
- Update the game and read the notes for the patch that adds support.
- Open the graphics settings and look for the feature described in the game’s documentation.
- Create a repeatable scene, save, and compare the same shot.
If the option is missing, check the game, the patch, the GPU and the required graphics mode. Don’t conclude that a file found online will unlock it. On a cloud platform, also check the game’s availability, your plan and the session you were actually given.
Keep a reference configuration #
Before changing anything, save the game settings or take a photo of them without personal data. Pick a scene where you control lighting and camera, and note resolution, preset and active features. If the game changes settings on its own between sessions, check them again before comparing.
Do a first run with your usual configuration, then change only the feature you are testing. Keep the frame limit and other options identical where the game allows it. If a change requires a restart, repeat the same route after launching.
Don’t use a first run full of loading or background activity as your only reference against a second run that has already settled. Make sure both runs represent comparable conditions. If you publish the results, always state the spot you tested and the combination of versions, and avoid generalizing to every game.
Compare quality, performance and third-party changes #
How to make a useful comparison #
Keep resolution, preset, lighting and camera position constant. First compare the image with DLSS 5 on and off, then test the other features separately.
Look at faces, materials, vegetation, text and thin objects while moving. A single screenshot can hide temporal instability: also record a short sequence along a repeatable path.
For performance, look at smoothness and responsiveness as well as the frame counter. If you change Super Resolution, Frame Generation and neural rendering at the same time, you won’t know which change produced the result. Don’t compare FPS figures collected with different settings as if they measured that feature alone.

Watch stability in motion #
Check lighting transitions, character edges, thin objects and camera cuts. A still image can look convincing and still show differences in motion. If you record the comparison, use consistent capture settings and don’t apply different corrections to the two videos.
Also ask whether the result matches the look you want in that game. An image that feels more realistic is not automatically better for every kind of content or every player’s taste. Describe what you see rather than summing it up with a universal verdict.
For performance, avoid mixing averages collected in different scenes. A useful result connects smoothness, configuration and scene, and makes the limits of the test visible.
Why you should avoid swapping DLLs at random #
Overrides and third-party mods can change specific components, but they don’t automatically add the integration and engine information that each technology needs. A menu item that changes name is not proof enough.
In games protected by anti-cheat, modifying files can also break the game’s rules. Before any experiment, keep the original configuration and check the manufacturer’s documentation. The official path remains the reference for understanding what you are really comparing.
If you need to troubleshoot a driver problem, use a dedicated procedure such as the DDU guide, without making it a required step for turning on DLSS.
If you decide to study a third-party change, keep the experiment clearly separate from the supported path. Keep the original files, the version and the instructions to restore them; also check whether the game verifies file integrity or forbids modifications. Don’t credit the manufacturer with a result you got using components other than the official ones.
To return to the reference test, restore the files and check the options again. A configuration left partly modified can make every later comparison hard to read, even after a driver update.
Frequently asked questions #
Does DLSS 5 always increase FPS? #
It shouldn’t be treated as a free, universal performance boost. Measure the actual combination of features and take into account the cost of neural rendering as well.
Does a game with DLSS 4 automatically get DLSS 5? #
No. Support has to be confirmed in the official announcements and in the game’s patch. Belonging to the same family doesn’t make all integrations interchangeable.
How can I tell which DLSS feature is causing the change? #
Start from a documented configuration and change one feature at a time at the same point in the game. Also check for required restarts, and don’t credit neural rendering with a result you got while also changing resolution or frame generation.


