What is DLSS?

DLSS 3 in Cyberpunk 2077
(Image credit: CD Projekt RED)

It wasn’t that long ago when the only way to make your favorite PC games look better was to buy the very best graphics card on the market, but that’s changed in recent years, largely thanks to an Nvidia technology known as Deep Learning Super Sampling, or DLSS for short.

Utilizing advanced, hardware-accelerated machine learning, DLSS lets your PC render fewer pixels per update cycle, while still pushing out sharper frames, steadier ray-traced lighting, and multiplayer-ready frame rates that once demanded significantly more processing power. It’s an AI engine that gives you extra frames per second, practically for free.

But how does this technology actually work? Which features do you get on which cards? And is “DLSS Quality” always the right call?

I’ve spent years working with Nvidia’s DLSS, and I’m here to break down the different versions, separating myth and marketing from reality, and explain when to use this technology, .

An Nvidia GeForce RTX 4080 in a testbench in a testing lab

(Image credit: Future / John Loeffler)

What is DLSS?

DLSS is a suite of AI-powered graphics techniques that run on the dedicated Tensor Cores inside GeForce RTX GPUs.

Instead of brute-forcing every pixel of every frame at your monitor’s native resolution, DLSS lets the game render a lower-resolution base image and then uses a neural network trained on thousands of high-resolution examples to reconstruct a crisper image at higher resolution in real time.

The result is often picture quality that practically matches that of a native rendered frame, but at a fraction of the workload, allowing for more frames per second to be generated for smoother gameplay.

AI upscaling Nvidia DLSS

(Image credit: Nvidia)

What does the DLSS do?

DLSS has grown over the years into an umbrella term that covers several related technologies:

DLSS Super Resolution – AI upscaling that turns a lower-res frame into a higher-res one.

DLSS Ray Reconstruction – Does away with multiple ray-tracing denoisers for different games and uses a single transformer AI model to produce better lighting and shadows.

DLAA (Deep Learning Anti-Aliasing) – Uses the same neural network purely for high-quality edge smoothing at native resolution.

DLSS Frame Generation (introduced in DLSS 3) – Inserts an AI-predicted frame between two real ones, effectively doubling motion smoothness.

DLSS Multi Frame Generation (new in DLSS 4) – Goes even further, creating up to three AI frames for every 1 game-engine frame.

Taken together, these technologies boost framerates, stabilise ray-traced visuals, and sharpen detail, with NVIDIA Reflex keeping latency to a minimum.

Image of Nvidia's DLSS 4 Frame Generation

(Image credit: Nvidia)

Is it better to turn DLSS on or off?

For nearly all situations, the answer will be to turn DLSS on. DLSS leverages AI-specific hardware that isn’t used elsewhere during the frame rendering process, so turning it on isn’t taking resources from something else, and using it can bring serious gains in a game’s frame rate. That said, there are some caveats to that advice:

Frame saturation: If your graphics card already produces more frames per second than your monitor’s refresh rate at native resolution can support (say, 200 fps on a 144Hz monitor), turning DLSS won’t add much of a noticeable difference.

Image-quality tolerance: Nvidia’s new DLSS transformer models have effectively fixed the problems of earlier DLSS models that sometimes produced wonky visual artifacts or blurriness, but some users may still spot rare artefacts such as ghosting on fine wires or HUD elements.

Game support: While many of the latest PC games support DLSS, there are notable exceptions, like Starfield, and in my testing, some games work better with AMD’s FidelityFX Super Resolution. You’ll want to test both technologies (and possibly Intel XeSS), where available, to find the best results for your game and settings.

Mode choice: DLSS generally offers Quality, Balanced, Performance, and Ultra-Performance modes, with performance modes trading some visual quality for much higher FPS. My advice is to enable DLSS Quality mode first and evaluate your FPS. If you need more frames, move down to Balanced or Performance modes until you hit your target frame rate.

DLSS 4 framerate comparisons

(Image credit: Nvidia)

Does DLSS increase FPS?

Yes—sometimes dramatically. Even without Frame Generation, which can be somewhat controversial among some users, DLSS can make games playable at resolutions and settings higher than recommended for a given graphics card. The 8GB VRAM in the RTX 4060 largely precludes its being able to effectively game at 1440p, but with DLSS, you largely can, especially if you stay away from intensive settings like ray tracing.

NVIDIA DLSS

(Image credit: NVIDIA)

Is DLSS only for RTX 40?

No. Every generation of GeForce RTX—from 20-series Turing cards onward—can run DLSS Super Resolution and Ray Reconstruction, provided the game supports them.

What was 40-series-exclusive at launch is Frame Generation (DLSS 3); that feature relies on the Optical Flow Accelerator introduced with the Ada Lovelace architecture, and so Nvidia RTX 20 and RTX 30 series cards cannot use it.

The Nvidia RTX 5070 GPU being held

(Image credit: Future)

Can any graphics card use DLSS?

DLSS requires NVIDIA’s Tensor Cores, so it is limited to GeForce RTX GPUs (desktop, laptop, or cloud). Owners of AMD or Intel cards can use rival, open-standard alternatives—AMD FSR and Intel XeSS—but DLSS is strictly limited to Nvidia RTX cards.

DLSS 3.5 in action in Cyberpunk 2077: Phantom Liberty

(Image credit: Nvidia / CD Projekt RED)

Is DLSS 4 only for 50 series?

DLSS 4's flagship feature, Multi-Frame Generation, is only available on Nvidia's new RTX 50-series graphics cards.

However, the underlying transformer AI model that sharpens Super Resolution and Ray Reconstruction can be enabled on earlier RTX cards through the NVIDIA App, so you still gain image-quality improvements, just not the three-for-one frame generation feature.

The best games with DLSS support

(Image credit: CD Projekt)

The future of DLSS

DLSS began life in 2019 as a single-purpose upscaler and has since become a full neural-rendering pipeline.

Whether you crave buttery-smooth esports frame rates or want to push path-traced lighting without melting your GPU, DLSS gives you free performance in exchange for Tensor-Core horsepower you already own.

And with each new generation, the AI gets smarter, so flipping that little toggle in the options menu is likely to remain one of the biggest “instant upgrade” buttons in PC gaming for years to come.

You might also like...

John Loeffler
Components Editor

John (He/Him) is the Components Editor here at TechRadar and he is also a programmer, gamer, activist, and Brooklyn College alum currently living in Brooklyn, NY.

Named by the CTA as a CES 2020 Media Trailblazer for his science and technology reporting, John specializes in all areas of computer science, including industry news, hardware reviews, PC gaming, as well as general science writing and the social impact of the tech industry.

You can find him online on Bluesky @johnloeffler.bsky.social

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.