What is DLSS 5?
NVIDIA Deep Learning Super Sampling (DLSS) is an image scaling technology that allows you to increase gaming performance with minimal loss of quality. DLSS is a pioneer of such technologies: its first version was released in 2019. How do different generations of NVIDIA smart scaling work, and what is the difference between them?
What is DLSS 5?
Announced at GTC 2026, DLSS 5 (Deep Learning Super Sampling 5) is NVIDIA's next-generation AI graphics technology that introduces real-time Neural Rendering.
Unlike previous versions that focused on frame interpolation, DLSS 5 uses generative AI models to "understand" scene semantics — automatically incorporating photorealistic lighting, subsurface scattering for skin, and intricate hair and light interactions to bridge the gap between real-time gaming and cinematic VFX.
DLSS 5 Features
Real-Time Neural Rendering
DLSS 5 analyzes a frame to determine scene semantics — distinguishing between skin, hair, and fabric — while understanding light conditions (backlit, cloudy, and more).
“Neural Rendering”
The model does not only upscale pixels. It reconstructs lighting and materials so the frame can look closer to offline VFX.
Understanding semantics
Because the AI model is trained from start to finish, it can accurately render Subsurface Scattering (SSS) on characters and the subtle luster on complex fabrics that traditional shaders find difficult to calculate in real time.
Closing the VFX Gap
CEO Jensen Huang described DLSS 5 as the moment when AI is no longer just "enhancing" graphics but actively "creating" photorealistic pixels to overcome the limits of raw hardware power.
“GPT Moment for Graphics”
— Jensen Huang
Generative models start producing the look of cinematic lighting instead of only interpolating or sharpening existing frames.
DLSS Version information
-
DLSS Super Resolution (2.x)
Temporal scaling. It is supported on any RTX card across a huge list of titles, from the first games with DLSS support to the most modern ones.
-
DLSS Frame Generation (3.0)
Frame generation. It only works on RTX 40 and RTX 50 series cards. The list of supported games is smaller — most of them shipped in 2023–2024. Updated DLSS 4 Frame Generation has largely replaced it.
-
DLSS Ray Reconstruction (3.5)
Ray reconstruction. It is supported on all RTX cards, but it is not widespread in games. The newer image-recovery algorithms work only when ray tracing is enabled — classic rasterization still uses Super Resolution techniques.
-
DLSS 4.x Super Resolution
Improved temporal scaling. Compatible with all RTX graphics cards. In any game with DLSS 2.x support, the older upscaling model can be replaced with one of the new ones. This feature is supported in the NVIDIA App starting with Game Ready 572.16.
-
DLSS 4 Ray Reconstruction
Improved ray reconstruction. It also works on any RTX cards, but support depends on the game. If a title is compatible with DLSS 5, DLSS 4.x Super Resolution still helps accelerate improved ray reconstruction.
-
DLSS 4 Frame Generation
Improved frame generation with one intermediate frame inserted. It works on RTX 40 and RTX 50 series cards in many recent games.
-
DLSS 4.x Multi Frame Generation
Multi-frame generation that inserts two to five frames. It is only compatible with RTX 50 series cards that support multi-generation, but over time it should appear in projects that already support regular Frame Generation.
DLSS 5 changes the essence of rendering. This is no longer just the smoothing of "ladders" or the generation of intermediate frames. It is a Neural Rendering technology — not scaling, but lighting and materials.
What is DLSS for in games?
As mentioned above, this technology helps increase the speed of the game without hurting graphics quality. In practice, it works with a bang.
Using DLSS, you do not give up FPS for prettier graphics — you raise both. When the technology is enabled, FPS in games often increases 1.5–2×. The load on the GPU drops, and the card does not heat up as much.