DLSS stands for Deep Learning Super Sampling. It is an AI-powered image upscaling technology developed by Nvidia that boosts video game performance by rendering graphics at a lower resolution and then using artificial intelligence to reconstruct them into a sharp, high-resolution output in real-time.
Nvidia introduced this feature to solve a massive hardware challenge: rendering modern video games at high resolutions like 4K with demanding graphics settings like ray tracing creates immense strain on graphics processing units (GPUs). By rendering fewer pixels initially, DLSS drastically lowers the workload on your graphics card, boosting frame rates without sacrificing visual quality. It is used exclusively on PC systems equipped with Nvidia GeForce RTX graphics cards.
DLSS leverages dedicated AI processors called Tensor Cores found only on Nvidia RTX graphics cards.
The technology allows games to run at higher frame rates while maintaining a sharp visual output.
It has evolved from a basic image upscaler into a full suite of neural rendering features, including frame generation and ray reconstruction.
DLSS requires explicit implementation by game developers within the game engine.
Nvidia has fundamentally shifted how DLSS processes graphics since its debut. The technology has progressed through three major generational pillars.
The original version required Nvidia to train its AI supercomputer network on a per-game basis. It struggled with motion blur and artifacting, meaning the visual quality often fell short of native resolution rendering.
A massive architectural leap. Nvidia introduced a generic AI network that no longer required training for specific games. It introduced temporal feedback, using vector data from previous frames to track moving objects, which vastly improved image stability and sharpness.
DLSS 3 introduced Optical Multi Frame Generation, which uses AI to insert entirely new frames between traditionally rendered ones, doubling perceived performance. DLSS 3.5 added Ray Reconstruction, replacing traditional denoisers with an AI network to generate cleaner lighting effects.
The core mechanics of DLSS rely on shifting the rendering burden away from traditional shader pipelines and onto specialized AI hardware.
Lower Resolution Rendering: The game engine renders the 3D scene at a fraction of your target display resolution (for example, rendering at 1080p for a 4K output).
Data Extraction: The graphics pipeline extracts low-resolution color data alongside temporal motion vectors, which track how pixels move from frame to frame.
Tensor Core Processing: This data is fed into the GPU Tensor Cores, where the pre-trained deep learning model analyzes the low-resolution input.
AI Reconstruction: The AI algorithm predicts what the missing pixels should look like based on millions of high-resolution reference images, outputting a final image that matches your monitor's native target resolution.
Modern DLSS is no longer a single tool; it is a suite of distinct AI features that can be toggled independently or combined in game settings.
The foundational element of DLSS. It takes a lower-resolution input frame and upscales it to your native screen resolution using spatial and temporal data.
Exclusive to RTX 40 Series GPUs. It analyzes sequential frames and optical flow data to insert a completely synthesized, AI-generated frame between every two rendered frames, bypassing CPU bottlenecks.
An AI model trained to improve ray-traced image quality. It replaces hand-tuned denoisers, recognizing patterns in lighting rays to produce sharper reflections and more accurate global illumination.
DLSS is a hardware-dependent technology tied directly to the Nvidia architecture. It does not work on graphics cards from other manufacturers.
| Feature | Supported Hardware |
|---|---|
| DLSS Super Resolution (2.x) | Nvidia GeForce RTX 20 Series, 30 Series, and 40 Series |
| DLSS Frame Generation (3.x) | Nvidia GeForce RTX 40 Series only |
| DLSS Ray Reconstruction (3.5) | Nvidia GeForce RTX 20 Series, 30 Series, and 40 Series |
Significant Frame Rate Boosts: Can double or triple game performance depending on the quality preset and hardware generation.
Overcomes Hardware Bottlenecks: Allows mid-range GPUs to achieve playable frame rates in heavy workloads like path tracing.
Cleans Up Visual Artifacts: The temporal anti-aliasing built into DLSS often eliminates shimmering and jagged edges better than native anti-aliasing techniques.
Hardware Lock-in: It requires an Nvidia RTX graphics card, offering zero utility for AMD or Intel GPU owners.
Visual Artifacts: In fast-moving scenes, it can occasionally introduce ghosting, minor blurring, or UI flickering.
Latency Overhead: Frame generation can introduce slight input latency, necessitating the use of Nvidia Reflex technology to compensate.
| Metric | Nvidia DLSS | AMD FSR (FidelityFX Super Resolution) | Intel XeSS (Xe Super Sampling) |
|---|---|---|---|
| Core Technology | Deep Learning AI (Tensor Cores) | Spatial/Temporal Upscaling | Deep Learning AI / Vector Math |
| Hardware Restriction | Locked to Nvidia RTX GPUs | Open-source; works on all GPUs | Works on most GPUs (Optimized for Intel Arc) |
| Image Quality | Industry Leader (High stability) | Good (Prone to shimmering at low resolutions) | Excellent on Arc hardware; good on others |
A standard resolution scaler simply stretches an image, resulting in blurriness. DLSS uses temporal data and AI models to intelligently reconstruct missing pixel details, often matching or exceeding native image sharpness.
Frame generation increases frame rates visually, but it adds a small amount of processing latency because frames must be held briefly in a buffer to calculate the generated frame. It requires anti-latency software to maintain a responsive feel.
Tensor Cores: Specialized hardware cores inside Nvidia RTX GPUs dedicated to accelerating AI and matrix math operations.
Ray Tracing: A rendering technique that simulates the physical behavior of light to achieve realistic reflections, shadows, and illumination.
Temporal Anti-Aliasing (TAA): A method of smoothing jagged edges by using data from past frames to clean up the current frame.
Spatial Upscaling: Scaling an image up based purely on the data present in a single frame, without looking at preceding frames.
Learn how AMD FSR 3 Fluid Motion Frames (AFMF) uses frame generation and optical flow technology to double game framerates across AMD, NVIDIA, and Intel GPUs.
Learn how Intel XeSS (Xe Super Sampling) uses AI-driven temporal reconstruction to boost PC gaming frame rates and performance across various GPUs.
Learn how path tracing works, how it differs from traditional ray tracing, and how it delivers photorealistic global illumination in modern games.
Learn about AMD Radeon RX graphics cards, from Polaris to RDNA architecture. Discover technical specs, key features, and how these GPUs work.
Learn about the Nvidia Quadro series. Discover how these workstation GPUs work, their key specifications, benefits, and how they differ from GeForce cards.