Technology News

NVIDIA DLSS 5: Exclusive to GeForce RTX 50 Series GPUs and Beyond

August 10, 2026Carlos Mendoza3 мин

NVIDIA utilized the SIGGRAPH 2026 event to provide a deeper explanation of how DLSS 5 will function. The significant news is that this technology is exclusively intended for GeForce RTX 50 graphics cards, built on the Blackwell architecture. Consequently, there has been no announcement of compatibi

NVIDIA utilized the SIGGRAPH 2026 event to provide a deeper explanation of how DLSS 5 will function. The significant news is that this technology is exclusively intended for GeForce RTX 50 graphics cards, built on the Blackwell architecture. Consequently, there has been no announcement of compatibility with GeForce RTX 40 Series graphics, let alone previous generations. This exclusivity is understandable given the substantial leap in image quality this technology is expected to deliver.

A prime example of this is that early live demonstrations of DLSS 5 technology were only feasible with two GeForce RTX 5090 cards. Following this revelation, NVIDIA stated that by the time the technology becomes available this autumn, it will have matured sufficiently for use with a single graphics card. It's important to remember that this technology goes beyond simply reconstructing an image at a higher resolution or generating intermediate frames to enhance fluidity. Instead, it incorporates a neural rendering model that modifies lighting, materials, and specific visual details of each scene in real-time.

GeForce RTX 50 GPUs Will Be the Sole Users of the Diffusion Model for DLSS 5 Lighting and Material Reconstruction

Concluding the main narrative, it's not surprising that NVIDIA DLSS 5 will remain exclusive to the company's most modern and powerful architecture. For instance, even a highly capable GPU like the GeForce RTX 3090 or RTX 3080 is not even compatible with DLSS Frame Generation technology, which debuted with DLSS 4. This feature should ideally be fundamental.

The model has been trained to recognize elements such as characters, hair, fabrics, translucent skin, reflections, and various lighting conditions. Based on this information, it can introduce complex effects like subsurface scattering on skin, more convincing light interaction with hair, or more detailed reflections and shadows. DLSS 5 achieves effects that would typically be extremely computationally expensive to render in real-time using traditional methods.

NVIDIA emphasizes that DLSS 5 does not replace rasterization, Ray Tracing, Path Tracing, or other DLSS technologies. The game engine will continue to be responsible for generating the geometry, materials, normals, motion vectors, and the original structure of the frame. DLSS 5 is added at the end of the pipeline to visually enhance the output, which is why the company presents it as a new category of neural rendering, not merely another upscaler.

It's worth recalling that NVIDIA recently presented the new features of DLSS 5 following its somewhat challenging debut. After the official presentation, it was revealed that studios will now have a choice of three DLSS 5 models. The idea is that each model will be trained with different data and parameters. These models can be switched within the game itself and applied independently to specific scenarios, characters, or objects. The aim is to allow developers to select the finish that best aligns with the artistic direction of each scene. This approach will prevent the application of a single, general filter across the entire image that might detract from the artistic vision.

Two main controls will also be incorporated: structural intensity and tonal intensity. The former regulates high-frequency details like ambient occlusion, reflections, or subsurface scattering, while the latter modifies more general components such as scene lighting, color, and tone. Furthermore, developers can utilize masks to exclude faces, characters, objects, or areas of the scene, or assign different configurations to them. NVIDIA is currently collaborating with its partners to further refine the technology.

As previously indicated, the most significant change affects hardware requirements. The initial demonstrations of DLSS 5 utilized two GeForce RTX 5090 cards. The reason for this was that one RTX 5090 was used to run the game, while the second RTX 5090 was dedicated almost exclusively to executing the AI model. NVIDIA assures that the commercial version has been optimized to function with a single graphics card through a more compact model and more efficient VRAM management. It will be interesting to see how DLSS 5 performs on much more modest graphics cards like the RTX 5070 or RTX 5060.