Version: 2021.3
Dynamic resolution
Multi-display

Deep learning super sampling

NVIDIA Deep Learning Super Sampling (DLSS) is a rendering technology that uses artificial intelligence to increase graphics performance and quality. You can use it to:

  • Run real-time ray traced worlds at high frame rates and resolutions.
  • Provide a performance and quality boost for rasterized graphics. This is particularly useful for virtual realityA system that immerses users in an artificial 3D world of realistic images and sounds, using a headset and motion tracking. More info
    See in Glossary
    applications to help them run at higher frame rates. This helps to remove disorientation, nausea, and other negative effects that occur at lower frame rates.

For more information about DLSS, see the NVIDIA documentation.

Requirements and compatibility

This section includes information about the hardware requirements and render pipelineA series of operations that take the contents of a Scene, and displays them on a screen. Unity lets you choose from pre-built render pipelines, or write your own. More info
See in Glossary
compatibility of DLSS.

For information about hardware and driver requirements, see the NVIDIA documentation.

Render pipeline compatibility

Feature Built-in Render Pipeline Universal Render Pipeline (URP) High Definition Render Pipeline (HDRP) Scriptable Render Pipeline (SRP)
DLSS No (1) No (1) Yes No (1)

Notes:

  1. The low-level DLSS framework code is in core Unity, but HDRP is the only render pipeline that uses it to implement DLSS. For more information, see the API documentation.

Supported platforms

Unity supports DLSS on the following platforms:

  • DirectX 11 on Windows 64 bit
  • DirectX 12 on Windows 64 bit
  • Vulkan on Windows 64 bit.

Unity does not support DLSS for Metal, Linux, Windows using x86 architecture (Win32), or any other platform. To build your project for Windows, use x86_64 architecture (Win64).

For information about the hardware requirements of DLSS, see NVIDIA’S DLSS requirements.

Using DLSS

For information on how to use DLSS in HDRP, see Deep learning super sampling in HDRP.

The Spaceship Demo Project rendered with a screen percentage of 50 without DLSS. The image is blurry, especially around the spark visual effect.

The Spaceship Demo Project rendered with a screen percentage of 50, but using DLSS. This image is less blurry than the previous image, especially around the spark visual effect.

Quality modes

The available quality modes are as follows:

  • Auto: Selects the best DLSS mode for the current output resolution.
  • Balance: Balances performance with image quality.
  • Quality: Emphasizes image quality over performance.
  • Performance: Emphasizes performance and image quality.
  • Ultra Performance: Produces the highest performance and the lowest image quality.
Dynamic resolution
Multi-display