Omniverse RTX Renderer

Overview

The NVIDIA Omniverse RTX™ Renderer is a scalable photo-real renderer leading the convergence of real-time and offline rendering to simulate the world.

• Built for NVIDIA RTX GPUs on standards such as USD and NVIDIA Material Definition Language, for the Windows and Linux operating systems

• Pushes the limits of RTX rendering with a combination of AI, Multi-GPU, and the Cloud

• Supporting various sectors such as autonomous vehicles, robotics, architecture, engineering, construction, design, manufacturing, scientific visualization, media and entertainment

It has one common core of features and two render modes for different target-uses, RTX - Real-Time 2.0 and RTX - Interactive (Path Tracing), both solving the rendering equation with a path tracing-based algorithm.

RTX - Real-Time 2.0 mode

Omniverse RTX Renderer provides the RTX - Real-Time 2.0 mode, a physically based path-tracing rendering mode leveraging the NVIDIA DLSS suite of neural rendering technologies, which enables rendering more geometry at higher fidelity than rasterization techniques, at real-time performance.

The RTX - Real-Time 2.0 mode is complemented by the RTX – Interactive (Path Tracing) mode, which can achieve higher fidelity results with some performance trade-offs. See the RTX – Interactive (Path Tracing) mode documentation for more details.

Note

In order to achieve real-time performance, some effects such as sub-surface scattering, depth of field, and volume rendering may diverge more than other effects when compared to the RTX – Interactive (Path Tracing) mode.

See the RTX - Real-Time 2.0 section for more details.

RTX – Interactive (Path Tracing) mode

Omniverse RTX Renderer provides the RTX – Interactive (Path Tracing) mode, a cutting-edge rendering mode designed to accurately simulate the behavior of light in order to predict how illumination interacts with materials and environments in real-world scenarios.

RTX – Interactive (Path Tracing) mode is complemented by the RTX - Real-Time 2.0 mode, which can achieve similar results with some fidelity trade-offs, at real-time performance. See the RTX - Real-Time 2.0 mode documentation for more details.

See the RTX – Interactive (Path Tracing) section for more details.

Common Features

Multi-GPU Rendering

The Omniverse RTX Renderer supports multi-GPU rendering for single and multi-view, with automatic load-balancing and near-linear speedup for each frame, on a single system with up to 16 GPUs.

Multi-GPU render settings are available for both RTX - Real-Time 2.0 and RTX – Interactive (Path Tracing) modes.

For more details, see:

• Multi-GPU rendering with RTX - Real-Time 2.0 mode

• Multi-GPU rendering with RTX - Interactive (Path Tracing) mode

Materials

The Omniverse RTX Renderer supports rendering physically-based materials such as:

• NVIDIA Material Definition Language (MDL)

• USD Preview Surface

Light Types

The Omniverse RTX Renderer supports rendering thousands of dynamic lights of a variety of types with no light baking, such as:

• Sphere

• Rectangle, including as Projector Light Type

• Disk

• Cylinder

• Distant

• Dome, including MDL-based procedurals

• Mesh Lights

Geometry types

• Triangle Meshes

• Subdivision Surfaces (OpenSubdiv)

• Basis Curves

• Points

Cameras

• Flexible camera lens distortion model (fTheta) and more

Post-Processing

• Tone Mapping and Auto-Exposure

• Color Correction

• Color Grading

• XR Compositing

• Matte Object

• Chromatic Aberration

• Depth of Field

• Motion Blur

• FFT Bloom

• TV Noise & Film Grain

• ReShade

Volume Rendering

• NVIDIA Flow

• Global Volumetric Effects

• Simple Fog

Geometry and Texture Streaming

The Omniverse RTX Renderer supports automated GPU resources management such as geometry and texture streaming, which enables high-quality rendering in real-time even for large datasets.

See the GPU Resources Management section for more details.

Debug Views

Debug Views are available for various debugging needs, such as:

• Path Tracing passes for AOVs, Multi Matte, Adaptive Sampling error

• Heat Maps to correlate colors to per-pixel measurements such as GPU timing spent as well as Any Hit and Intersection shader invocation count

See the Debug Views section for more details.

Feature Support per GPU Architecture

Some Omniverse RTX Renderer features are only supported under certain conditions. The ability to run Omniverse SDKs on non-RTX GPUs is provided without any support guarantees. The table below details feature availability on each GPU architecture:

GPU Architecture [3]	Compute Capability	DLSS Ray Reconstruction [1]	DLSS Frame Generation	OptiX Denoiser	Shader Execution Reordering	Opacity Micro-Map	Motion BVH
Blackwell (RTX)	12.0	YES [2]	YES	YES	YES (DX12)	YES	YES (Vulkan)
Blackwell B200/B300	10.0	YES [2]	NO	YES	NO	NO	NO
Hopper H100/H200/H800	9.0	NO	NO	YES	NO	NO	NO
Ada Lovelace	8.9	YES	YES	YES	YES (DX12)	YES	YES (Vulkan)
Ampere (RTX)	8.6	YES	NO	YES	NO	NO	YES (Vulkan)
Ampere A100/A30/A800	8.0	NO	NO	YES	NO	NO	NO
Turing	7.5	YES	NO	YES	NO	NO	NO

Note

[1]

Without DLSS Ray Reconstruction support, images appear significantly noisy when using the RTX Real-Time render mode.

[2] (1,2)

For Blackwell GPUs on Linux, DLSS Ray Reconstruction is not supported on Kit versions lower than 106.5.3.

[3]

Older GPU architectures such as Tesla, Fermi, Kepler, Maxwell, Pascal and Volta are not supported by the Omniverse RTX Renderer. Learn more about CUDA GPU Compute Capability and Legacy GPU CUDA Compute Capability.

Render Settings

• RTX - Common
• RTX - Real-Time 2.0 mode
• RTX – Interactive (Path Tracing) mode
• RTX Real-Time (Legacy) mode
• Post-Processing
• Feature Support per GPU Architecture

Release Notes

• RTX Renderer Release Notes