RTX - Common

This page documents common render settings that apply to any RTX Renderer render mode.

Geometry

Misc

Display Name	Setting Name	Type and Default Value	Description
Normal & Tangent Space Generation Mode	/rtx/hydra/TBNFrameMode	Int 0	Mode selection for vertex Normals and Tangents generation. - Auto (0): Selects the mode depending on available data and data update pattern. - CPU (1): Uses mikktspace to generate tangent basis on the CPU. - GPU (2): Allows normal and tangent basis update on the GPU to avoid the CPU overhead, such as for deforming meshes. - Force GPU (3): Forces GPU in any circumstances.
Back Face Culling	/rtx/hydra/faceCulling/enabled	Bool False	Enables back face culling for ‘Single Sided’ primitives. This applies to UsdGeom prims set to ‘Single Sided’ since the Double Sided flag is ignored.
USD ST Main Texcoord as Default UV Set	/rtx/hydra/uvsets/stIsDefaultUvSet	Bool True	If enabled, ST is considered to be the default UV set name.
Hide Geometry That Uses Opacity (debug)	/rtx/debug/onlyOpaqueRayFlags	Bool False	Allows hiding all objects which have opacity enabled in their material.
Instance Selection	/rtx/hydra/instancePickingEnabled	Bool True	Enables the picking of instances.
Renderer-internal Meters Per Unit	/rtx/scene/renderMeterPerUnit	Float -1.0	Number of units per meter used by the renderer relative to the scene scale. Some materials depend on scene scale, such as subsurface scattering and volumetric shading.
Ray Offset	/rtx/raytracing/rayOffset	Float 0.0	Offset used to prevent ray-triangle self intersections.
Points Default Width	/persistent/rtx/hydra/points/defaultWidth	Float 1.0	Uses this value if width is not specified in UsdGeomPoints.

Wireframe

Display Name	Setting Name	Type and Default Value	Description
World Space Thickness	/rtx/wireframe/wireframeThicknessWorldSpace	Bool True	Interprets the wireframe thickness value in world space instead of screen space.
Thickness	/rtx/wireframe/globalWireframeThicknessWorldSpace	Bool False	Wireframe thickness in wireframe mode.

Subdivision

Display Name	Setting Name	Type and Default Value	Description
Global Refinement Level	/rtx/hydra/subdivision/refinementLevel	Int 0	The refinement level for all primitives with Subdivision Schema not set to None. Each increment increases the mesh triangle count by a factor of 4.
Subdivision Feature-adaptive Refinement	/rtx/hydra/subdivision/adaptiveRefinement	Bool False	Feature-adaptive refinement automatically increases or reduces refinement level based on geometric features. This reduces the number of polygons used in flat areas for example.

Curves

Display Name	Setting Name	Type and Default Value	Description
Global Number of BVH Splits	/rtx/hydra/curves/splits	int 4	Higher number of splits results in faster rendering but longer BVH build time. The speed up depends on the geometry: long and thin curve segments tend to benefit from more splits. Memory used by the BVH grows linearly with the number of splits.

Materials

MDL

Display Name	Setting Name	Type and Default Value	Description
Translucency as Opacity	/rtx/material/translucencyAsOpacity	Bool False	When enabled, the translucent material will be converted into opacity blending based surface.
Animation Time Override	/rtx/animationTime	Float -1.0	Overrides the time value provided to MDL materials.
Animation Time Use Wallclock	/rtx/animationTimeUseWallclock	Bool False	Use actual elapsed time for animation instead of simulated time.
MDL Displacement	/rtx/material/enableMDLDisplacement	Bool False	Enable MDL material displacement. Enabling it can negatively impact stage load time when there are many materials with displacement support (like OmniSurface) in a stage. Requires stage reload to take effect.

Background

Display Name	Setting Name	Type and Default Value	Description
Background Override	/rtx/background/source/type	Int 0	By default, the background is sourced from a dome light if one is rendered, otherwise the Background Texture if one is assigned, otherwise the Background Color. This can be overridden to force the use of either the Texture or Color. - None (0): Default behavior, no override. - Texture (1): Forces the assigned texture to be rendered as the background. - Color (2): Forces the color to be rendered as the background.
Background Texture	/rtx/background/source/texture/path	Asset None	The path to the texture to use as the background.
UV Sampler	/rtx/background/source/texture/textureMode	Int 0	Determines the texture lookup behavior: - Repeat (0): The texture UV value is modulated with 1. - Mirror (1): The UV value is modulated by 2 then mirrored by x=1 and y=1. - Clamp (2): The UV value is clamped to [0, 1].
Color Space	/rtx/background/source/texture/colorSpace	Int 0	Sets the color space for the background texture to linear space. - sRGB (0): sRGB color space. - lin_rec709 (1): Linear color space based on the Rec. 709 standard.
Luminance Scale	/rtx/background/source/texture/luminanceScale	Float 1.0	Scales the background texture luminance.
Background Color	/rtx/background/source/color	COLOR3 0.0, 0.0, 0.0	The color to use as the background.

Lighting

Light Visibility

Display Name	Setting Name	Type and Default Value	Description
Show Area Lights in Primary Rays	/rtx/raytracing/showLights	Int 0	Defines if area lights are visible or invisible in primary rays. - Per-Light Enable (0): The light’s value is used. - Force Enable (1): True for all lights, regardless of their value. - Force Disable (2): False for all lights, regardless of their value.
Invisible Lights Refractions Ghost Factor	/rtx/raytracing/invisLightRefractionsGhostFactor	Float 0.0	Reduces glossy and specular refractions of lights invisible in primary rays. At 0, all lights will be fully visible. In (0,1], refractions will gradually fade out until all glossy refractions vanish at 1 and only diffuse contributions remain. - May be overwritten per light source. - Replaces deprecated /rtx/raytracing/invisLightRoughnessThreshold.
Invisible Lights Reflections Ghost Factor	/rtx/raytracing/invisLightReflectionsGhostFactor	Float 0.0	Reduces glossy and specular reflections of lights invisible in primary rays. At 0, all lights will be fully visible. In (0,1], reflections will gradually fade out until all glossy reflections vanish at 1 and only diffuse contributions remain. - May be overwritten per light source. - Replaces deprecated /rtx/raytracing/invisLightReflectionsRoughnessThreshold.
Use First Distant Light & First Dome Light Only	/rtx/scenedb/skipMostLights	Bool False	Disable all lights except the first distant light and first dome light.
Shadow Bias	/rtx/raytracing/shadowBias	Float 0.001	Offset applied for shadow ray origin along the surface normal. Reduces self-shadowing artifacts.

Dome Light

Display Name	Setting Name	Type and Default Value	Description
Lighting Mode	/rtx/domeLight/upperLowerStrategy	Int 0	Select how to sample the Dome Light. - Image-Based Lighting (0): Most accurate even for high-frequency Dome Light textures. Can introduce sampling artifacts in real-time mode. - Approximated Image-Based Lighting (4): Fast and artifacts-free sampling in real-time mode but only works well with a low-frequency texture, for example a sky with no sun disc where the sun is instead a separate Distant Light. Requires enabling Direct Lighting denoiser. - Limited Image-Based Lighting (3): Only sampled for reflection and refraction. Fastest, but least accurate. Good for cases where the Dome Light contributes less than other light sources.
Baking Resolution	/rtx/domeLight/baking/resolution	Int 4096	The baking resolution of the Dome Light texture when an MDL material is used as its image source.

Section Plane

The Section Plane feature allows creating and placing one or many infinitely-sized planes to clip geometry and lights. This feature is particularly useful for visualizing cross-sections of complex models and scenes.

Display Name	Setting Name	Type and Default Value	Description
Section Plane	/rtx/sectionPlane/enabled	Bool False	Enables toggling the section plane feature on and off.
Plane	/rtx/sectionPlane/plane	Float Array 0.0, 0.0, 0.0, 0.0	A float array in multiples of four floats to define planes using the equation (Ax + By + Cz + D = 0), where (A, B, C) is the Normal Vector and (D) is a scalar value that, when combined with the normal vector, defines the plane’s position in space. For example, /rtx/sectionPlane/plane = [1.0, 0.0, 0.0, 0.0, 0.0, -1.0, 0.0, 0.0] would create two section planes oriented perpendicularly from one another at the world origin.

Note

For an example of a Kit extension using the Section Plane feature to create a single section plane, refer to the Section Tool Kit Extension documentation.

Simple Fog

Display Name	Setting Name	Type and Default Value	Description
Simple Fog	/rtx/fog/enabled	Bool False	Enables the use of a simple and fast exponential fog approximation which only models some absorption, no light scattering.
Color	/rtx/fog/fogColor	Float3 0.75, 0.75, 0.75	The color or tint of the fog volume.
Intensity	/rtx/fog/fogColorIntensity	Float 1.0	The intensity of the fog effect.
Height-based Fog - Use +Z Axis	/rtx/fog/fogZup/enabled	Bool False	Use positive Z axis for height-based fog. Otherwise use the positive Y axis.
Height-based Fog - Plane Height	/rtx/fog/fogStartHeight	Float 1.0	The starting height (in meters) for height-based fog.
Height Density	/rtx/fog/fogHeightDensity	Float 1.0	Density of the height-based fog. Higher values result in thicker fog.
Height Falloff	/rtx/fog/fogHeightFalloff	Float 1.0	Rate at which the height-based fog falls off.
Start Distance to Camera	/rtx/fog/fogStartDist	Float 0.0	Distance from the camera at which the fog begins.
End Distance to Camera	/rtx/fog/fogEndDist	Float 5000.0	Distance from the camera at which the fog achieves maximum density.
Distance Density	/rtx/fog/fogDistanceDensity	Float 1.0	The fog density at the End Distance.

Global Volumetric Effects

The global volumetric effects use a three-dimensional “voxel grid”, a camera aligned frustum divided into several depth slices parallel to the camera plane along the camera direction. These depth slices are themselves subdivided into a number of frustum shaped voxels along the width and height of the slice (the amount of voxels is controlled by the “Pixel Density”).

Display Name	Setting Name	Type and Default Value	Description
Fog Height	/rtx/raytracing/inscattering/atmosphereHeight	Float 1000.0	Height in world units (centimeters) at which the medium ends. Useful for atmospheres with distant lights or dome lights.
Fog Height Falloff	/rtx/pathtracing/ptvol/fogHeightFallOff	Float 10.0	Exponential decay of the fog above the Fog Height.
Maximum Inscattering Distance	/rtx/raytracing/inscattering/maxDistance	Float 50000.0	Maximum depth in world units (centimeters) the voxel grid is allocated to. If set to 10,000 with 10 depth slices, each slice will span 1,000 units (assuming a slice distribution exponent of 1). Ideally this should be kept as low as possible without causing artifacts to make the most of the fixed number of depth slices.
Density Multiplier	/rtx/raytracing/inscattering/densityMult	Float 1.0	Scales the fog density.
Transmittance Measurement Distance	/rtx/raytracing/inscattering/transmittanceMeasurementDistance	Float 10000.0	Controls how far light can travel through fog. Lower values yield thicker fog.
Transmittance Color	/rtx/raytracing/inscattering/transmittanceColor	Float3 0.5, 0.5, 0.5	Assuming a white light, it represents its tint after traveling a number of units through the volume as specified in Transmittance Measurement Distance.
Single Scattering Albedo	/rtx/raytracing/inscattering/singleScatteringAlbedo	Float3 0.9, 0.9, 0.9	The ratio of scattered-light to attenuated-light for an interaction with the volume. Values closer to 1 indicate high scattering.
Anisotropy Factor (g)	/rtx/raytracing/inscattering/anisotropyFactor	Float 0.0	Anisotropy of the volumetric phase function, or the degree of light scattering asymmetry. -1 is back-scattered, 0 is isotropic, 1 is forward-scattered.
Apply Density Noise	/rtx/raytracing/inscattering/useDetailNoise	Bool False	Enables modulating the density with a noise. Enabling this option can reduce performance.
World Scale	/rtx/raytracing/inscattering/detailNoiseScale	Float 0.002	A scale multiplier for the noise. Smaller values produce more sparse noise.
Animation Speed X/Y/Z	/rtx/raytracing/inscattering/noiseAnimationSpeed(X,Y,Z)	Float 0.0	The X/Y/Z vector for the noise shift when animated.
Scale Min/Max	/rtx/raytracing/inscattering/noiseScaleRangeMin	Float 0.0	A range to map the noise values of each noise octave. Typically these should be 0 and 1, but to make sparse points in the noise less sparse a higher min can be used, or a lower max for dense points to be less dense.
Octave Count	/rtx/raytracing/inscattering/noiseNumOctaves	Int 4	A higher octave count results in greater noise detail, at the cost of performance.

Flow

Display Name	Setting Name	Type and Default Value	Description
Flow	/rtx/flow/enabled	Bool False	Enables simulation and rendering of Flow volumes.
Flow in Real-Time Ray Traced Shadows	/rtx/flow/rayTracedShadowsEnabled	Bool False	Flow volumes in ray-traced shadows in real-time mode.
Flow in Real-Time Ray Traced Reflections	/rtx/flow/rayTracedReflectionsEnabled	Bool False	Flow volumes in ray-traced reflections in real-time mode.
Flow in Real-Time Ray Traced Translucency	/rtx/flow/rayTracedTranslucencyEnabled	Bool False	Flow volumes in ray-traced translucency in real-time mode.
Flow in Path-Traced Mode	/rtx/flow/pathTracingEnabled	Bool False	Flow volumes in Path-Traced mode.
Flow in Path-Traced Mode Shadows	/rtx/flow/pathTracingShadowsEnabled	Bool False	Flow volumes in shadows cast by light primitives in Path-Traced mode.
Composite with Flow Library renderer	/rtx/flow/compositeEnabled	Bool True	Use Flow’s built-in renderer and composite the results instead of ray tracing through Flow volumes in the RTX renderer.
Use Flow Library Self Shadowing	/rtx/flow/useFlowLibrarySelfShadow	Bool True	Enable the casting of shadows by the volume on to itself when using Flow’s built-in renderer.
Max Blocks	/rtx/flow/maxBlocks	Int 0	Maximum number of Flow volume blocks. To keep rendering performance high and memory consumption low, this number needs to be kept as small as possible, while also high enough to contain the entire volume.

GPU Resources Management

The texture and geometry streaming system optimizes GPU memory and renderer resource usage by dynamically loading the most relevant texture mips and geometry relative to the view, available memory, and renderer limits.

When geometry streaming is enabled, the RTX renderer can:

• Select the subset of geometry that fits within the geometry GPU memory budget.

• Prioritize loading geometry that is most relevant to the active view.

• Cache GPU-ready geometry so that geometry streamed out and later streamed back in does not need to be fully reprocessed from USD.

• Select the most relevant instances and unique meshes when instance or unique geometry budgets are reached.

• Balance geometry memory consumption with other streaming systems, such as texture streaming.

While the RTX renderer is able to render millions of instances, high instance counts can negatively impact performance because top-level acceleration structure (TLAS) construction can become expensive. The streaming system allows reducing the maximum rendered instance count while dynamically selecting the most relevant subset of loaded instances to render.

How Geometry Is Prioritized

Geometry streaming prioritizes geometry, instances, and unique meshes by estimating their importance to the current view. For each relevant camera, the system estimates the projected solid angle of each geometry instance’s bounding sphere. For a unique mesh or shared geometry, it sums the projected solid angles of all of its instances and uses that total for relevance ordering. This keeps geometry used by many small visible instances from being ranked as unimportant solely because each instance is small. This is a coarse approximation: cameras are currently represented by position, and relevant cameras have equal weight.

This ordering is used to decide what to load, what to keep resident, which instances and unique meshes to render when budgets are reached, and which geometry requests have higher GPU memory priority. Because large scenes can contain many instances, this computation runs on a time-sliced background thread. After a camera move, it can take a few frames for streaming to settle on the new most-relevant set.

Unique Meshes and Resource Descriptors

Large CAD, manufacturing, and digital twin datasets can contain very high numbers of unique meshes after renderer de-duplication. The RTX renderer uses GPU resource descriptors for geometry buffers and for other GPU resources such as textures. A unique mesh can require multiple resource descriptors, and the descriptor pool is shared with other renderer resources. As a result, the practical unique mesh limit is lower than the raw descriptor count and can vary by renderer version, content, and scene resource usage.

Geometry streaming provides a unique geometry budget to avoid exceeding this limit. When the number of unique meshes exceeds the budget, the renderer selects the most relevant unique meshes to display and omits the rest until the relevant set changes.

Note

In RTX statistics, the number of meshes after de-duplication is the value most directly related to unique mesh descriptor pressure. If a scene has a very high value there, reducing duplicate or near-duplicate mesh data can improve stability and reduce popping when geometry streaming budgets are active.

How to Enable Geometry Streaming

Currently, geometry streaming is disabled by default. There are two common ways to enable geometry streaming:

• In Application Preferences:

  Open the application’s Preferences panel, under Rendering > RTX Geometry Streaming and toggle Enable Geometry Streaming on.

  Note

  You must restart the application for the setting’s change to take effect.

• Command Line:

  At application startup, add the following commands:

  --enable omni.ujitso.client
  --/UJITSO/enabled=true
  --/persistent/UJITSO/geometry=true
  

  Use --/UJITSO/geometry=true instead of --/persistent/UJITSO/geometry=true to enable the setting for the current session only.

Limitations

Some types of geometries are not yet supported for geometry streaming and fall back to the uncached processing and loading.

These are:

• Meshes with the flag IsVolume set to true

• Skeletal meshes with subdivision enabled

• Meshes with instance primvar data

• Dynamic meshes with displacement

Performance and Authoring Guidance

When geometry streaming is active in a budget-constrained scene, the structure of the source data has a large effect on visual quality and performance:

• Very large objects or assemblies with large bounding volumes are difficult to prioritize accurately. If a large merged object spans both foreground and background areas, the renderer cannot lower detail only for the background portion of that object.

• Prefer object scopes and bounding boxes that match the meaningful object size in the scene. For example, many small detailed parts should not be merged into one large mesh if those parts need independent streaming decisions.

• Repeated mesh data should be de-duplicated or authored as instances where possible. This reduces unique mesh count, descriptor pressure, and geometry memory usage.

• Replacing repeated simple mesh shapes with appropriate primitive or instanced representations can reduce unique mesh count.

• Lowering the instance budget can improve TLAS construction time, but setting it below the number of visible or important instances can cause visible popping while the renderer selects the most relevant subset.

• Streaming geometry in for many instances at once can cause frame-rate stutter. Scene graph instancing can also reduce streaming effectiveness when many instances of the same mesh need different streaming states.

Settings

Display Name	Setting Name	Type and Default Value	Description
Texture Streaming Budget Priority	/rtx/resourcemanager/gpuBudgetPriority	Float 0.5	Sets the priority at which GPU memory is assigned to texture streaming over other memory requests like geometry streaming.
Geometry Streaming GPU Memory Budget	/rtx/hydra/geometrystreaming/gpuBudgetMB	Int 64000	Overrides the maximum GPU memory budget, in MB, available to geometry streaming.
Geometry Streaming Budget Priority	/rtx/hydra/geometrystreaming/gpuBudgetPriority	Float 0.5	Sets the priority at which GPU memory is assigned to geometry streaming relative to other memory requests such as texture streaming.
Instance Streaming Budget	/rtx/hydra/geometrystreaming/instanceBudget	Int 16777216	The most relevant instances among all instances in memory are rendered, up to this instance budget count. A high instance count can result in slow top-level acceleration structure (TLAS) construction time and suboptimal real-time rendering performance.
Unique Geometry Budget	/rtx/hydra/geometrystreaming/UniqueGeometryBudget	Int 200000	The most relevant unique meshes are rendered, up to this unique geometry budget. This budget helps avoid GPU resource descriptor exhaustion in scenes with many unique meshes after de-duplication.

View

Data Window

The Data Window settings are used to adjust the data window dimensions, such as for cropping or overscan, by controlling the rendering data window dimensions in normalized device coordinates (NDC) separately from the display window.

• Cropping: If the data window is smaller than the display window, areas outside the data window are filled with black rather than being rendered, which speeds up rendering compared to rendering the full display window.

• Overscan: If the data window is larger than the display window, areas outside the display window are rendered, which can enable sampling extra pixels beyond the aperture boundary to support some post-processing effects.

Note

EXR headers include both data window and display window attributes.

Display Name	Setting Name	Value Type	Default Value	Description
Fit output to data window	/rtx/dataWindow/fitOutputToDataWindow	Bool	False	When enabled, the output AOV sizes will match the data window dimensions. Otherwise, those AOV sizes are set to the display window dimensions.
X Min	/rtx/dataWindowNDC/0	Float	0.0	X min coordinates, in normalized device coordinates (NDC).
Y Min	/rtx/dataWindowNDC/1	Float	0.0	Y min coordinates, in normalized device coordinates (NDC).
X Max	/rtx/dataWindowNDC/2	Float	1.0	X max coordinates, in normalized device coordinates (NDC).
Y Max	/rtx/dataWindowNDC/3	Float	1.0	Y max coordinates, in normalized device coordinates (NDC).

Debug View

Various render targets are available to be previewed in the viewport. Some are useful to identify potential areas of optimization, and may offer specific viewing settings such as to remap the values to a heat map representation for convenience.

Display Name	Setting Name	Type and Default Value	Description
Render Target	/rtx/debugView/target	String Off	A list of all render passes which can be visualized.
Output Value Scaling	/rtx/debugView/scaling	Float 1.0	Scales the output value by this factor. Useful to accentuate differences.

RTX - Interactive (Path-Tracing) Views

Display Name	Description
PT Adaptive Sampling Error	Allows visualizing the normalized standard deviation of the Monte Carlo estimator of the pixels: warm colors represent high variance, which indicate that additional samples would lead to improved convergence for those pixels.
PT Denoised Result	Final render output, after denoising.
PT Noisy Result	Final render output, before denoising.
PT AOV Background	Shading of the background, such as the background resulting from rendering a Dome Light.
PT AOV Diffuse Filter	The raw color of the diffuse texture.
PT AOV Direct Illumination	Shading from direct paths to light sources.
PT AOV Global Illumination	Diffuse shading from indirect paths to light sources.
PT AOV Illuminance	The amount of light falling onto (illuminating) and spreading over a given surface area.
Value	The value at the last clicked position.
Display As Heatmap	Displays the aov as a heatmap.
Auto Range	Automatically scales the min and max values for the heatmap.
Min	Minimum range value for the heatmap.
Max	Maximum range value for the heatmap.
Show Heatmap Legend	Enables displaying the heatmap legend.
PT AOV Luminance	The intensity of light emitted from a surface per unit area in a given direction.
Value	The value at the last clicked position.
Display As Heatmap	Displays the aov as a heatmap.
Auto Range	Automatically scales the min and max values for the heatmap.
Min	Minimum range value for the heatmap.
Max	Maximum range value for the heatmap.
Show Heatmap Legend	Enables displaying the heatmap legend.
PT AOV Motion Vectors	Motion vectors.
PT AOV Reflections	Shading from indirect reflection paths to light sources.
PT AOV Reflection Filter	The raw color of the reflection, before being multiplied for its final intensity.
PT AOV Refractions	Shading from refraction paths to light sources.
PT AOV Refraction Filter	The raw color of the refraction, before being multiplied for its final intensity.
PT AOV Self-Illumination	Shading of the surface’s own emission value.
PT AOV Subsurface Filter	The raw color of the subsurface scattering texture.
PT AOV View Normal	The surface’s normal in view-space.
PT AOV Volumes	Shading from VDB volumes.
PT AOV World Normal	The surface’s normal in world-space.
PT AOV World Position	The surface’s position in world-space.
PT AOV Z-Depth	The surface’s depth relative to the view position.

Refer to AOV render settings and Multimatte render settings for more details about each AOV and how to enable them for preview.

Performance Heat Map Views

Heat Maps correlate colors from a color gradient to per-pixel measurements from a rendering pass which is useful for debugging and for profiling. The warmer the color, the higher the value it represents.

Display Name	Description
Timing Heat Map	Measures the amount of frame time spent on a pixel.
Any Hit Heat Map	Any Hit shaders are invoked when ray intersections are not opaque, which can incur a high performance cost. This heat map counts how many Any Hit shaders were invoked in a pixel for the selected pass. It is often used to find materials that have unwarranted use of translucency.
Intersection Heat Map	Custom Intersection shaders are invoked when rays intersect with curves. This heat map counts how many custom Intersection shaders were invoked in a pixel for the selected pass.
Pass	Selects a pass to measure.
Maximum Time (microseconds) / Any Hit / Intersection	The color palette is remapped between 0 and the maximum value.
Color Palette	Different color palettes can be selected for viewing convenience.
Show Color Bar	Toggles the display of the Color Bar in the viewport.

Debug

Streaming Settings

Display Name	Setting Name	Type and Default Value	Description
Texture Streaming (toggling requires scene reload)	/rtx-transient/resourcemanager/enableTextureStreaming	Bool True	Enables texture streaming which allows to load textures at a resolution which meets the view’s needs. This reduces memory usage and load times. Toggling the setting requires reloading the scene for it to be applied.
Texture Streaming Budget (% of GPU memory)	/rtx-transient/resourcemanager/texturestreaming/memoryBudget	Float 0.6	Limits memory budget used for texture streaming.
Texture Streaming Budget Per Request (in MB)	/rtx-transient/resourcemanager/texturestreaming/streamingBudgetMB	Int 200	Maximum budget per streaming request. 0 = unlimited but could lead to stalling during streaming. High or unlimited budget could lead to stalling during streaming.
Geometry Streaming Budget Per Frame (in MB)	/rtx-transient/hydra/geometrystreaming/streamingBudgetMB	Int 250	Maximum geometry disk streaming budget per frame. 0 = unlimited but could lead to stalling during streaming. High or unlimited budget could lead to stalling during streaming.

Materials

Display Name	Setting Name	Type and Default Value	Description
Disable Material Loading	/app/renderer/skipMaterialLoading	Bool False	Scenes are loaded without materials. This can lower scene loading time.

Texture Compression Settings

Display Name	Setting Name	Type and Default Value	Description
Texture Compression Quality	/rtx-transient/materialdb/blockCompression/quality	Int 0	Fastest: 0, Normal: 1, Production: 2, Highest: 3
Compression Size Threshold	/rtx-transient/resourcemanager/compressionMipSizeThreshold	Int 7	Textures smaller than this size won’t be compressed. 0 disables compression.
Max Resolution	/rtx-transient/resourcemanager/maxMipCount	Int 16	Textures larger than this are downsampled to this resolution. 7 = 64, 8 = 128, 9 = 256, 10 = 512, 11 = 1024, 12 = 2048, 13 = 4096, 14 = 8192, 15 = 16384, 16 = 32768
Normal Map Mip-Map Generation (toggling requires scene reload)	/rtx-transient/resourcemanager/genMipsForNormalMaps	Bool True	Enables mip-map generation for normal maps to reduce memory usage at the expense of quality. Toggling the setting requires reloading the scene for it to be applied.
Normal Map Roughness generation (toggling requires scene reload)	/rtx-transient/resourcemanager/createNormalRoughness	Bool False	Enables roughness generation for normal maps for improved specular reflection with mip mapping enabled.