You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: doc/classes/Environment.xml
-4Lines changed: 0 additions & 4 deletions
Original file line number
Diff line number
Diff line change
@@ -220,10 +220,6 @@
220
220
Number of ray samples cast per pixel per frame. Higher values reduce noise but increase GPU cost proportionally. When using a temporal denoiser such as DLSS Ray Reconstruction, a low sample count (1-2) is typically sufficient since the denoiser accumulates information across frames.
221
221
[b]Note:[/b] Only has an effect when [member pathtracing_enabled] is [code]true[/code].
Copy file name to clipboardExpand all lines: doc/classes/ProjectSettings.xml
+10-6Lines changed: 10 additions & 6 deletions
Original file line number
Diff line number
Diff line change
@@ -3245,30 +3245,34 @@
3245
3245
[b]Note:[/b] Enabling occlusion culling has a cost on the CPU. Only enable occlusion culling if you actually plan to use it. Large open scenes with few or no objects blocking the view will generally not benefit much from occlusion culling. Large open scenes generally benefit more from mesh LOD and visibility ranges ([member GeometryInstance3D.visibility_range_begin] and [member GeometryInstance3D.visibility_range_end]) compared to occlusion culling.
3246
3246
[b]Note:[/b] Due to memory constraints, occlusion culling is not supported by default in Web export templates. It can be enabled by compiling custom Web export templates with [code]module_raycast_enabled=yes[/code].
If [code]true[/code], path tracing ray hit shaders are compiled asynchronously in the background. This avoids frame stalls when new shader variants are first encountered during rendering. While a shader is compiling, a fallback shader may be used, which can cause temporary visual artifacts. Disable this if you require fully deterministic rendering from the first frame at the cost of potential stalls on shader warm-up.
Number of frames a skinned or vertex-displaced deformed-mesh BLAS is kept alive in the path tracer cache after its last visible frame. Once this threshold is exceeded the BLAS and its associated buffers are freed. Increase this value if deformed meshes flicker back into view after brief occlusion; decrease it to reclaim GPU memory sooner.
Number of frames a merged MultiMesh BLAS is kept alive in the path tracer cache after its last visible frame. At 60 fps, the default of 3600 corresponds to approximately one minute. Once this threshold is exceeded the BLAS and all merged vertex, attribute, and index buffers are freed. Decrease this value to reclaim GPU memory when MultiMeshes disappear from the scene.
When enabled, setting [member MultiMesh.buffer] keeps a CPU-side mirror of the instance transform data in sync with the GPU buffer. This allows the path tracer to read per-instance transforms without performing a GPU-to-CPU readback on first access, eliminating the one-time stall that occurs otherwise.
3261
3261
[b]Warning:[/b] Enabling this setting increases CPU memory usage. Every [MultiMesh] that assigns [member MultiMesh.buffer] will maintain a full duplicate of its transform data in RAM.
3262
3262
Enable this only when [MultiMesh] nodes with frequently updated transforms are present in path-traced scenes and the GPU readback stall is measurable.
Maximum total triangle count ([code]instance_count * triangles_per_surface[/code]) for a [MultiMesh] surface to use a single merged BLAS in the path tracer. When the combined triangle count exceeds this value, the path tracer falls back to one TLAS entry per instance sharing a common BLAS instead. Lower this value to reduce peak GPU memory for large MultiMeshes; raise it to prefer the merged path (which avoids per-instance TLAS overhead).
If [code]true[/code], enables NVIDIA Shader Execution Reordering in the path tracer. SET reorders ray tracing shader invocations at runtime to improve GPU occupancy and reduce divergence between threads executing different materials. This can significantly improve path tracing performance on NVIDIA Ada Lovelace and later GPUs. Has no effect on hardware that does not support SET.
If [code]true[/code], shadow rays use Ray Queries instead of fully traced rays. Ray Queries are significantly cheaper but rely on a quick alpha texture fetch, which can produce incorrect shadows on materials with generated/animated alpha textures. Fully traced shadow rays are more accurate but incur a higher GPU cost.
3274
+
[b]Note:[/b] Only has an effect when path tracing is enabled on the active [Environment].
Number of cubemaps to store in the reflection atlas. The number of [ReflectionProbe]s in a scene will be limited by this amount. A higher number requires more VRAM.
If [code]true[/code], adds [code]hint_alpha[/code] to the uniform declaration. This marks the texture's alpha channel as the source of opacity for path tracing ray query alpha tests (simple shadows and Opacity Micromap baking). Can be combined with any [member texture_type], for example [constant TYPE_COLOR] for a color-plus-alpha albedo texture or [constant TYPE_DATA] for a dedicated grayscale alpha mask.
0 commit comments