vk_lod_clusters --help prints all registered options with their descriptions, this page only highlights the
noteworthy ones. Options can also be put into a .cfg file and passed with --configfile <file.cfg>.
The scene_overrides directory next to shaders holds per-scene settings that ship with the sample instead of
with the model. Loading <name>.gltf/.glb/.cfg also runs scene_overrides/<name>.cfg if it exists, after the
scene's own .cfg and before the scene is loaded, so options that affect loading take effect. An override cannot
change --scene.
--renderer 0starts with rasterization.--supersample 0disables the super sampling that otherwise doubles rendering resolution in each dimension.--clasallocator 0disables the more complex gpu-driven allocator when streaming and uses the simple move based CLAS compaction instead. It moves all resident CLAS whenever a group is unloaded, which makes--blascachingineffective while it is used, see BLAS Caching.--gridcopies Nset the number of model copies in the scene.--gridunique 0disables the generation of unique geometries for every model copy. Greatly reduces memory consumption by truly instancing everything. It's on for the sample bunny scene by default, but off otherwise.--streaming 0disables streaming system and uses preloaded scene (warning this can use a lot of memory, use above--gridunique 0to reduce)--vsync 0disable vsync. If changing vsync via UI does not work, try to use the driver's NVIDIA Control Panel and setVulkan/OpenGL present method: native.--autoloadcache 0disables loading scenes from cache file.--mappedcache 1keeps memory mapped cache file persistently, otherwise loads cache to system memory. Useful to save RAM on very large scenes.--autosavecache 0disables saving the cache file.--meshoptarena 0disables the per-thread stack arena that servesmeshoptimizer's temporary allocations during cluster building. The arena is on by default; it removes most of the global allocator contention that otherwise limits how well lod processing scales with thread count.--meshoptarenabudget <MiB>caps how much each thread keeps between calls (default 8), below which the arena starts thrashing on chunk churn.--forcepreprocessmegabytes 1024if a scene's raw geometry (vertex & indices) is greater than this cutoff, use a dedicated preprocess pass. Can be quicker and allows using memory mapped cache file. Default is 2048 for 2 GiB.--multimaterials 1 --attributes 7 --texturedmaterials 1enables textured PBR materials (see Materials below).--maxtexturemegabytes <MiB>sets an upper VRAM budget for material textures (default 4096; 0 = no limit). See Materials.- All megabyte budgets (
--maxgeomegabytes,--maxclasmegabytes,--startclasmegabytes,--clasgrowmegabytes,--maxblascachingmegabytes,--maxtransfermegabytes,--maxtexturemegabytes) also accept a negative value, which is interpreted as a percentage of the device local heap. For example--maxgeomegabytes -10 --maxclasmegabytes -10reserves 10 % of VRAM each, independent of the GPU in use. --dlss 1enables DLSS when built withUSE_DLSS(Super Resolution in rasterization, denoising in ray tracing). Use--dlssquality <0-3>to set quality (max performance through ultra performance).--camerastring "..."sets the initial camera (copy/paste from the Misc Settings → Camera widget).--addcamerapath "..."defines a camera fly-through path (repeatable, see Camera Path).--loadcamerapaths "file.txt"replaces all paths with the definitions from a text file (see Camera Path).--runcamerapath <index> <framecount>deterministically plays back a defined path for benchmarking (see Camera Path).