Textures & lighting
Shadow Resolution & World-Texel Estimator
Direct answer
World-space shadow detail is coverage divided by shadow-map resolution. A 2048² map covering 100 m gives 4.88 cm per shadow texel, so a one-metre feature spans about 20.5 texels. Four 32-bit cascades at that resolution require roughly 64 MiB before filtering, page tables or renderer overhead.
Result
- World-space shadow texel
- —cm
- Entered feature spans
- —texels
- Resolution required for target
- —px
- Next power-of-two resolution
- —px
- Approximate allocated map memory
- —MiB
- Total allocated shadow texels
- —million
- Target detail check
- —
Method & assumptions
How this is worked out
World texel size is coverage width / map resolution. The number of texels across a feature is feature size divided by that world texel size.
Required resolution rearranges the first equation: coverage / target texel size, then rounds up to a power of two for a conventional map allocation.
Memory multiplies resolution squared, bytes per texel, cascade count and equivalent shadowed lights. It excludes cube-map faces, filtering buffers, virtual-page tables, cache residency, mip levels and implementation-specific compression.
What the estimate does and does not model
| Included | Not included | Reason |
|---|---|---|
| Map resolution | Projection warping | Assumes even orthographic coverage |
| World coverage | Perspective distribution | Reports average linear density |
| Map count | Virtual page residency | Conventional allocation estimate |
| Bytes per texel | Filter and cache buffers | Renderer-specific overhead varies |
Virtual Shadow Maps distribute resolution dynamically and cache pages, so use this calculator as a world-detail comparison, not as an exact VSM residency predictor.
Questions
How many centimetres does one shadow texel cover?
Divide world coverage in centimetres by shadow-map resolution. A 100 m span is 10,000 cm; divided by 2048 it gives about 4.88 cm per texel.
Why do distant cascades look softer?
They cover more world space with a similar map resolution, so each texel represents a larger area. Cascade distribution, filtering and bias further affect the visible result.
Does this predict Virtual Shadow Map memory exactly?
No. Virtual Shadow Maps allocate and cache pages based on visibility and demand. The calculator provides comparable world-space density and a conventional-map memory baseline, not exact VSM page residency.
Sources and assumptions
Sources define the engine settings or mathematical model. Project-specific performance and final playability still need measurement in the target build.