Three OIT techniques implemented in Unity's Built-in Render Pipeline with GPU instancing, command buffers, and compute shaders.
Platform: PC · DirectX 11 / Shader Model 5.0 · Unity 2022.3 LTS
A single-pass approximation. Each fragment writes into a weighted colour accumulation buffer and a revealage buffer simultaneously via MRT. A composite pass resolves them analytically — no sorting, no extra geometry passes.
Four weight functions are available (Weight0–3) to control how much depth influences each fragment's contribution, tunable per scene.
Scene: Assets/Scenes/WBOIT.unity
Two modes implemented:
Each pass peels the single frontmost unresolved layer by clipping against the previous pass's depth texture. After N passes, layers are composited back-to-front.
Peels the frontmost and backmost layers simultaneously using a min/max depth buffer (RGFloat) with BlendOp Max. Halves the pass count needed for the same quality as front-to-back.
Both modes support configurable layer count (1–50) and independent resolution scaling for peel passes and the composite step.
Scene: Assets/Scenes/DepthPeeling.unity
An exact single-pass technique using SM 5.0 UAV writes. Each fragment atomically prepends itself to a per-pixel linked list stored in a structured buffer. A compute shader resets the buffers each frame. The composite shader walks each pixel's list, sorts fragments by depth, and blends back-to-front over the skybox. Supports alpha blend and additive composite modes.
Configurable _maxNodesPerPixel and _resolutionScale let you trade VRAM against quality. At 1080p with 4 nodes per pixel the node buffer is ~126 MB.
Scene: Assets/Scenes/LLOIT.unity
| WBOIT | Depth Peeling (F2B) | Depth Peeling (Dual) | Linked List OIT | |
|---|---|---|---|---|
| Correctness | Approximate | Exact | Exact | Exact |
| Geometry passes | 1 | N (layers) | N/2 (layers) | 1 |
| Sorting | None | Implicit | Implicit | Per-pixel, runtime |
| VRAM overhead | Low | Medium | Medium | High |
| SM requirement | 4.0 | 4.0 | 4.0 | 5.0 |
| Dense overdraw | Degrades gracefully | Capped by _layers |
Capped by _layers/2 |
Capped by _maxNodes |
- Order-Independent Transparency
- Order Independent Transparency with Dual Depth Peeling by nvidia https://developer.download.nvidia.com/SDK/10/opengl/src/dual_depth_peeling/doc/DualDepthPeeling.pdf
- Order Independent Transparency with Per-Pixel Linked Lists https://www.cs.cornell.edu/~bkovacs/resources/TUBudapest-Barta-Pal.pdf
- Weighted Blended Order-Independent Transparency https://jcgt.org/published/0002/02/09/