Unity Rendering and Performance Optimization for the VI Sales Crown Tool

Background Unity optimization covers a broad knowledge base; as the VI Sales Crown tool iterated, various performance metrics showed deviations, prompting a systematic optimization effort.

1. Rendering Process The article explains the three main stages of the rendering pipeline—Application, Geometry, and Rasterization—detailing the responsibilities of CPU, GPU, and memory at each stage.

2. Terminology

Drawcall : a command issued by the CPU to the GPU that triggers rendering operations.

Batching : merging multiple draw calls (static or dynamic) to reduce CPU‑GPU communication.

Atlas : combining many small 2D textures into a single large texture to improve batching and lower rendering cost.

3. Optimization Overview After understanding CPU/GPU/memory responsibilities, each module can be targeted for optimization. The article lists typical optimization categories for CPU (rendering, animation, physics, UI, particle, I/O, GC, scene loading, code), GPU (fill rate, pixel/geometry complexity, memory bandwidth, LOD/occlusion), and memory (model/texture size and format, network resources, managed heap, allocation patterns).

Module

Optimization Points

CPU

Reduce draw calls, batch, compress resources, optimize animation, physics, UI, particles, I/O, garbage collection, scene loading, and code execution time.

GPU

Lower fill rate, pixel complexity, geometry complexity, memory bandwidth usage, reduce vertex count, compress textures, use mipmaps, apply LOD and occlusion culling.

Memory

Control model polygon count, texture size/format, avoid large one‑time allocations, reduce overall memory footprint.

Other

AssetBundle management, texture compression, heat reduction, high‑cost code paths, decoupling modules.

4. Tools & Methodology

Unity Profiler, UPR (official profiler), UWA (third‑party profiler), Memprofile.

Mobile/PC debugging tools: Mali Graphic Debugger, Snapdragon Profiler, ADB & Android Studio, Xcode GPU Frame Capture.

Methodology includes quantitative measurement before/after optimization, isolation of suspect code, comparative analysis, early specification of target devices, and leveraging experience to prioritize functional iteration before performance tuning.

5. VI Sales Crown Product Performance Optimization

5.1 AR Optimization Issues such as low recognition rate, large model files, and platform‑specific code coupling were addressed by consolidating scenes, separating Android/iOS logic, asynchronous loading, and retraining a reduced‑size model, resulting in a 28 MB package reduction and higher recognition stability.

5.2 Texture Optimization Large, duplicated textures were replaced with a new shader and smaller images, improving visual quality while lowering rendering pressure and memory usage by about 200 MB.

5.3 Memory Optimization By decoupling module dependencies, staggering initialization, using smaller textures, and providing mesh‑reduction guidance, memory peaks were lowered and total memory consumption decreased by roughly 200 MB.

5.4 Heat Optimization Frame rate was capped at 45 fps, draw calls reduced via UI atlasing, and textures compressed, resulting in CPU temperature dropping from 65 °C to 55 °C and delayed heat onset.

5.5 Package Size Optimization Unused assets were removed, Vuforia model size reduced, and remote asset delivery introduced, shrinking the APK from 210 MB to 80 MB.

6. Conclusion The author reflects that optimization and architecture are core competencies for developers; the VI Sales Crown project demonstrates measurable gains in package size, memory, heat, and AR performance, and the experience informs future work on holographic projects.

Author Bio Hu Chunyuan, member of the Innovation Project Team at Autohome, responsible for architecture and optimization of the VI Sales Crown and holographic warehouse projects.