Is UI Particle the Main Overhead When UI Time Cost Is Much Higher on Low-End Devices? How to Investigate Further?
In performance tests across devices of different tiers, the GameOptim GOT Online report shows that average UI time cost on high end devices is approximately 2.66 ms, while it reaches 8 ms on low end devices, and some intervals even consistently exceeded 30 ms. Upon further inspection of the function stack, it was found that nodes such as ParticleSystem.Simulate are called frequently, and the call count is relatively consistent with variations in UI time cost. Does this mean UI Particle is the main source of overhead, and how should this be investigated further? https://uwa overseas public.oss us east 1.aliyuncs.com/uploads/markdown/c0b0dd41 77f4 4c7f 98e7 55de649a16f4.png
From the current stack data in the GameOptim GOT Online report, UI Particle is the part that requires priority inspection. The call count of ParticleSystem.Simulate is consistent with the variation in UI time cost, but the actual time cost proportion still needs to be checked to determine its impact on overall UI overhead. https://uwa overseas public.oss us east 1.aliyuncs.com/uploads/markdown/518aadb1 2fe3 43fb 8d56 6210bc8fdc6e.png Combined with the test footage, some time cost peaks appeared when the upgrade button was clicked, corresponding to the flashing particle effects in the interface. In the main city scene in the later stage, UI time cost remained at a relatively high level. UI Particle is also used in positions such as activity entrances and button glow in the project. It is recommended to investigate from the following two aspects: 1. Check the update overhead of persistent UI effects Effects that play briefly on click mainly affect the trigger frame, while effects that run continuously in the main city such as button glow may continuously generate particle simulation overhead. It is recommended to compare the call count and time cost of ParticleSystem.Simulate, and check which effects are still updating even when there is no user interaction. If UI Particle dynamically updates meshes through CanvasRenderer, also pay attention to related nodes such as Canvas.BuildBatch and Canvas.SendWillRenderCanvases, to confirm whether extra Canvas rebuilding overhead exists. For effects that update frequently, consider using an independent sub Canvas to separate dynamic effects from static UI, and compare batches and CPU time cost before and after the split. https://uwa overseas public.oss us east 1.aliyuncs.com/uploads/markdown/7ad44211 55ec 4683 8620 49e0eb7d9580.png 2. Adjust the effect presentation for low end devices For simple effects such as button flash and glow, test UI Shader solutions using sequential frames or UV scrolling to reduce particle simulation overhead; for other effects, control the quantity by device tier to avoid running too many UI Particle instances simultaneously on low end devices. When replacing, also pay attention to alpha blending, GPU time cost, and memory changes — do not focus only on CPU overhead. Finally, compare UI module time cost, ParticleSystem.Simulate time cost, and FPS under the same interface and operation workflow to confirm whether the adjustment is effective.