COMPARISON OF THE PERFORMANCE OF GAMEOBJECT AND ECS ARCHITECTURES IN MODELING A CROWD OF N AGENTS IN UNITY

Ishchenko Nikita Nikolaevich

The paper presents a comparative performance analysis of two approaches in Unity: the classical GameObject/MonoBehaviour architecture and the new Entity Component System (ECS). For validation, a crowd simulation based on the well-known Boids algorithm, widely used for modeling collective behavior, was implemented under both paradigms. Experiments were conducted on two hardware platforms: a PC (Intel Core i5-12400F, NVIDIA RTX 3070) and a MacBook Pro (Apple M4). For each implementation, performance metrics were gathered, including FPS, frame time, memory usage, and the number of draw calls. The collected data made it possible to analyze differences in scalability and efficiency between the architectures. The results demonstrated that ECS achieves significantly higher frame rates and stability as the number of agents increases, which confirms its efficiency for tasks requiring real-time performance and large-scale simulations.

Keywords:Unity, Data-Oriented Tech Stack, DOTS, GameObject, Entity Component System, parallel computing, performance, rendering, GPU instancing, optimization.