Exploring instancing is essential for your specific use case. It is designed to handle scenarios like yours.
Alternatively, opting for BufferGeometry instead of traditional geometry can help reduce memory consumption.
edit
The memory consumption issue is primarily due to the overhead of the merge operation when working with THREE.Geometry. This is attributed to the extensive allocation of JS objects such as Vector3, Vector2, Face3, etc., which eventually get discarded because the original geometry is no longer present. This situation causes strain on the system, leading to potential performance degradation even if no crashes occur. Buffer geometry proves to be more efficient because it utilizes typed arrays, ensuring that only primitives are copied around, and there is no need for object allocation.
Nevertheless, it results in increased GPU memory usage since multiple instances are stored in the same buffer, causing repetition and pre-transformation of data. This is where instancing can offer assistance by optimizing the process.
Mesh (Node, Object)
A Mesh describes a 3D object within a scene graph, detailing its parent-child relationship, position, rotation, scale, geometry, material, and other attributes.
Geometry
This component holds the fundamental geometry data, including vertices, UVs, normals, etc. It serves as the foundation for modeling programs, potentially creating ambiguity.
It is crucial to grasp that these elements exist in "object (model) space," which entails initial translation, rotation, and scale set by the modeler. This allows the object to be placed and manipulated within a 3D environment effectively.
To address the issue mentioned in the title, one must understand the concept of instancing and streamline the rendering process to focus on individual objects instead of processing an entire forest model.
By utilizing instancing, draw calls can be minimized while rendering tree models singularly rather than collectively as a forest, leading to improved performance and reduced memory overhead.
/edit
In scenarios with a high volume of geometries like 100k, optimizing draw calls becomes crucial for enhancing performance. It is vital to leverage techniques such as instanced rendering and efficient buffer management to address performance bottlenecks associated with handling large numbers of geometries.
The game experiences a significant memory saving transformation, but the performance is hindered by the multitude of draw calls linked to the 100k geometries.
The challenge lies in managing the overhead generated by a large number of draw calls rather than the individual geometries. Various strategies can be employed to overcome this performance issue effectively.