Before getting on with the recipes, let's step back and think about the different parts of Unity games and how their construction and runtime behavior can impact on game performance.

Games are made up of several different kinds of components:

When a game is running, there are many competing processing requirements for your CPU and GPU, including:

One way to reduce the complexity of graphical computations and to improve frame rates is to use simpler models whenever possible—this is the reduction of the Level Of Detail (LOD). The general strategy is to identify situations where a simpler model will not degrade the user's experience. Typically, situations include where a model is only taking up a small part of the screen (so less detail in the model will not change what the user sees), when objects are moving very fast across the screen (so the user is unlikely to have time to notice less detail), or where we are sure the users' visual focus is elsewhere (for example, in a car racing game, the user is not looking at the quality of the trees but on the road ahead). We provide a LOD recipe, Improving performance with LOD groups, in this chapter.

Unity's draw call batching may actually be more efficient than you or your team's 3D modelers are at reducing the triangle/vertex geometry. So, it may be that by manually simplifying a 3D model, you have removed Unity's opportunity to apply its highly effective vertex reduction algorithms; then, the geometric complexity may be larger for a small model than for a larger model, and so a smaller model may lead to a lower game performance! One recipe presents advice collected from several sources and the location of tools to assist in different strategies to try to reduce draw calls and improve graphical performance.

We will present several recipes allowing you to analyze actual processing times and frame rates, so that you can collect data to confirm whether your design decisions are having the desired efficiency improvements.

At the end of the day, the best balance of heuristic strategies for your particular game project can only be discovered by an investment of time and hard work, and some form of profiling investigation. Certain strategies (such as caching to reduce component reflection lookups) should perhaps be standard practice in all projects, while other strategies may require tweaking for each unique game and level, to find which approaches work effectively to improve efficiency, frame rates, and, most importantly, the user experience when playing the game.

Perhaps, the core strategy to take away from this chapter is that there are many parts of a game that are candidates for possible optimization, and you should drive the actual optimizations you finally implement for a particular game based on the evidence you gain by profiling its performance.