Does a Drop in CPU Temperature and Power Consumption During Long Multi-Player Combat Mean the CPU Is Throttling?
In a long duration multi player combat test, CPU temperature is noticeably higher than GPU temperature, and device power consumption stays at a relatively high level. However, as the combat progresses, temperature and power consumption instead decrease in the later stage, and the frequency of some CPU cores also fluctuates. Does this mean the CPU is throttling, and how should this be determined? https://uwa overseas public.oss us east 1.aliyuncs.com/uploads/markdown/35843ca7 ea91 4338 a437 360336779d4d.png
A decrease in temperature and power consumption may be caused either by a reduction in workload or by thermal throttling. It cannot be determined from curve changes alone; CPU frequency, the time cost of each module, and the actual game logic need to be analyzed together. https://uwa overseas public.oss us east 1.aliyuncs.com/uploads/markdown/2093348f afb8 4a9b bae5 ca5554767693.png In the test, CPU temperature on some devices was more than ten degrees higher than GPU temperature, and device power consumption was approximately 5W. In the later stage, temperature and power consumption decreased, and the frequency of some CPU cores fluctuated. It is recommended to investigate from the following three aspects: 1. Compare CPU frequency with FPS https://uwa overseas public.oss us east 1.aliyuncs.com/uploads/markdown/a2c47d4a ba69 4296 9d8d 4d771079eae5.png Focus on the CPU core that carries the game's main thread tasks, and check whether its frequency continuously decreases and whether FPS shows an obvious change over the same period. Note that frequency fluctuation on a single core does not necessarily indicate thermal throttling — it may also be related to normal frequency scaling or thread scheduling. 2. Compare module time cost under similar workload CPU throttling may affect the execution efficiency of multiple CPU side modules. When business conditions such as UI update frequency and particle effect scale are similar, the CPU time cost of modules such as the UI and particle systems can be compared. If the time cost of multiple modules increases noticeably, whether thermal or power limits exist should be further determined together with the device's thermal state and thread scheduling. 3. Check for workload changes during combat A decrease in temperature and power consumption may also result from an actual reduction in computation, so it is necessary to confirm whether the combat state changed before and after. For example, some combat gameplay includes a wounded unit mechanism; as combat progresses, some units no longer participate, so the computation for logic such as pathfinding and AI decision making decreases accordingly, and CPU load may also drop. It is recommended to compare the number of participating units and the related function call counts before and after the temperature drop, to confirm whether there is an obvious change in workload. Overall, a decrease in temperature and power consumption cannot be directly attributed to CPU throttling. CPU frequency, FPS, and module time cost under similar workload should be combined to distinguish the impact of limited hardware performance from a reduction in business computation. If necessary, this can be further verified through a controlled test under the same workload.