androidtutorialshow-todeveloper

How to Profile Android Games with APA

Step-by-step guide to install Android Performance Analyzer and capture system traces for games. Analyze CPU, GPU, memory, and power data.

October 17, 2026 Yum News

Android Performance Analyzer (APA) unifies CPU, GPU, memory, and power data into one tool. Game developers can capture system traces without switching between separate profilers. The standalone desktop app works on Windows, macOS, and Linux and targets Vulkan-based titles and other graphics-heavy apps.

APA is currently in open beta. Official documentation states it is intended for profiling games and Vulkan-based graphics. For standard apps, continue using the profilers inside Android Studio.

Developer workspace with laptop showing code

Download and Install APA

Visit the official Android Performance Analyzer page on developer.android.com. Agree to the terms and download the package for your operating system.

On macOS you need an ARM-based Mac running macOS 12 or higher. Intel Macs are not supported. Windows requires 64-bit Windows 10 or later. Linux users must install the required 64-bit libraries listed in the Android Studio install guide.

After installation, set the ANDROID_HOME environment variable so APA can locate Platform-Tools. Connect your test device with USB debugging enabled. The first connection triggers a device validation check. Do not interact with the phone while validation runs.

Prepare Your Game for Accurate Traces

Use a release build or a build with performance compiler flags enabled. For Vulkan games, set android:debuggable=“true” in the manifest so Vulkan-specific data appears in traces. For Java or Kotlin code paths, set debuggable to false so the Android Runtime runs at full optimization. This produces data closer to real-world player performance.

Your device must run Android 12 or higher. Enable USB debugging and, if present, the Install via USB option.

Smartphone on desk next to laptop

Capture a System Trace

Open APA and create a new project. Give it a name and choose a directory. Click Record Trace in the title bar.

The Configure a Recording window opens with default options. Select your validated device from the dropdown. Choose Launch app and record if you want the game to start automatically. Adjust duration, GPU counters, and other filters according to what you need to measure. Click OK.

The Control Recording window appears and the game launches on the device. Play the section you want to profile. Click Stop when finished or wait for the preset duration. APA retrieves the data and opens the trace view.

Analyze the Trace

The trace view shows system-wide activity. Inspect CPU core frequencies and scheduled work. Review GPU counters from Qualcomm, Arm, Imagination, and Samsung hardware. Examine memory and power rails in the same timeline.

Vulkan debug markers appear if you added them in your engine. Split the view to compare two regions of the same trace or open two traces side by side for A/B testing.

APA loads and renders traces up to 26 times faster than the previous GPU Inspector according to the Android team. This speed helps when you iterate on frame timing or thermal behavior.

Tips for Better Results

Profile on the lowest-end device you support. Thermal throttling and memory pressure show up more clearly there. Keep the phone on a stable surface and avoid background apps that can skew power data.

If validation fails, check that adb sees the device, then disconnect and reconnect. Retry the validation button if needed.

For non-Vulkan apps that still need deep system traces, the updated System Trace viewer inside Android Studio Canary (Panda 4 and later) uses the same APA technology. See our earlier note on frame profiling with APA for additional context.

Conclusion

Android Performance Analyzer gives game teams a single place to capture and inspect CPU, GPU, memory, and power data. Download the open beta, validate your device, and record a short trace of the scene that drops frames. The unified timeline usually reveals the bottleneck faster than switching between separate tools.