Android XR Performance
App performance is critical for developers, particularly in AR/MR, where poor performance can significantly impact user comfort.
Our SDK is designed to be resource-efficient while delivering robust analytics. We are continually optimizing to ensure it remains both lightweight and effective. For specific requirements, please contact us, and our customer success team will assist you.
Overall, the performance impact of our SDK is minimal compared to the demands of rendering and MR/AR interaction systems. Recent performance test results are provided below.
Best Practices
- Web requests can have a notable performance impact. If necessary, increase automatic_send_timer in
cognitive3d.jsonto send fewer, larger requests. - Initializing a session should be done when the user is least likely to notice it - immediately when the application begins or during a scene load.
Performance Testing Overview
To provide comprehensive performance insights, we tested our SDK using a purpose-built Android XR application that exercises key platform features, including spatial UI panels and hand-anchored objects placed on the user's palms. This test app represents a typical Android XR experience, giving you a clear understanding of the SDK's performance impact in real-world conditions.
Test Device
All tests were conducted on the following device:
Samsung Galaxy XR
- CPU: Qualcomm Snapdragon XR2+ Gen 2
- RAM: 16 GB
- GPU: Adreno 740
Methodology
All performance values were recorded using a custom lightweight performance profiler for Android devices, designed with minimal overhead to ensure accurate FPS and CPU usage readings.
For each test, we executed the app twice - once without the SDK to establish baseline performance, and once with our SDK included to determine the SDK's overhead.
Test Conditions
- Date: February 2026
- Android XR SDK version:
1.0.2 - Android Studio version:
2025.2.1 Patch 1 - App:
Android XR Sample
Test Setup
The SDK implementation included:
- Gaze tracking
- Hands interactions
- 4 dynamic objects (including hands)
- 1 sensor value being written per frame (in addition to our default ones)
- 1 custom event being written per second (in addition to our default ones)
Note
Recording a custom sensor every frame and a custom event every second far exceeds typical usage. These aggressive settings were intentional to stress-test the SDK as a worst-case scenario; real-world integrations would see even lower overhead. The CPU usage increase is largely driven by these stress-test conditions. Typical integrations with standard event and sensor frequencies will see significantly less impact.
Test Results
| App without Cognitive3D | App with Cognitive3D | Relative % difference | |
|---|---|---|---|
| Average framerate | 71.58 fps | 71.38 fps | -0.28% |
| CPU usage (%) | 60.54% | 64.01% | +5.73% (+3.47 percentage points) |
If you have a question or any feedback about our documentation please use the Intercom button (purple circle) in the lower right corner of any web page or join our Discord.