MAGNETAR is an all-in-one kernel tuning module that applies scenario-aware CPU, GPU, memory, I/O, and process adjustments on supported rooted ARM devices, with behavior described by the project as adapting to device usage.
Scenario-Based System Tuning
A background tuning stack that coordinates several kernel-facing controls rather than applying one fixed performance profile.
CPU and GPU Control
The project includes frequency-scaling, GPU-detection, renderer initialization, and cluster-management logic.
Memory and Scheduling
Version 19.6.0 refines virtual-memory settings, low-memory handling, schedtune configuration, and process prioritization.
Storage-Aware I/O
The current changelog describes scheduler improvements that account for the device's flash-storage behavior.
Requirements
- Android API 21 or newer.
- An ARM or ARM64 device; x86 and x86_64 are rejected by the installer.
- Magisk build 26404 or newer, KernelSU kernel version code 11422 or newer, or APatch version code 10472 or newer.
- Installation from a running root-manager application; recovery installation is explicitly unsupported.
- A kernel that exposes the controls needed by the selected tuning functions. Results can differ substantially between devices and vendor kernels.
Installation and Diagnostics
- Verify the Android version, CPU architecture, and root-manager version against the requirements above.
- Install the MAGNETAR release ZIP through Magisk, KernelSU, or APatch while Android is running.
- Confirm that the installer completes without an architecture or minimum-version error.
- Reboot the device to start the module's daemon and monitoring components.
- Observe thermals, battery drain, application stability, and idle behavior before deciding whether the profile is suitable for the device.
For a reproducible diagnostic package, the project provides the root-terminal command magnetar --bugreport.
Version 19.6.0 Focus
The release changelog highlights extended GPU detection, flash-aware I/O changes, Android 16 and multi-user improvements, daemon-start fixes, CPU scaling corrections, and refinements to LMK, schedtune, rendering, and monitoring. These are implementation goals, not a guarantee that every device will become faster or use less power.