meowrch-settings

This repository contains a set of system configurations. Some of them are inspired by CachyOS.

These settings are designed to improve system performance, speed, and resource management.

More about optimizations

Udev Rules: Device Event Automation

Udev rules automatically apply system configurations upon device detection or state changes.

• Audio Power Management: Manages snd-hda-intel power saving to mitigate audio crackling, disabling it when AC-powered and re-enabling on battery.
• ZRAM Swap Optimization: Configures ZRAM to prefer anonymous page compression (vm.swappiness=150) and disables Zswap for efficient RAM-based swap.
• Device Permissions: Sets rtc0 and hpet device group to "audio" for proper application access.
• SATA Performance: Configures SATA host link power management to max_performance.
• I/O Scheduler Assignment: Dynamically assigns optimal I/O schedulers: bfq for HDDs, mq-deadline for SATA SSDs, and none for NVMe SSDs.
• HDD Performance Tuning: Applies hdparm settings (-B 254 -S 0) to rotational disks.
• NVIDIA Runtime Power Management: Enables/disables NVIDIA GPU runtime power management on driver bind/unbind events.
• CPU DMA Latency Access: Sets permissions for the cpu_dma_latency device.

Sysctl: Kernel Runtime Configuration

Sysctl parameters modify kernel behavior at runtime for system-wide performance and stability.

• Memory & I/O Management: Adjusts vm.swappiness, vfs_cache_pressure, dirty_bytes, dirty_background_bytes, and dirty_writeback_centisecs for balanced memory usage and efficient disk I/O. Disables vm.page-cluster.
• System Stability & Security: Disables kernel.nmi_watchdog, enables kernel.unprivileged_userns_clone, restricts kernel.kptr_restrict, and disables kernel.kexec_load_disabled.
• Logging & Network: Configures kernel.printk to hide messages from console, increases net.core.netdev_max_backlog, and sets fs.file-max.

Modprobe: Kernel Module Parameters

Modprobe configurations control module loading and behavior for hardware-specific optimizations.

• AMD GPU Driver Enforcement: Forces amdgpu driver for GCN 1.0+ and 2.x GPUs, blacklisting radeon.
• Watchdog Module Blacklist: Prevents loading of Intel TCO and AMD SP5100 watchdog timers.
• NVIDIA Driver Optimizations: Applies parameters like NVreg_UsePageAttributeTable=1 (PAT for CPU performance), NVreg_InitializeSystemMemoryAllocations=0 (disables memory clearing for GPU), NVreg_DynamicPowerManagement=0x02 (mobile GPU power saving), and NVreg_RegistryDwords=RMIntrLockingMode=1 (frame-pacing).

Systemd: Service & System Management

Systemd unit and configuration files for streamlined boot, resource management, and service control.

• Journal Log Limits: Sets journald size limit to 50MB.
• Service Timeouts: Defines DefaultTimeoutStartSec (15s) and DefaultTimeoutStopSec (10s) for services.
• File Descriptor Limits: Increases DefaultLimitNOFILE for both system (2048:2097152) and user (1024:1048576) services.
• Time Synchronization: Configures systemd-timesyncd with Cloudflare and Google NTP servers.
• ZRAM Generator: Configures ZRAM with zstd or lz4 compression, ram size, and swap-priority=100.
• PCI Latency Service: Enables a systemd service to apply pci-latency script at boot.
• User Service Resource Delegation: Delegates CPU, cpuset, IO, memory, and pids to user services.

Tmpfiles: Temporary File & THP Management

Configurations for temporary file cleanup and Transparent Huge Page (THP) behavior.

• Coredump Retention: Clears coredumps older than 3 days.
• THP Defragmentation: Sets transparent_hugepage/defrag to defer+madvise for tcmalloc-using applications.
• THP Shrinker: Configures khugepaged/max_ptes_none for Kernel 6.12+ to optimize THP memory usage.

Runtime Optimization Scripts

Boot-time scripts in /usr/bin/ that apply dynamic optimizations requiring hardware detection or runtime adjustments.

meowrch-apply-optimizations

Purpose: Applies runtime optimizations that cannot be configured via static files.
Systemd Service: meowrch-optimizations.service (enabled at boot)

Optimizations Applied:

• CPU Governor Management: Automatically sets performance governor for all CPU cores when available, ensuring maximum clock speeds and minimal frequency scaling latency.
• Intelligent I/O Scheduler Selection:
  • SSDs/NVMe: Assigns none scheduler (bypasses kernel I/O scheduling, direct-to-hardware)
  • HDDs: Assigns mq-deadline scheduler (optimizes for rotational media seek patterns)
  • Detection via /sys/block/*/queue/rotational attribute
• Transparent Huge Pages (THP) Tuning:
  • Sets THP to madvise mode (enabled only for applications explicitly requesting it)
  • Configures THP defrag to defer (reduces memory allocation stalls)
• Network Packet Steering:
  • Enables RPS (Receive Packet Steering) with ff mask for all network interfaces
  • Distributes network interrupt processing across all available CPU cores
• GPU Performance Modes:
  • NVIDIA: Activates persistence mode (nvidia-smi -pm 1) and locks maximum GPU/memory clocks
  • AMD: Forces manual power profile and highest performance level via /sys/class/drm/card0/device/ sysfs interface

pci-latency

Purpose: Optimizes PCI device latency timers to reduce audio crackling and improve peripheral responsiveness.
Systemd Service: pci-latency.service (enabled at boot)

Configuration Applied:

• Default PCI Devices: Sets latency timer to 20 cycles (balances throughput and latency)
• Root Complex (0:0): Resets latency timer to 0 (immediate data transfer)
• Audio Controllers (PCI class 04xx): Sets latency timer to 80 cycles (prevents audio buffer underruns and glitches)

Uses setpci to directly modify PCI configuration space at boot.

Systemd Service Integration

Auto-start Units:

• meowrch-optimizations.service - Executes /usr/bin/meowrch-apply-optimizations after sysinit.target
• pci-latency.service - Executes /usr/bin/pci-latency before graphical/multi-user target

Both services run once at boot with Type=oneshot and RemainAfterExit=yes to maintain optimization state.

earlyoom: Early Out-Of-Memory Daemon

Purpose: Prevents complete system freezes by proactively killing memory-hungry processes before OOM killer activation.
Configuration: /etc/default/earlyoom-meowrch

Protection Strategy:

• Memory Threshold: Triggers at 5% free RAM (-m 5)
• Swap Threshold: Triggers at 10% free swap (-s 10)
• Protected Processes: System-critical daemons (init, systemd, Xorg, ssh, NetworkManager, pipewire, wireplumber)
• Kill Candidates: Memory-intensive applications prioritized (firefox, chromium, electron, code, discord, telegram, steam)
• Testing Mode: Includes --dryrun flag by default (remove to enable actual process termination)

Automatically enabled via earlyoom.service at boot, ensuring system remains responsive during memory pressure scenarios.

ananicy-cpp: Automatic Nice Daemon

Purpose: Dynamically adjusts process priorities (nice levels, I/O class, OOM scores) based on predefined rules for optimal responsiveness.
Rules: cachyos-ananicy-rules - CachyOS-curated ruleset for gaming, desktop, and server workloads

Key Features:

• Game Boost: Automatically prioritizes game processes with high CPU/IO priority
• Background Tasks: De-prioritizes compilers, package managers, and system maintenance
• Desktop Responsiveness: Keeps GUI applications responsive under heavy background load
• OOM Score Adjustment: Protects important processes from being killed first

NVIDIA Hyprland VRAM Fix

/etc/nvidia/nvidia-application-profiles-rc.d/50-limit-free-buffer-pool.json

Saves 2-4GB VRAM on Wayland by disabling excessive buffer pool GLVidHeapReuseRatio=0 - NOT fixed in drivers 575+ (Jan 2026)

Refs

• github.com/hyprwm/Hyprland/issues/7704
• reddit.com/r/hyprland/comments/1m71ukt

🙏 Acknowledgements

Optimizations are based on projects:

• CachyOS - for proven performance settings
• ARU - Arch Linux optimization guide
• Arch Linux community - for best practices in system configuration