Intro
Let’s not go for “every single knob” (many are irrelevant or hardware-specific), but instead create a focused, high-impact sysctl configuration that truly pushes Linux to its performance limits.
Below is your “Insane Performance Mode” of sysctl.conf, written like a professional tuning template with explanations inline.
This configuration is designed for benchmarking, gaming, and raw throughput testing — not for production or secure environments.
Insane Performance Mode – sysctl.conf
Purpose: Maximize CPU, memory, and I/O performance on Linux by reducing latency, disabling safety checks, and optimizing kernel scheduling and networking.
Warning: Reduces system security and stability. Use only for isolated performance testing or gaming rigs.
CPU & Scheduler Tweaks
# Reduce scheduler latency and favor performance over fairness
kernel.sched_latency_ns = 6000000
kernel.sched_min_granularity_ns = 750000
kernel.sched_wakeup_granularity_ns = 1000000
# Enable aggressive CPU utilization (disable automatic balancing delays)
kernel.sched_migration_cost_ns = 50000
# Disable automatic NUMA balancing (may improve consistency on gaming rigs)
kernel.numa_balancing = 0
# Disable scheduler autogrouping (less latency under load)
kernel.sched_autogroup_enabled = 0
This configuration tunes the Linux CPU scheduler to prioritize lower latency and higher performance. It reduces scheduling delays and time slices so tasks can switch faster and respond more quickly under load. NUMA auto-balancing and scheduler autogrouping are disabled to avoid unpredictable task movement and grouping behavior, which can improve consistency on performance-focused systems such as gaming machines.
Virtual Memory & Cache Optimization
# Reduce swap use (keep processes in RAM)
vm.swappiness = 1
# Minimize dirty page flushing latency for lower write stalls
vm.dirty_ratio = 10
vm.dirty_background_ratio = 5
# Force kernel to drop caches more aggressively
vm.vfs_cache_pressure = 200
# Disable zone reclaim (important for single-node performance)
vm.zone_reclaim_mode = 0
# Set higher minimum free memory to reduce allocation latency
vm.min_free_kbytes = 65536
This configuration tunes Linux memory management to favor RAM usage and reduce latency. It minimizes swap activity, keeps frequently used data in memory, and flushes dirty pages more frequently to avoid large write stalls. The settings also increase cache turnover and maintain a larger pool of free memory, helping the system respond faster during heavy workloads.
Networking & Throughput
# Maximize TCP buffer sizes
net.core.rmem_max = 268435456
net.core.wmem_max = 268435456
net.ipv4.tcp_rmem = 4096 87380 268435456
net.ipv4.tcp_wmem = 4096 65536 268435456
# Enable window scaling, timestamps off for lower latency
net.ipv4.tcp_window_scaling = 1
net.ipv4.tcp_timestamps = 0
# Use more aggressive TCP congestion control
net.ipv4.tcp_congestion_control = bbr
# Disable slow start after idle
net.ipv4.tcp_slow_start_after_idle = 0
# Reduce latency in packet queuing
net.core.netdev_max_backlog = 100000
net.core.somaxconn = 8192
# Disable TCP memory pressure limits
net.ipv4.tcp_no_metrics_save = 1
This configuration optimizes the Linux networking stack for high throughput and low latency. It increases TCP buffer limits to allow larger data transfers, enables modern congestion control with BBR, and reduces delays caused by slow-start and packet queuing. Together, these settings help maintain faster and more consistent network performance, especially on high-speed connections.
I/O & Filesystem
# Improve I/O throughput
fs.aio-max-nr = 1048576
# Reduce inode cache expiration latency
fs.inotify.max_user_watches = 1048576
fs.inotify.max_user_instances = 1024
# Enable fast file descriptor recycling
fs.file-max = 2097152
# Disable access time updates on file reads (also use noatime in fstab)
vm.block_dump = 0
This configuration improves filesystem responsiveness and scalability. It raises limits for asynchronous I/O, file descriptors, and inotify watchers so the system can handle many files and events without hitting kernel limits. These tweaks are useful for workloads that monitor many files or perform heavy disk operations, such as development environments, containers, or large applications.
Logging & Debug Overhead
# Minimize printk noise
kernel.printk = 3 3 3 3
# Disable kernel log rate limiting (for benchmarks)
kernel.printk_ratelimit = 0
# Disable debug exception handlers
kernel.perf_event_paranoid = -1
This configuration reduces kernel logging overhead and restrictions to keep the system quieter and slightly more performance-focused. It limits console log verbosity, disables rate-limiting for kernel messages during testing or benchmarking, and relaxes perf access restrictions so performance analysis tools can collect deeper system metrics.
Security → Performance Trade-offs
# Disable ptrace restrictions (some profilers need this)
kernel.yama.ptrace_scope = 0
# Disable address space randomization
kernel.randomize_va_space = 0
# Allow unrestricted dmesg access (for debugging)
kernel.dmesg_restrict = 0
# Disable kernel log restrictions
kernel.kptr_restrict = 0
This configuration relaxes several kernel security restrictions to make low-level debugging and profiling easier. It allows unrestricted ptrace access for debuggers, disables address space randomization, and permits full access to kernel logs and symbols through dmesg and /proc. These settings are useful for development and performance analysis but reduce system security if left enabled on production systems.
Advanced Networking Stack (optional for online games)
# Faster ARP cache refresh
net.ipv4.neigh.default.gc_interval = 30
net.ipv4.neigh.default.gc_stale_time = 60
# Optimize UDP performance
net.ipv4.udp_rmem_min = 16384
net.ipv4.udp_wmem_min = 16384
# Disable ICMP redirects
net.ipv4.conf.all.accept_redirects = 0
net.ipv4.conf.all.send_redirects = 0
This configuration tunes network responsiveness and security behavior. It shortens ARP cache cleanup intervals so stale network neighbor entries are refreshed more quickly, which can help maintain accurate routing on active networks. The UDP buffer minimums are slightly increased to improve reliability for small UDP workloads. Finally, ICMP redirects are disabled to prevent the system from accepting or sending routing redirection messages, which reduces the risk of certain network-based attacks and keeps routing decisions predictable.
Real-Time / Low-Latency Mode (Optional)
# Favor low latency over throughput globally
kernel.sched_rt_runtime_us = -1
kernel.sched_rt_period_us = 1000000
# Reduce tick timer overhead
kernel.timer_migration = 0
This configuration adjusts the kernel scheduler to prioritize low-latency workloads. Removing the runtime limit for real-time tasks allows them to run without strict throttling, which can reduce delays for time-sensitive processes such as games, audio, or benchmarking tools. Disabling timer migration keeps timer events on the same CPU instead of moving them between cores, which slightly reduces scheduling overhead and can improve consistency in latency-sensitive workloads.
Apply Configuration
Save all above lines into:
sudo nano /etc/sysctl.d/99-performance-tweaks.conf
Then apply immediately:
sudo sysctl -p /etc/sysctl.d/99-performance-tweaks.conf
Verification Commands
Check whether values applied successfully:
sudo sysctl -a | grep -E "sched_|vm\.|net\.|fs\.|kernel\."
Monitor kernel parameters:
cat /proc/sys/vm/swappiness
cat /proc/sys/net/ipv4/tcp_congestion_control
Notes & Recommendations
- Combine this sysctl config with kernel boot parameters like
mitigations=off,intel_idle.max_cstate=1,nohz_full=1, andisolcpus=for true low-latency mode. - For benchmarking or gaming, disable services like
irqbalance,snapd, and background telemetry. - Use
cpupower frequency-set -g performanceto fix your CPU frequency at max. - Always back up your system before applying such configurations.