Debian-based, empowered by a real-time RT kernel for strict microsecond-level scheduling determinism and zero-compromise low-latency execution.
At the core of the distribution is a fully preemptive RT kernel. High-priority processes execute with microsecond-level timing determinism without being interrupted by the system.
Utilizes aggressive CPU scheduler optimization, hardware interrupt (IRQ) affinity steering, and optimized virtual memory configurations to eliminate input and bus latencies.
Bypasses system-wide memory restrictions. Audio and data processing engines are permitted unlimited RAM locking (memlock), eliminating disk swap overhead and audio crackles.
Perfected specifically for Digital Audio Workstations (DAW/Studio), high-precision robotics, industrial automation, and real-time high-throughput data streaming.
Features a highly customized, stripped-down XFCE4 desktop environment completely free of system overhead, background services, and bloatware to reserve maximum CPU power for your tasks.
Hardware power-saving features and security mitigations (Spectre/Meltdown patches) are disabled. Guarantees zero latency from power state transitions and unlocks raw silicon speed.
During the boot phase, the system transmits hardware bypass, latency elimination, and kernel preemption policies directly to the Linux kernel:
Completely disables all kernel-side software mitigations for hardware vulnerabilities (Spectre, Meltdown, MDS, L1TF, ZombieLoad, SRBDS, GDS). This removes pipeline barriers and KPTI overhead, yielding raw CPU speed.
Enables maximum kernel-space preemptibility. Low-priority kernel tasks (such as background disk writes) are immediately interrupted at the microsecond level the moment a high-priority RT thread (such as DAW buffers) demands CPU cycles.
Converts hard-IRQ hardware interrupt handlers into prioritizable kernel threads. This allows assigning high real-time priorities (like 95) directly to sound cards or USB hubs, leaving disk/network interrupts at a lower tier.
Fully disables hardware and software watchdog detectors. Eliminates per-core Non-Maskable Interrupts (NMI) and polling ticks, leaving CPU cores uninterrupted for actual task execution.
Offset per-core scheduling clocks across the timeline, avoiding global lock contention and memory bus saturation spikes caused by simultaneous processor checks.
Deactivates system-wide auto-Transparent Huge Pages, restricting them to programs explicitly requesting them via madvise. This prevents background memory defrag runs from introducing millisecond-level freeze spikes.
Disables automatic NUMA memory page balancing. Eliminates periodic background memory scanning and page migration across processor sockets, preserving fixed memory latencies.
Locks Active State Power Management and port PM to maximum. Prevents PCIe lanes from entering low-power states, eliminating wake-up delays that cause dropouts on sound cards and NVMe drives.
Silences RCU CPU stall detectors and soft-lockup warnings when long-running, high-priority real-time threads dominate processor cores without yielding, avoiding logging bottlenecks.
Alt-seviye kernel parameters adjusting scheduler behavior and memory page flushes to minimize CPU overhead:
Forces parent processes to execute before child processes after fork calls, preserving warm L1/L2 caches and eliminating cache-miss delays.
Disables terminal session autogrouping. Calculated weights are processed per-thread, giving heavy multi-threaded DAW or automation tasks direct access to CPU slices.
Enforces process affinity by delaying core-to-core migrations for warm tasks up to 5ms, avoiding expensive L1/L2 cache invalidations.
Reduces the scheduler epoch interval to 4ms, shrinking process wait times in the queue for responsive operations.
Sets task minimum execution time and wake-up preemption delta to 0.5ms. Ensures rapid wakeup preemption and low timing jitter.
Lowers the kernel's swap-out tendency to 5%, ensuring all active real-time processes stay inside physical RAM to bypass disk access bottlenecks.
Retains file system path and inode metadata in RAM longer, speeding up access and cataloging on huge sample directories.
Forces flush daemons to write dirty data to storage at 2% dirty cache, blocking writes at 5%. Prevents massive storage I/O write spikes that occur when huge buffers build up.
Limits dirty data age to 3 seconds and sets flush daemon wake-ups to 1 second, guaranteeing smooth, consistent disk synchronization.
Reserves 128MB of RAM for atomic kernel allocations, allowing instantaneous memory acquisitions under heavy workloads without page-reclamation halts.
Expands virtual memory areas allocation limits to 1 million, accommodating dense memory indexing required by massive DAW samplers.
Slows memory stat logs, disables background memory compaction, and turns off zone reclaim. Keeps memory allocations deterministic and completely static.
System boundaries, file indexing capacity, and TCP stack configurations adjusted for lowest possible ping times and zero packet dropouts:
Expands Process ID limits to 4.19M and system file handles to 2M to accommodate massive concurrent execution lines.
Lifts performance counters restrictions and resolves kernel symbol addresses, allowing latency profiling and eBPF tracing tools to run without privilege blocks.
Increases file watch directories to 524k, resolving bottlenecks when DAWs or database engines index massive directories.
Raises maximum packet backlog to 4096, preventing network interface card packet dropouts during sudden data streaming bursts.
Sets socket read/write buffers limits to 25MB and connection limits to 4096, preventing socket queues from bottlenecking.
Enables TCP Fast Open on client and server lines, allowing data transfer directly within the initial SYN packet, bypassing round-trip handshake delays.
Switches TCP congestion control to BBR, which models real bottleneck bandwidth and RTT instead of interpreting packet loss as congestion, preventing bufferbloat.
Allocates a 50-microsecond active socket polling window. Forces the CPU to check sockets directly for incoming data, bypassing hardware interrupt overhead to minimize ping delay.
Boot and hardware automations enforcing stable, unthrottled, low-latency device performance:
Pins all peripheral interrupts (USB, SATA, Network, GPU) strictly to CPU0. Isolates CPU1, CPU2, CPU3, etc., from answering hardware polling, allowing real-time applications bound to these isolated cores to run uninterrupted.
Disables RCU expedited grace period scheduling on isolated cores to prevent unnecessary processor wakeups (IPIs) and deploys active hugepage and memory configurations.
Deactivates scheduling (none) on NVMe drives to bypass redundant OS queue calculations, while deploying the latency-fair BFQ scheduler on SATA SSDs.
Forces SATA link power management policies to max_performance, preventing SATA controllers from entering low-power states, bypassing uyanma (wake-up) lags.
Locks mechanical hard disk drives to their peak power profile and blocks disk spin-downs, keeping disk platters permanently active to avoid spindle spin-up delays.
Completely disables USB auto-suspend power management, keeping USB ports and attached interfaces locked in 100% active power states to prevent wake-up lags or connection drops on external audio interfaces.
Locks the processor frequency governor and energy-performance preference strictly to performance mode, blocking processor C-states. Keeps all cores pinned at maximum turbo clock frequencies constantly to eliminate scaling delays.
Overrides default system boundaries, allowing high-priority real-time workloads to control RAM and CPU priorities directly:
Allows user processes in the audio group to capture up to 95% real-time scheduling priority (rtprio) inside the kernel. DAW audio buffers can execute at higher priorities than standard kernel tasks, preventing dropouts.
Grants the audio group unlimited RAM memory pinning (memlock). Keeps VST instrument samplers, buffers, and active memory pinned in physical RAM, completely preventing page swapping and associated audio cracks.
Permits audio processes to adjust their static execution priority (nice value) down to the maximum level of -19.
Grants all standard system users a baseline 50% Real-Time priority and 512MB RAM locking budget to support general low-latency operations.
Key release details and system features:
performance mode with blocked C-states.irq-affinity.sh) deployed as a systemd service.