Skip to content

Why Latency Matters

For X-Plane, temporal predictability matters more than raw throughput — a stable 35 FPS image looks smoother than one fluctuating between 25 and 50. Micro-stutters rarely stem from insufficient computing power but from latency: scheduling delays, CPU wake-up times from sleep states, interrupts at the wrong moment, and blocking memory operations. The tuning page presents two kernel profiles — the stock profile forces application priority, the Liquorix profile removes external disturbances — each tailored to the kernel's scheduling model. Applying the same parameters on the wrong kernel will degrade results.

The monitoring page provides the tools to verify every tuning measure: Is the governor actually active? Are interrupts landing on the protected cores? Is the NVMe causing wake-up latencies? Each tool — from turbostat to mpstat to ioping — maps to a specific tuning setting.

The theoretical foundations — why latency matters more than throughput and which system sources generate latency — are covered in the Latency and Predictability chapter.

  • Tuning — Kernel parameters, CPU governor, interrupt affinity, NVMe tuning
  • Monitoring — btop, turbostat, perf, mpstat and other analysis tools

Further Reading

Topic Page Focus
Kernel Tuning Kernel Tuning CPU governor, interrupt affinity, NVMe tuning
Monitoring Monitoring Verify tuning with turbostat, mpstat, ioping
Latency and Predictability Latency and Predictability Theoretical foundations of system latency
Liquorix Kernel Liquorix Kernel Low-latency kernel with PDS scheduler
Load Dimensions Load Dimensions CPU, GPU, IO — where bottlenecks form