Methodology & Data Limits

Transparent analysis matters. This page explains what Arbiter measures, how insights are generated, what confidence means, and where interpretation should stay cautious.

Scope clarity: Arbiter automatically identifies patterns, but system changes are only applied when you explicitly choose and confirm a Quick Fix.

How Data Is Collected

Analysis combines frame-level and system-level telemetry.

Frame Timing Pipeline

  • Per-frame timing and present mode capture
  • 1% low and frame pacing analysis
  • Session-level stability indicators

System Event Tracing

  • CPU usage, interrupts, DPC/ISR patterns
  • Memory pressure and paging events
  • Disk and background process contention

Correlation Layer

  • Aligns stutter moments with system events
  • Flags likely bottlenecks and causal chains
  • Ranks findings by impact and confidence

Capture Reliability & Cleanup Verification

Operational safety controls reduce stuck sessions and improve reproducibility.

Session Safety Model

  • Single-instance lock prevents concurrent capture collisions
  • Capture lease + heartbeat enables stale-session recovery
  • Startup janitor clears stale flags and prior orphan state

Two-Phase Stop Strategy

  • Graceful stop path first, then forced escalation when required
  • Watchdog timeout triggers automatic escalation and verification
  • Reason-code tagging supports postmortem triage

Post-Stop Verification

  • Checks for lingering trace sessions and orphan capture processes
  • Sweeps stale artifacts with retention for debugging
  • Writes reliability telemetry consumed by release KPI gates
Support workflow: Export Diagnostics (Ctrl+D) to attach cleanup verification evidence when reporting capture-stop issues.

Hardware Information & Motherboard Graphic

Arbiter includes hardware-aware diagnostics to add context, not just raw FPS numbers.

Hardware Profile in Tool

The app surfaces a hardware profile including CPU, GPU, RAM, VRAM, storage behavior, and thermal context so recommendations are tied to your actual system tier.

Motherboard Component Graphic

The motherboard-style visual diagnostic panel maps health signals across components (CPU, memory, storage, thermals, and selected buses) to make bottlenecks easier to spot for non-experts.

How It Is Used

Hardware context is used to prioritize recommendations. Example: a thermal limit warning and clock drops may prioritize cooling and fan curve steps before graphics-quality changes.

Interpretation tip: The motherboard graphic is a diagnostic aid. It should be validated with before/after captures when making system changes.

Confidence & Known Limits

Not all findings have equal certainty. Arbiter reports confidence to reduce overconfidence.

High Confidence

Repeated patterns across capture windows with consistent metric alignment (for example sustained GPU saturation + stable CPU headroom).

Medium Confidence

Correlated symptoms with less repetition or mixed context (for example intermittent paging during specific scenes).

Lower Confidence

Short captures, inconsistent workloads, highly dynamic scenes, or simultaneous overlapping bottlenecks.

Where Limits Exist

  • Very short sessions can misclassify transient spikes
  • Overlay/streaming workloads can mask game-only behavior
  • Driver and OS policy changes may alter baseline interpretation
  • Unsupported graphics paths can reduce telemetry richness

Best Practice

  • Capture 5-10 minutes of real gameplay
  • Apply one or two fixes at a time
  • Re-capture and compare deltas before keeping a change