FIELD MANUAL / 14
Device interrupts & verification
Version 2.2.2.3 reorganizes interrupts into observation, device selection and evidence-based core configuration. Use multiple sessions of the same game and configuration.
- Where to find it
- Advanced → Device interrupts
- When it applies / restoration
- Observation is diagnostic collection. Affinity writes persist and require a restart and subsequent-session verification.
Step by step
- Enable observation as administrator and collect normal game sessions. Collection has overhead; disable it when diagnosis is complete.
- Browse or search input, storage, network, display and audio groups. Look for a specific issue backed by evidence.
- Select one device and inspect candidate cores and exclusion reasons. Candidates based on fewer than three sessions are preliminary, not proof of improvement.
- Apply, restart and collect comparable sessions. Inspect adjustment verification; restore one selected device or all recorded changes as needed.
How candidate cores are selected
Candidates use comparable sessions with at least 80% per-core coverage, average load below 60% and busy time below 20%, excluding game physical cores. The first entry is not a universally best core, especially with incomplete evidence.
Applied does not mean improved
Three post-restart and three baseline sessions compare slow DPC on game cores, long frames and target-core load. Results distinguish improvement, no clear improvement and insufficient evidence. Write time and baseline history are retained.
Device-specific tradeoffs
Pinning multi-message MSI-X devices can reduce parallelism; StorPort completion DPC can follow the initiating CPU. Busy or low-frequency cores can worsen input/audio latency. Change one device at a time and restore its recorded settings if results regress.
Archived 2.2.2.3 instructions; source references are pinned to the public v2.2.2.3 tag. The original document path is absent from that public tag; this link opens the archive root.