Audio latency in games is more impactful than most people realize. If the sound of a gunshot arrives 80ms after the actual event, your brain is processing information that's already out of date. In competitive games, audio cues — footsteps, reload sounds, ability sounds — are as important as visual information.
Where the latency comes from
Windows audio goes through several layers:
- Game generates audio sample
- Passes through Windows Audio Session API (WASAPI)
- Processed by audiodg.exe
- Sent to audio driver
- Played by hardware
Each layer adds buffering. The default buffers are sized for stability (no dropouts) rather than latency.
audiodg.exe priority
Audiodg.exe is the Windows audio processing process. By default it runs at normal priority — same as any background application. When the system is under load, it can get starved of CPU time, causing audio dropouts or delayed processing.
HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\audiodg.exe\PerfOptions
CpuPriorityClass = 3
IoPriority = 3
MMCSS Audio profile
The Multimedia Class Scheduler Service (MMCSS) is supposed to give audio processing elevated priority. The default Audio task configuration is conservative:
HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile\Tasks\Audio
Priority = 6
Scheduling Category = High
HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile
SystemResponsiveness = 0
NoLazyMode = 1
LazyModeTimeout = 1
SystemResponsiveness = 0 tells MMCSS to give 100% of CPU time to multimedia tasks when needed, rather than reserving some for other processes.
Sound enhancements
Windows applies sound enhancements (equalization, spatial audio processing, bass boost) in the audio pipeline. These add processing time. For gaming, disable them:
Go to Sound settings → your playback device → Properties → Enhancements → Disable all enhancements.
Or via PowerShell to hit all devices at once.
Timer resolution
Audio buffers are flushed on a timer. At 15.6ms default resolution, your audio buffer processes at most 64 times per second. At 0.5ms resolution, it processes 2000 times per second — much finer grain.
IzanagiOP's Audio Pack handles all of this: audiodg.exe priority, MMCSS Audio/ProAudio profiles, sound enhancement disable across all endpoints, timer resolution. Discord: https://terweb.lt/
Top comments (0)