while gaming industry is obsessed about low latency (screens, mice, graphics cards etc...) reaching 1's of ms latency, audio in games is stuck at ~100 ms latency....

....aand ASIO isn't an option in games (game engines) sadly

thoughts?

 

Spoiler: Source: Grok KI

You're absolutely right that the gaming industry has pushed hard for ultra-low latency in visuals, input devices, and graphics processing, often achieving single-digit millisecond performance. However, audio latency in PC gaming remains a stubborn bottleneck, frequently hovering around 100 ms or more, which can feel jarring in fast-paced games where audio cues are critical. Let's break this down and explore why audio latency lags behind and what can be done.

Why Audio Latency is High in PC Gaming

1. Audio Stack Overhead:
   • On Windows, the default audio stack (WASAPI or older DirectSound) introduces significant latency. WASAPI in shared mode, commonly used by games, buffers audio to ensure compatibility across diverse hardware, resulting in 50-100 ms of latency.
   • Unlike graphics, where DirectX or Vulkan can directly interface with hardware, audio often passes through multiple layers (OS, drivers, middleware), adding delays.
2. Game Engine Limitations:
   • Game engines like Unreal, Unity, or proprietary ones prioritize compatibility and ease of use over low-latency audio. They rely on high-level audio APIs (e.g., XAudio2, FMOD, Wwise) that aren't optimized for sub-10 ms latency.
   • ASIO, as you mentioned, is a low-latency audio driver protocol, but it's designed for professional audio applications (e.g., DAWs like Ableton or Pro Tools). Game engines don't support ASIO because it’s not practical for consumer-grade hardware, lacks universal compatibility, and requires exclusive hardware access, which conflicts with other system audio needs.
3. Hardware and Driver Variability:
   • Unlike GPUs or monitors, audio hardware (soundcards, DACs, or onboard chips) varies widely in quality and driver optimization. Consumer audio devices prioritize cost and compatibility over latency, unlike pro-grade interfaces.
   • Bluetooth headsets, popular for gaming, often add 100-200 ms of latency due to encoding/decoding (even with low-latency codecs like aptX LL).
4. Buffering for Stability:
   • Games use large audio buffers to prevent glitches or dropouts, especially on systems with varying performance. Smaller buffers reduce latency but risk audio stuttering if the CPU is under heavy load, so developers err on the side of stability.
5. Lack of Industry Focus:
   • Audio latency is less perceptible to most gamers than visual or input lag, so it gets less attention. The human brain is more forgiving of slight audio delays (up to ~100 ms) compared to visual stuttering or mouse lag, which are immediately noticeable.
   • Competitive gamers, however, notice audio latency in scenarios like shooters, where sound cues (e.g., footsteps, gunshots) need to sync perfectly with actions.

Current State of Audio Latency in Gaming

• Typical Latency: Most PC games with standard setups (WASAPI, onboard audio, or gaming headsets) have 50-150 ms of audio latency, depending on the hardware, drivers, and engine.
• Best-Case Scenario: Using high-end DACs or soundcards with optimized drivers (e.g., Creative Sound Blaster, ASUS Xonar) and WASAPI exclusive mode, latency can drop to 20-50 ms, but this is rare and often requires manual tweaks.
• Wireless Audio: Bluetooth or 2.4 GHz wireless headsets add latency. Even "gaming-grade" wireless headsets (e.g., SteelSeries Arctis, Logitech G Pro) typically achieve 15-30 ms at best with proprietary protocols.
• Console Comparison: Consoles like the PS5 or Xbox Series X often have tighter audio pipelines (thanks to custom hardware and OS integration), achieving 20-50 ms latency, but still not near the 1-5 ms of graphics/input.

Why ASIO Isn't Viable for Games
You mentioned ASIO, and you're right to lament its absence. ASIO achieves 1-10 ms latency in pro audio, but here’s why it’s not used in games:

• Exclusive Hardware Access: ASIO locks the audio device, preventing other apps (e.g., Discord, Spotify) from outputting sound, which is unacceptable for most gamers.
• Driver Support: Only high-end audio interfaces (e.g., Focusrite, RME) fully support ASIO. Consumer soundcards or onboard chips often lack proper ASIO drivers.
• Complexity: Game engines would need to implement ASIO support, increasing development complexity and limiting compatibility with budget PCs.
• CPU Overhead: ASIO’s small buffers demand more CPU cycles, which could impact game performance, especially on lower-end systems.

Potential Solutions and Workarounds
While we’re stuck with ~100 ms latency in most cases, there are ways to mitigate it:

1. Use Wired Audio:
   • Wired headsets or speakers connected via 3.5mm or USB DACs have lower latency than wireless. High-end DACs like the Schiit Modi or Focusrite Scarlett can approach 20-30 ms with good drivers.
2. Optimize Windows Audio Settings:
   • Set your audio device to the lowest supported buffer size in the driver control panel (e.g., Creative or Realtek software).
   • Use WASAPI exclusive mode if the game supports it (rare), as it bypasses some Windows audio mixing overhead.
3. High-End Soundcards:
   • Cards like the Creative Sound Blaster AE-7 or EVGA Nu Audio support low-latency modes (20-40 ms) with custom drivers, though not ASIO-level.
4. Game Engine Tweaks:
   • Some engines (e.g., Unreal) allow developers to adjust audio buffer sizes via config files. Modding communities occasionally share tweaks for specific games, but this is game-dependent and risky.
5. Custom Audio Middleware:
   • Developers could adopt low-latency audio middleware like Steam Audio or Resonance Audio, but adoption is slow due to added development costs.
6. Future APIs:
   • Microsoft’s Windows Sonic or DirectX 12 audio improvements could reduce latency in future games, but no major breakthroughs have been announced.
   • There’s hope for custom audio APIs tailored for gaming, similar to how NVIDIA Reflex optimizes input latency, but nothing concrete exists yet.

Why It Matters and Future Outlook
High audio latency can desync audio cues from gameplay, impacting immersion and competitive performance. For example, in Valorant or CS2, a 100 ms delay in hearing footsteps could mean the difference between reacting in time or getting outplayed. As VR and spatial audio grow, low-latency audio will become more critical for immersion.

The industry could address this by:

• Developing gaming-focused low-latency audio APIs.
• Optimizing game engines for smaller audio buffers on high-end hardware.
• Educating developers on audio latency’s impact, similar to the push for low input lag.

For now, gamers can minimize latency with wired setups and driver tweaks, but a systemic fix requires industry-wide effort. If you’re curious about specific games or hardware, I can dig deeper into their audio performance or check X for community insights on low-latency setups. Let me know!

 

A lot in games is tied to the frame rate, in a standard game loop it goes something like this: read input, update world state (like move based on the input, do enemy AI, object physics, what have you, dispatch audio events), render graphics, repeat. Now, of course you can run some stuff in threads but those tend to do other things, like feed the GPU, do some background loading for assets etc. etc.

At the moment given a refresh rate of 60Hz then that means each step happens every 16ms or so. So while you could shoot for a more aggressive latency, that's the granularity and there's not really escaping it.

Our brains will adjust to it and it feel simultaneous anyhow, so most games would see no improvement in dropping the audio latency. And for game devs... lower latency would not improve sound quality, and any chance of audio dropouts is a no-no (likely given how much games can tax the system).

So not going to happens unless things improve organically, like standard audio drivers lowering their latencies by default. Or maybe some bright spark does market research and comes to the conclusion that low-lat audio in games makes them sell 5% more.

*edit:* grammar

 

Last edited: Sep 26, 2025

I hate that, I hate that soooooo much! C64, near zero sound latency, Amiga, same. Same with fcking 8bit Gameboy and og NES. PC sucks at everything sound ever since IBM stuck it together at lowest effort possible.
I have some rapis with retroarch für retro gaming. That bloody little bastard gets down to 16ms on that linux based system. Not great, but good enough for me. Feels realtimey enough for me. My fully blown 16 core workstation has 100ms above from mouseclick to 'BLAAM'. And people just don't care. They don't even feel the sound latency. But they spend a shit load of money for displays they think have 2ms lag. And say stuff like "uh yes, there is a huge difference between those 8ms monitors and those 2ms ones." Sure, sure... but you don't even get when sound is 1/10 of a second late.

 

This used to be the case, but even 15 years ago that was already considered bad form. You don't tie things to frame rate because that has disastrous consequences, it's a console thing from the times where you could output stable 30/60fps. Hasn't been like that in a long time, last games that did this were on the PS3 era. Games doing that these days are just wrong, period.

 

You misread me, I know that you don't tie the game logic or physics to frame rate, and implement timesteps decouple everything else from the frame rate (in case it drops, for instance), the simplest method is to update the game state by the time it took to process a frame. That's not perfect but it already prevents the game logic and such to be bound to how long it takes to draw.

But you still need to take input and update the game state before rendering, so none of this changes the fact that you will have a master game loop based on the graphics rendering and framerate.

 

I think almost everyone (in competitive gaming) trained themselves to mostly rely on visual cues.

The older Counter-Strike games had a snd_mixahead setting, but that would start crackling and tanking performance as soon as you got down to the <45ms range.

Some amount of fixing could probably be had just by supporting newer APIs (WASAPI changes in later Windows 10 revisions).

But convincing the company's suits to allocate resources for such a "whatever" subsystem is difficult when the money is with couch gamers who play on consoles, which use all sorts of latency-inducing fuckery, on even slower giant-ass TVs developed for visual "experiences".

Just not worth the CPU overhead (for decision makers).