The main reason why ASIO (or WASAPI exclusive mode) is never used in games is because it's the audio equivalent of exclusive fullscreen mode. the application detaches the OS sound mixer and sends the audio signal directly to the hardware, which means you can't hear anything else (just like you can't see any other windows while in exclusive fullscreen, except for overlays that intercept the render pipeline). Also one important note about LDAT: depending on what audio equipment you use, the output device itself can actually cause a notable amount of audio latency which LDAT doesn't reflect (just like debounce or pixel response time, but much bigger)
@@AuxHex i know, thats why i specifically said "WASPAPI exclusive mode" @Howard Nyugen but for that all the applications would have to implement ASIO, dont they? Edit: i misread. If that's actually possible that would be really cool and i'd wonder why it isnt made use of
@@LastExceed These days, there are many options and workarounds for generic exclusive-mode Asio drivers such as Voicemeeter (you can route less important audio at a delay compared to the asio source) and other multi-client Asio interfaces. As one example, some DAWs like FL Studio actually come built-in with a high quality non-exclusive Asio driver, allowing everything else in the Windows mixer to come through while still providing low latency audio inside the DAW at the same time. While it isn't _quite_ as low latency compared to standalone exclusive asio drivers, it's still a huge improvement nonetheless, especially for the convenience factor. If games developers had the know-hows to write an audio driver like FL's Asio, or companies like THX or Dolby developed their own non-exclusive asio standard for developers to license, we could theoretically have huge latency improvements.
That is correct, Steinberg's ASIO has the "downside" that only one application can use the sound card at a time. So if you listen to music or use discord at the same time then ASIO is not a solution - *this is why I mentioned that in the context of supporting it in competitive games* where players usually only want the game audio with no distractions - ofc voice communication might still be an issue here if an additional application like discord is used. That is where WASAPI could come in handy as an alternative to ASIO. I will see if I can produce some testresults using that driver/api. :)
I've never even thought about audio delay in a competitive environment before. Seeing those huge differences in audio delay can mean a world of a difference when trying to react to something like Reinhardt's "HAMMER DOWN" audio or Zarya's graviton and such. Amazing detailed video and topic! Great stuff.
Sound is a lot more complex topic than video delay. The way humans perceive sound is very weird, audio and video delays allow us to judge distances. For our bodies light travels instantly, it's too fast to notice any delay between something happening in front of our face and something happening across the street. But sound WILL have a delay. And our brains use this difference to adjust perception of distance. This leads to two interesting effects. First of all, if sound is coming from a fixed point in space and doesn't change distance, then our brains can adjust to a very high sound latency, up to 200ms or so and everything will sound natural unless you're looking for this delay specifically. Second effect is that in a dynamic environment like a competitive game, sound effects should not only be modelled between left and right ears, but also in time depending on a distance from the player. Otherwise the game sounds "flat" and unnatural. Which is the case for most games of today. And while we're adding distance based delay, we should also add correct Doppler effect emulation. DE + delays together create a fully immersive 3D picture of the in-game world without which in-game sound is nothing more than a blinking icon in a status bar.
_"The way humans perceive sound is very weird"_ That might be so, yet it is also a fact that we already perceive sound as out of sync at a difference of 45ms, which is well known among those who have to deal with audio to video synchronization (i.e. lip sync, ...). The audio delay in games is design to not be *ahead* of the image - and that intentional audio delay is based on 60Hz v-sync where you have an average B2P of > 100ms. On console we still do not have an issue (maybe with next gen as that supports 120Hzs an VRR). But on PC where we have reduce the B2P to less than 20% of the B2A, we are well past the threshold of 45ms where humans notice that image and audio are out of sync. The other aspect is competitive gaming where game audio is an important cue - 20ms instead of 150ms delay is a noticeable improvement, especially on a high level of play. In story driven games it is less of an issue, unless the delay is noticeable when it comes to lip-sync - if the animations and character models are detailed enought for that.
This video demonstrated a 36ms to 176ms+ audio delay for things like gunfire that happen immediately close to the observre. That is in no way a realistic emulation of sound. The fact that such a difference exists indicates that a very large portion of that delay is unintentional and not intended to be an accurate simulation of sound delay during travel. If the actual difference between audio hardware is up to 10 ms as demonstrated by the ASIO driver, you should not be seeing a 140ms swing. Humans perceive sound based off: 1) the distance of the observer from the sound source 2) The difference in intensity / delay between your two ears. I know of very few games that simulate the difference in intensity of the same sound between two ears. Overwatch has a feature called Dobly atmos (which perhaps should also be tested) which does positional sound better than many games but it is still a far cry from the technology used in VR titles that make use of HRTFs (or a game like Senua's sacrifice). It's really a shame that audio delay is so high when it could be drastically lower and for that matter, much more immersive.
Lovely! Many thanks for the info provided and introducing me to ASIO. IMO absolute delay values are to be taken with a grain of salt, coming from video games. Sound design is not checked as thoroughly, as other disciplines of a game. It is common to insert zero-samples at the beginning of a sound clip to prevent popping and crackling. Next of, having a too fast of an attack in a sound effect may make it sound unnatural. As such, a slower attack may be chosen for a more pleasent sound effect. This may result in the sound staying longer below the noise floor and thus a couple ms will be added by the LDAT measurement. So a higher LDAT measurement may not be an engine limitation, but a choise by the sound designers, prioritising how it sounds, as opposed to a perfect sync, especially with consoles hooked up to TVs with very high button-to-pixel delays. The relative delays are of course rock solid and this video is an exellent source of info! edit: In a perfect world, a sound mod would be an interesting thing to make: Replace all sounds of a game with a square wave going instantly to 100% volume for a cleaner measurement.
Yes, (as I mentioned) considering that at 60Hz with v.sync the Button2Pixel delay is over 100ms, the audio delays of >120ms are in the safe spot to prevent that the audio gets *ahead* of the video. On console we currently do not have an issue _(though that might change with the 120Hz support of "nextgen)_ . On PC the situation is different as we have dramatically lowered the B2P delay during the last few years, increasing the rift between video and audio.
@@BattleNonSense Sadly, we do not have a "gaming" oriented audio company to develop low latency audio technologies or standards for games :( It feels like we are going to be doomed to high latency audio in games for a very long time.
@@GENKI_INU We don't need gaming oriented audio companies. The standards already exist in the professional audio world, mainly tunable buffers, ASIO, and realtime-priority audio threads. I use a Native Instruments Komplete DJ mixing interface for my headphones/microphone, which runs at a internal buffer of only 192 samples, compared to the 768 tested in the video, and output buffer delays
@@henryzhang7873 We do need new standards. Not everyone owns an audio interface or something that even supports ASIO. Game engines also have different pipelines and different needs to suit them better with their workflow, different from what producers need. If Realtek audio chips in the future came with an ASIO-like standard that developers started mass adopting, or if AMD and Nvidia's graphics cards focused on their built-in audio with low-latency game drivers that developers adopted, that would benefit everyone.
@@GENKI_INU yeah but then the drivers should use ASIO. ASIO isn’t a driver but also named after a programming technique. It’s the idea that as soon as you know the content of a sample, you pass it immediately to the sound card with minimal buffering (more complicated than this, but this is the important part), using special OS or hardware systems. This reduces delay because you are not waiting for more samples to accumulate. Not only audio uses it: I use it in the web servers I write to have the best possible latency. It’s just hard to program for because humans are not used to dealing with asynchronous streams as a concept. Developers that know how to do this get paid way more working in the tech space than gaming. The server you are watching this video off of is likely also based on ASIO principles because Google/YT has been pushing these kind of event driven paradigms for years.
I think this is really interesting, and something I have experienced in games myself. Games like Escape From Tarkov, Insurgency: Sandstorm and CoD MW, uses sound as awareness mechanic, and I find that when playing those games the sound lags behind when you turn sharply. Would you consider testing some of the games that use sound as an awareness mechanic? Thanks for an interesting video!
I would second sound tests or more tests in general include COD:MW2019 . Its sound is great quality wise... but is seemingly broken in many ways so it doesn't deliver in Multiplayer how it has in the past (near perfect for player ability to use it as intended as spacial awareness assistance etc).
I've started experiencing audio delay after upgrading to a TCL c635. The delay is only in 60fps games and more noticeable when going from stereo uncompressed to DA for headphones or any of the other Dolby formats. I can't understand what causes it and don't know if getting a standalone audio system will get rid of the audio lag
As someone interested in rythm games, this is very enlightening. I did not expect the usb Sound cards to be that bad in terms of latency but feel reinsured that sticking with 48kHz and pushing the buffer size down in pipewire was the right choice. Thanks for doing the empirical test and presenting the results in such a nice way!
I knew I wasn't going crazy when some games felt more out of sync than others regarding weapons visually firing and producing the gunfire sounds. Sucks to see there isn't any reasonable solution to reducing it beyond what I'm already doing. Thanks for the video!
My X-Fi Audigy 2 sound card used to have a "creator mode" which seemingly drastically reduced audio latency across the system. Problem is that card came out in 2002 and is PCI (not PCIe) only, and ever since then I've been bothered by audio latency. It really sucks that it's such a hugely overlooked part of the equation, with no good solutions available.. I hope that some good solution comes out of this!
At the end where he's talking about whether people really want lower delay, and mentions footsteps and such, I would just like to mention that rhythm game players have a huge stake in audio delay being lower.
There's a program for windows called REAL (reduce audio latency) that forces the buffer to be lower than default of 10ms, on my usb dac it defaults to 7ms which works well for me. I've tried my motherboard built in audio, but my onboard sound uses 3ms buffer which is broken in every game
Pro tip: If you use the Windows 10/11 "High Definition Audio Device" audio driver, that's the default driver for a fresh install for all chipsets, the latency is insanely low. With Realtek S1200A driver I get 95ms latency. With High Definition Audio Device driver I get between 30 and 38ms. With Astro A40 mixamp I get 110ms.
do you know a way to keep it, because windows instantly update the driver to realtek, i always felt that games run very well after a new windows installation but i didn't know why
Oh gosh please please please run this test with osu!. Audio delay is very important in rhythm games, and the osu! developer seems to believe that you'll only have a delay >10ms if something is wrong with your setup. It would be amazing to have some concrete evidence to support adding ASIO driver compatibility to the game.
Could you perhaps play around with Voicemeeter Banana in the future? It supports pushing audio through ASIO, as well as has adjustable buffer size both for the sources, and its internal virtual audio cables (via the .exe's contained in program's folder). I'd love to see not only whether and how much Voicemeeter can increase latency, but also whether adjusting its buffer scaling could actually improve latency over standard Windows audio mixing. Thank you very much for this video! An interesting topic, to be sure.
3 ปีที่แล้ว +12
The increased audio latency he's recording in his testing is not related to Windows audio mixing. It's related to the buffer size every game is set to have. So from my understanding, I would say Voicemeeter Banana can only maintain the same latency or increase it, but definitely not reduce it.
I have done setup guides for VoiceMeeter (Banana) and I'm using it on a daily basis, but I do plan to look into the audio delay in a separate video. Early testing _(with reduced WDM buffer size as well as ASIO)_ was not promising - though I want to look into it more to make sure I don't miss anything. :)
@@BattleNonSense I would guess it would only add latency because the game will still be using the WDM/MME audio output and THEN sending it to the ASIO driver via VoiceMeeter.
Audio delay's a huge deal in the rhythm game community We discovered ASIO a loong time ago and are actively looking to get every single game to support it
Would be interesting to know how the delays are if the sound is transfered through hdmi and the headphones are connected to the monitor or are the sound drivers on the graphic cards exactly the same.
@@BattleNonSense I was thinking of this as well, I use external EMU 0404, which has digital input as well as USB connection, I use the USB connection for ASIO, and for other uses than music creation, I route the sound from the integrated soundcard using S/PDIF into my EMU 0404. Would be interesting to see the different digital options such as digital output from integrated sound card, integrated graphics card and PCI-e graphics card. Also if you do end up testing digital audio, does it make a difference if the output is HDMI or Displayport, and does it make a difference if the DAC is a cheap monitor one or something more expensive like a professional/entertainment soundcard or like Andrej mentioned AV receiver.
@@BattleNonSense I also use a USB microphone, namely the AT2020, so I'm wondering if using heaphones through there is a bad idea. I'd imagine it is considering how Dacs cause delay.
Yea. I use hdmi for audio but I used to use on board audio a couple years ago becuase I noticed more audio lag when using hdmi versus on board audio. My thought process as to why on board audio was less had to do with audio technically coming off the cpu closer to the metal versus hdmi which had to convert the signal again. Also. It is a similar using for external docs that have to process audio a second time. I used to use hdmi for video and on board audio using analog stereo connections to my headphones or to analog only speakers. If I plugged in an analog stereo plug to certain speakers that accept digital signal certain speakers re-digitize the analog signal before sending it to the speaker. Also, you didn't test if anything changed from using audio exclusive mode in windows versus non exclusive. I had games that had more audio lag with exclusive on and some games that had more audio lag with exclusive off.
For anyone using Linux: RetroArch has a JACK Audio (the equivalent of ASIO) driver built-in. I've achieved ~7ms of audio latency using it and it feels great.
Hi amazing video (as always) I noticed that they were an audio delay in Overwatch, I hope your video will have an impact on this topic like the one on Gpu bound (you invented Nvidia reflex, congrats). As I have a focusrite sound card I know an issue about ASIO, you need to have all application at same audio frequencies or you'll have heavy sizzling or no sound at all but it's a pretty good idea latency AND quality wise as it bypasses windows. Very good channel you make competitive gaming better.
@ Battle(non)sense Chances of going back and owning Tarkov netcode again? I know it's rough. I heard there may have been dev drama with you and that's why it hasn't been done? But man it's been over 2 years since your videos AND they haven't done a 'network' patch since then either. Feelsbadman. We NEED you to wake them up again! Love you!
There is a hybrid mode which allows to run both ASIO and WASAPI without taking exclusive control over the audio device (at the cost of it not being a complete ASIO). For example, FL Studio ASIO, included with the latest versions, can go as low as 62 ms, although it's still pretty high. But it still sucks that there is a fast AF boi called UR12 that can drop the audio delay to fractions of milliseconds if the system can handle it (minimum 4 samples at 44.1kHz per stream window). This mode takes exclusive control of the device though, yet I'd love to pay the price of being able to only hear the game just to experience nearly instant audio response. EDIT: Also, ASIO reduces CPU time spent on processing audio by a ton. That might be the more important in DAWs, since digital processing can be quite cumbersome even for very fast CPUs.
This video opened my eyes to this issue, but has left me with questions. What about SP/DIF Optical audio output? What about using a dedicated headphone amp connected to the line out on the motherboard? In theory using line out to an amp should have very little delay added since the amp won't be processing the audio, but only testing will confirm. SP/DIF I would love to know about as well to know what delay it adds compared to usb and pci-e. This video makes me hope the topic of audio delay begins to receive the attention it needs in the industry.
I am using a Creative AE-9 PCIe soundcard in "Direct Mode" and your video inspired me to compare the latency to my onboard ALC1220. I am using your technique and am testing in a Modern Warfare (2019) Custom Lobby. I collected ~8 samples from for each soundcard and both are consistently at around 110ms latency. I also tested in normal mode, and it was around ~115ms, but much less consistent.
@@tawl-x Yes, same delay with onboard ALC1220 and PCIe. You have to consider that the Creative AE-9 is a rather expensive high end sound card, maybe they figured out how to do it without delay. Not all sound cards are equal. I dont think battlenonsense's results are false.
Using wasapi give exclusive audio access to the game and usually have waaaay lower latency. However that obviously doesn't play well with discord and other voices apps. Would still be cool for games to give the option though. (The only "game" I know that gives that option is the Dolphin Emulator)
I'm not a competitive gamer (far from it) but this audio delay has been bothering me for years. From the sound confirming a button press in a menu to sound effects in game, it's simply annoying when it doesn't sync up.
Audio processing has a realtime priority in os schedulers for a reason. Was a much bigger problem when cpus were like 1/20 the speed. Honestly don't remember when I last experienced an audio sync issue. Must have been over 4 years ago. Also something I don't think mr. battlenonsense is aware of is that you need at least some sort of audio delay. If audio cues are instantaneous ie sub 1ms it can be disorienting due to delays in how your brain processes an input with the corresponding sound. It'll seem like the sound occurred a fraction of second before you pressed a button as nothing we do on computers is really capable of doing this. You can of course get used to it though. However with the ~30ms in a best case scenario he measured I don't think that'll be a problem.
This aligns with my experience between a PCIe sound card vs motherboard sound, the sound card had about 10-15ms longer delay, indicated by the need to adjust timing offset in a rhythm game. Audio and video desync however, especially just delayed audio, is not easy to perceive, and usually won't result in weird experience like input-to-visual lag does. Sound travels slowly, humans are used to hearing sounds delayed. But I can see how reacting to sounds entirely will result in a competitive advantage.
The reason the ASUS STX card is laggier is due to PLX converter chip (The DSP used is native PCI inteference). You should test a native PCI-E sound card. Results should be in-between i2c Realtek and USB2.0. NOTE: Modern Realtek 4080/4082 has been converted to a USB 2.0 interface vs legacy I2C in ALC1220. Latency should be closer to the FiiO and other USB card you tested. Maybe inbetween since its a more direct path to chipset or CPU depending on AMD/INTEL platforms. I just want people to be aware that the results are likely skewed from what is "possible" on PCI-E. Old creative cards like Titanium HD (Native PCI-E) were always the fastest for delay back around 2010 era.. IMO. Realtek had terrible drivers in that time frame. (Creative was bad too, but "faster") Maybe that's changed.. cant tell you since I use a 2.0 interference these days :D
i'd like to recommend audio interfaces like the focusrite scarlet 2i2, behringer umc 204hd, steinberg ur22 or similar ones. those interfaces are something between a external sound card and a small mixing table. you can get really decent ones for about 60-100€. as guitarist i need a very low audio output and input delay since i play my guitar "through the pc" instead of an amp for recording. any kind of delay between hitting the strings and the processed signal is super obvious, so using ASIO is the only way to go here. but even when not on asio, you can reduce the buffersize in the driver to much lower amounts than on regular sound cards without crackling/ destroying the sound. plus they offer some nice features like different volume controls, headphone amp, mic preamps, outputs and connectivity options.
Man I love these headphones. My Sennheiser Game Zero died recently and I almost bought another set of "gaming" headphones. The DT 770 *with* a mod mic was not only cheaper, it also sounds sooooooo much better!
@@BattleNonSense When the original pads will start to collapse, check out Dekoni Elite Velour pads! They're made out of memory foam that's muuuuch more comfortable and they boost a bit the sub bass (I needed to actually lower the 30 and 60 frequency bands by -2 and -1 db via my sound card EQ after installing them)
@@BattleNonSense Hey, can you test Linux (ALSA or JACK, PulseAudio has a delay) or MacOS? They should have less audio latency (comparable to ASIO on Windows). CSGO should be a good game for that, it has native support on MacOS an Linux
I have DT 770 Pro 250 ohm with IFI ZEN DAC V2, blue snowball mic. I download APO and PEACE EQUALIZER, with DOLBY ATMOS. Still trying to figure out if HESUVI will be better than ATMOS, or an other one. I dont care to spend more money to get the best VSS. Im playing WARZONE. Somebody have a recommandation? Thanks in advance 🙏 @Battle(non)sense
Whenever I think that there are many misconceptions, myths and misunderstood concepts in gaming, I only have to remind myself there's always audio stuff ;) So not surprised with the results in the slightest.
I've been curious of lowering audio latency for a while and I mostly agree with your findings. Onboard sound chip have the lowest latency, and external gaming DAC and internal gaming sound card have more latency. This is why I sold my Creative Sound Blaster Omni, and I changed into using regular external DAC (Topping E30) + headphone amp. Its latency is a bit slower than onboard sound chip by few milliseconds, but increasing sample size from 48khz to 384khz did improve latency by around 3ms. It might matches with onboard sound chip but haven't gone into testing it yet. But my twist is I'm not a competitive gamer, and I prefer using AV receiver + speakers. Audio latency became more complicated as it varies from different brand. For my case, two best solutions for lowest 7.1 sound latency is using HDMI connection (+25ms) or multi channel analog input (+0ms using onboard sound chip). One final note is bitstreaming audio sucks for latency. I tried Dolby Digital Live and DTS connect in the past and latency is as bad as around +100ms. And I cringe when Dolby Atmos is becoming more of feature, as it is also bitstreaming audio and have poor latency too. I rather use 7.1 audio then use upmixer to 7.1.4, and although upmixer add some latency it is still much better than Dolby Atmos.
I have tried some 3rd party sound enhancers to make footsteps and directional sound more easy to point out. But the audio delay with processing the sound in the software was to great to be useful. This video really shows how important that is.
If anything, I'm glad to know that I'm not crazy about the audio delay and that it being around 100 ms is normal. I kept noticing that if I slide my headset half-off, I could very obviously hear the difference between the click and the gunshot. I play guitar through my focusrite scarlett interface using the ASIO driver, that's where I have my headphones plugged in, and the latency is completely unnoticeable, so I just assumed it couldn't be my audio hardware. The mouse also felt perfectly fine so it couldn't be my mouse's input latency.
Would be interesting, to see how an external DAC using an optical connection compares to usb connected DAC. I doubt it would make much of a difference, because your still adding an extra processor processor in there, but it would be interesting nonetheless. One thing to consider, when deciding between using the onboard DAC and an external DAC, is the potential gains in clarity, sound stage, and imaging (where sounds sound like they are coming from), this is also dependent on the quality of the headphones, that you are using. For reference (Take this with a grain of salt.): I use to run my headphones and my speakers through the onboard audio on my Gigabyte Aorus Gaming 7 (Z370) Mother board. When I upgraded my audio setup I to a SMSL SU-8 (DAC) / THX 789 (AMP) setup and DT1990 headphones (I tried the headphones with the on board audio when I first got them), their was a notical improvement in sound stage, imaging, and clarity, from not only the headphones, but also the improvement in clarity and richness of the sound produced by my Klipsch Promedia 2.1 THX speakers seems vastly improved. PC -USB> DAC -XLR> AMP -Single ended 1/4> DT1990 PC -USB> DAC -RCA> Promedia 2.1 With that noted, I will gladly give up a little bit of sound latency, (Especially when I don’t perceive the change) for the bump in clarity and sound stage.
Optical would add delay, since there are two opticalelectrical modem devices/operations. What I think you want is coaxial digital output. If that is carefully designed at both ends, then most of the after-buffer latency would depend on your choice of DAC. RAM buffers and the OS and drivers are the big problems though. Everything else is negligible until those are fixed.
Hi Battle(non)sense! I'm a long time lurker (Now subscriber!) I've always wondered about the feasibility of "Timer Resolution" namely the use of a program called ISLC to manualy set, yes, Timer Resolution to reduce input delay, which isrelated to your recent scope of topics. Is ISLC snake oil or is there legitimate differences in hardware latency to really affect perceived FPS smoothness and Input Delay response.
I think you should have pointed out that when using those DACs, they were at default settings? so they would be on the higher end of the buffer setting. Most DACs allow you to change the buffer size to reduce delay. The GO XLR for example, is default at 512 samples, but can be reduced all the way down to 8 samples. Obviously the lowest sample rate may cause issues, but you can usually find the compromise at around 64 samples or 128 samples which is around the 2 to 4 millisecond output delay.
BLUETOOTH headphones add a lot of audio delay too, and are typically the worst culprit. They can add as much as 300ms of delay! Even the special low-latency "Gaming Bluetooth" headphones add upto 150ms of extra delay (kinda unacceptable for gaming). If you like using wireless headsets, try to make sure they also have an *optional cable jack* so you can plug them in to eliminate the Bluetooth audio-delay when you need to, without needing to keep a 2nd set of WIRED headphones just for gaming.
Very interesting topic, and I'm really surprised about the numbers you get. I'm not gaming on PC so I have no actual experience or knowledge how PC deals with internal sound cards or external interfaces - but those numbers are really huge compared to what you would get in professional music software/hardware integration. Here's a comparison (although might be completely obsolete for this kind of situation): When using Logic Pro with an Universal Apollo interface via Thunderbolt 2 running at 48 kHz sample rate I get the following: 1024 samples buffer size = 49.5 ms delay (roundtrip), 512 = 28.2 ms, 256 = 17.5 ms, 128 = 12.2 ms, 64 samples = 9.5, and 32 samples = 8.2 ms. Switching to 96 kHz the numbers are: 1024 = 25.1 ms, 512 = 14.4 ms, 256 = 9.1 ms, 128 = 6.4 ms, 64 = 5.1 ms, 32 = 4.4 ms. At 192 kHz the numbers are: 1024 = 12.9 ms, 512 = 7.5 ms, 256 = 4.5 ms, 128 = 3.5 ms, 64 = 2.9 ms, 32 = 2.8 ms. You'll probably see similar when you look at your audio settings in Ableton Live? Conclusion should be that there's tons of delay added by the games themselves?
Yep, added by the game engine, and also standard, non-professional audio drivers. He showed that with the right drivers and buffer settings plus proper audio software the delay can be reduced significantly.
@@shopt6933 This is why I avoid PC unfortunately. Things are so complicated (as a user.) : ) Would be interesting to see the same test on consoles. And on PC with “professional grade” audio interfaces perhaps. I suppose also netcode would be part of the audio latency?
@@cornerliston I suspect that on other platforms you will have similar issues, and just have fewer options to deal with them. Game engine induced delay wont disappear on a console, but driver based delays *might*. Although given how consoles these days keep OSes and notifications services running in the background, I doubt the latency path is magically any shorter there.
Audio is unfortunately always left behind compared to graphics/visuals. I had never thought that button to audio delay is higher then button to pixel delay.
From what I read, ASIO is not really all that useful these days anymore since we have WASAPI which you can get equally low latency with and you don't need any additional drivers. But as already pointed out by others, exclusive device access is something probably most enthusiasts won't want since you can't output any sound from other applications. And fixing audio delay on that level is probably not something the gamer will be able to fix with a setting when you have a complicated audio processing framework. That's up to the developers. Though, in the end of the day ~50ms delay more than video delay is probably going to be sufficient in most cases (unless you play a virtual piano / etc.). Veritasium did a video about how much you can tolerate ("What Exactly is the Present?"), highly recommended.
Would like to see is it different in Windows 7 and especially XP, also with a pci sound card and not a pcie. Oh, and another thing - pcie bus frequency overclock may lower the sound latency too.
I believe audio enchantments and sound card features also affect the audio delay. So, if audio enchantments are enabled and the sound card firmware or software does some additional processing it is affecting the audio delay. It can even cause additional input lag if the game itself is badly optimized and the main thread is affected by the audio thread or maybe there is not even an audio thread.
Audio delay sounds like a small problem for most people. But for people who rely on it alot, yes it is actually an annoying significant problem. However, its even more of a problem to rhythm gamers such as myself. I get about 88ms sound delay. 88 ms might not sound alot, but in rhythm games its basically a miss or hit. I hope developers put more effort into sound performance
Great video, been waiting for someone to cover this for quite awhile. Sound and sound cards in general have gotten the short end of the stick for the last decade and a half, basically since EAX was killed off by MS. I also agree about ASIO and not sure exactly why more developers don't care about it and I do believe this is a topic that gamers care about too, they just don't realize they do because they haven't been shown why they should care about it. As far as USB and PCIE go, PCIE is definitely faster, however, it goes to show you the amount of processing that's happening in the devices and the drivers that matters a lot more then the interface it is using. If you could test some other sound cards in the future, that would also be interesting. With all the gimmicks on the market, I do wonder if Creative is still remotely competitive in this regard or if they're just a shell of their former self.
I kept finding myself going back to my onboard sound for gaming often after using various external DACs (audiophile ones) over the last few years, and I think this might be the reason why. I'm very sensitive to latency, and it might have had an impact on my competitive FPS game experience. But I might just be talking rubbish, hindsight and all that.
@@incediumignis to be honest, for gaming a pair of decent gaming headsets is enough for most people. For music listening you might need to go shop somewhere else though 😉
@@MLWJ1993 gaming headsets are trash audio wise, gimmicky af. There are plenty audiophile grade headsets for cheap that dont need amps. Personally got an akg mkii k240 for 55 euros. Relatively inexpensive compared to the "premium" corsair,logitech,hyperX and razer counterparts. And for a better audio quality.
Chris, question. What about an interface like goXLR? my understanding is it would be similar to a USB DAC, but its not really the same as others. Any chance this is testable? Both as say a game PC running realtek 3.5mm into line in as well as USB?
I noticed in VoiceMeeter when i have the digital output: SPDIF ( REALTEK Digitial Output) and the Analog output ( Speakers Realtek High Defenition Audio) both enabled ( Voicemeeter combined with my goxlr app alows me to play both simultaneously) i hear a a slight echo. Now the big question is: What do i hear first? The digital output or the Analog output? Cause the one i hear first i wanna play with! I hope you guys can help me out!
For context, you need an audio delay of 10-15 ms max if you're trying to record a performance on a virtual instrument - any higher than that and the musician's rythm is very likely to get worse, as their muscle memory (trying to play in real-time) fights their instinct to adjust for the latency. This is why rhythm games play pre-determined tracks in the background and rarely involve the player generating sound (and those that do play like garbage).
Nice to see my sound card being tested (Xonar). I didn't get it for latency, but it's still pretty weird to see it slower. Wondering if you were testing the external sound cards with the on-board disabled in BIOS?
Unfortunate... I'll still hopefully be getting a Schitt Modi and Magni to pair with some DT 990s and my HD 280s but I'm sad to see the audio latency go up. Audio is a whole dimension that all too often gets shoved to the way side. Maybe you could compare Windows audio latency w/ and w/o external dacs to MacOS? From my understanding the fundamental way Apple handles audio is much better and why most music producers go to that platform so I'd be curious to see if that leads to significant gains in other use cases. Thank you for shedding light on this topic!
I first noticed the audio delay in Rainbow 6 Siege, but I couldn't find any reasonable solution for the longest time. After this video, I've basically realized that none exists :/
hi @Battle(non)sense Audio-engineer / studio techinician and musician and allround audio fanatic ( and gamer :P ) here... this video confuses me slightly. You are saying that games force your audio interface to change its driver set buffersize?? because i use a RME fireface 802 which has an extreme low latency, i can set it as low as 48 samples, which is about 1ms on 44khz but even lower on higher... however its important to realize that setting your audio to 96khz or even higher is also forcing your pc to work much much harder because the data package would be much larger (perhaps i say it weird, i mean that 1 min audio at 16 bit 44khz is much smaller in size compared to a 32bit 96khz wav file ) so i would not recommend that personally, in fact, 48khz which you claim to be default should not be default, it should be 44k, which is standard cd quality (at 16bit) however, and this is where my confusion comes, when i turn my buffer size down to 48 smp on my RME driver, i do feel the audio delay is smaller (i got a pretty good ear for it) i just now tested it in the apex legends firerange and once again i do feel the audio is waay snappier on 48 smp, but if i believe your video ( and i always believe your videos) it should not matter because apex would force its own buffer size on my audio card... I am very curious to your response on this and would be interested to test it further given the right parameters...
I noticed in VoiceMeeter when i have the digital output: SPDIF ( REALTEK Digitial Output) and the Analog output ( Speakers Realtek High Defenition Audio) both enabled ( Voicemeeter combined with my goxlr app alows me to play both simultaneously) i hear a a slight echo. Now the big question is: What do i hear first? The digital output or the Analog output? Cause the one i hear first i wanna play with! I hope you guys can help me out!
The problem is that you cannot affect what the sound renderer inside the game engine does - how large its buffer size is, which driver/API it uses, etc. By using ASIO for the output device of i.e. VoiceMeeter you can only reduce the delay *added* by VoiceMeeter, not reduce the delay caused by the game engine. Same applies to the ASIO Bridge. However I do plan to make a video on how VoiceMeeter affects the audio delay.
Would be interesting to know if and how much the OS matters? Same sort of tests in linux, since more and more games support linux and you have greater control over all sorts of settings, while probably not for everyone, I'm pretty sure serious gamers would make the jump for better performance.
Hey, Chris! What is your opininon on Voice Meeter Banana? I'm using that with my GSX1200pro to fine tune my sound. Banana can reduce buffer sizes also to reduce latency as far as I know. I'm not an expert tho, I'd love to get your opinion on this.
For me, it's more important to hear footsteps, but also not blow my ears out when a grenade goes off, so i use sounds equalization, and a magewell sound card, works really well.
Missing variables from this analysis worth a revisit: Since audio is compute sensitive (with regard to latency), competitive gamers may want to test lowering bitrate and sample size, in addition to number of channels in hopes of striking the perfect balance of latency/player performance This video while useful does not sacrifice bit depth, sample size, or number of streams and would benefit from analyzing these variables
This explains a lot - i wondered why some games felt like the audio and video felt out of sync - it's because i've ran console ports at 165hz Now to figure out ASIO4ALL, and how to make games use it...
Wondering if streamer devices like GoXLR or even Astro Mixamps for gaming headsets also create that delay as well. Based on your conclusions it seems like they do...interesting. Thanks for creating this!
Yo can you do an escape from tarkov analysis again its been 2 years and its had a whole game engine migration and so much optimization its insane. Id reeally like to know where the netcode is at.
Only came here to hear an German say "buffaah"
_"Get to the buffaaaaaah!"_ 🤣
And "powaah".
but both him and arnold are austrian, not german, which is even a bit better
Yes
Plot twist: you didn’t listen to a German
we need to protect this beautiful, austrian man at all cost, he is a treasure of the whole pc gaming community
Indeed
Indeed.
Protect him from what?
@@hestonsverige Everything and anything. You stay back! You hear me?!
@@hestonsverige from any harm, brother
This guy is changing the gaming industry and that's awesome! Thanks Chriss, I hope some developers listen to this and provide the options.
The main reason why ASIO (or WASAPI exclusive mode) is never used in games is because it's the audio equivalent of exclusive fullscreen mode. the application detaches the OS sound mixer and sends the audio signal directly to the hardware, which means you can't hear anything else (just like you can't see any other windows while in exclusive fullscreen, except for overlays that intercept the render pipeline).
Also one important note about LDAT: depending on what audio equipment you use, the output device itself can actually cause a notable amount of audio latency which LDAT doesn't reflect (just like debounce or pixel response time, but much bigger)
WASAPI has non exclusive mode.
There are multi-client ASIO drivers that allow ASIO apps and other apps to be heard simultaneously.
@@AuxHex i know, thats why i specifically said "WASPAPI exclusive mode"
@Howard Nyugen but for that all the applications would have to implement ASIO, dont they?
Edit: i misread. If that's actually possible that would be really cool and i'd wonder why it isnt made use of
@@LastExceed These days, there are many options and workarounds for generic exclusive-mode Asio drivers such as Voicemeeter (you can route less important audio at a delay compared to the asio source) and other multi-client Asio interfaces.
As one example, some DAWs like FL Studio actually come built-in with a high quality non-exclusive Asio driver, allowing everything else in the Windows mixer to come through while still providing low latency audio inside the DAW at the same time. While it isn't _quite_ as low latency compared to standalone exclusive asio drivers, it's still a huge improvement nonetheless, especially for the convenience factor.
If games developers had the know-hows to write an audio driver like FL's Asio, or companies like THX or Dolby developed their own non-exclusive asio standard for developers to license, we could theoretically have huge latency improvements.
That is correct, Steinberg's ASIO has the "downside" that only one application can use the sound card at a time. So if you listen to music or use discord at the same time then ASIO is not a solution - *this is why I mentioned that in the context of supporting it in competitive games* where players usually only want the game audio with no distractions - ofc voice communication might still be an issue here if an additional application like discord is used.
That is where WASAPI could come in handy as an alternative to ASIO. I will see if I can produce some testresults using that driver/api. :)
I've never even thought about audio delay in a competitive environment before. Seeing those huge differences in audio delay can mean a world of a difference when trying to react to something like Reinhardt's "HAMMER DOWN" audio or Zarya's graviton and such. Amazing detailed video and topic! Great stuff.
Sound is a lot more complex topic than video delay. The way humans perceive sound is very weird, audio and video delays allow us to judge distances. For our bodies light travels instantly, it's too fast to notice any delay between something happening in front of our face and something happening across the street. But sound WILL have a delay. And our brains use this difference to adjust perception of distance. This leads to two interesting effects.
First of all, if sound is coming from a fixed point in space and doesn't change distance, then our brains can adjust to a very high sound latency, up to 200ms or so and everything will sound natural unless you're looking for this delay specifically.
Second effect is that in a dynamic environment like a competitive game, sound effects should not only be modelled between left and right ears, but also in time depending on a distance from the player. Otherwise the game sounds "flat" and unnatural. Which is the case for most games of today. And while we're adding distance based delay, we should also add correct Doppler effect emulation. DE + delays together create a fully immersive 3D picture of the in-game world without which in-game sound is nothing more than a blinking icon in a status bar.
_"The way humans perceive sound is very weird"_
That might be so, yet it is also a fact that we already perceive sound as out of sync at a difference of 45ms, which is well known among those who have to deal with audio to video synchronization (i.e. lip sync, ...).
The audio delay in games is design to not be *ahead* of the image - and that intentional audio delay is based on 60Hz v-sync where you have an average B2P of > 100ms.
On console we still do not have an issue (maybe with next gen as that supports 120Hzs an VRR).
But on PC where we have reduce the B2P to less than 20% of the B2A, we are well past the threshold of 45ms where humans notice that image and audio are out of sync.
The other aspect is competitive gaming where game audio is an important cue - 20ms instead of 150ms delay is a noticeable improvement, especially on a high level of play.
In story driven games it is less of an issue, unless the delay is noticeable when it comes to lip-sync - if the animations and character models are detailed enought for that.
This video demonstrated a 36ms to 176ms+ audio delay for things like gunfire that happen immediately close to the observre. That is in no way a realistic emulation of sound. The fact that such a difference exists indicates that a very large portion of that delay is unintentional and not intended to be an accurate simulation of sound delay during travel. If the actual difference between audio hardware is up to 10 ms as demonstrated by the ASIO driver, you should not be seeing a 140ms swing.
Humans perceive sound based off:
1) the distance of the observer from the sound source
2) The difference in intensity / delay between your two ears.
I know of very few games that simulate the difference in intensity of the same sound between two ears. Overwatch has a feature called Dobly atmos (which perhaps should also be tested) which does positional sound better than many games but it is still a far cry from the technology used in VR titles that make use of HRTFs (or a game like Senua's sacrifice).
It's really a shame that audio delay is so high when it could be drastically lower and for that matter, much more immersive.
Lovely! Many thanks for the info provided and introducing me to ASIO.
IMO absolute delay values are to be taken with a grain of salt, coming from video games.
Sound design is not checked as thoroughly, as other disciplines of a game. It is common to insert zero-samples at the beginning of a sound clip to prevent popping and crackling. Next of, having a too fast of an attack in a sound effect may make it sound unnatural. As such, a slower attack may be chosen for a more pleasent sound effect. This may result in the sound staying longer below the noise floor and thus a couple ms will be added by the LDAT measurement. So a higher LDAT measurement may not be an engine limitation, but a choise by the sound designers, prioritising how it sounds, as opposed to a perfect sync, especially with consoles hooked up to TVs with very high button-to-pixel delays.
The relative delays are of course rock solid and this video is an exellent source of info!
edit: In a perfect world, a sound mod would be an interesting thing to make: Replace all sounds of a game with a square wave going instantly to 100% volume for a cleaner measurement.
Yes, (as I mentioned) considering that at 60Hz with v.sync the Button2Pixel delay is over 100ms, the audio delays of >120ms are in the safe spot to prevent that the audio gets *ahead* of the video.
On console we currently do not have an issue _(though that might change with the 120Hz support of "nextgen)_ .
On PC the situation is different as we have dramatically lowered the B2P delay during the last few years, increasing the rift between video and audio.
@@BattleNonSense Sadly, we do not have a "gaming" oriented audio company to develop low latency audio technologies or standards for games :(
It feels like we are going to be doomed to high latency audio in games for a very long time.
@@GENKI_INU We don't need gaming oriented audio companies. The standards already exist in the professional audio world, mainly tunable buffers, ASIO, and realtime-priority audio threads. I use a Native Instruments Komplete DJ mixing interface for my headphones/microphone, which runs at a internal buffer of only 192 samples, compared to the 768 tested in the video, and output buffer delays
@@henryzhang7873 We do need new standards. Not everyone owns an audio interface or something that even supports ASIO. Game engines also have different pipelines and different needs to suit them better with their workflow, different from what producers need. If Realtek audio chips in the future came with an ASIO-like standard that developers started mass adopting, or if AMD and Nvidia's graphics cards focused on their built-in audio with low-latency game drivers that developers adopted, that would benefit everyone.
@@GENKI_INU yeah but then the drivers should use ASIO. ASIO isn’t a driver but also named after a programming technique. It’s the idea that as soon as you know the content of a sample, you pass it immediately to the sound card with minimal buffering (more complicated than this, but this is the important part), using special OS or hardware systems. This reduces delay because you are not waiting for more samples to accumulate. Not only audio uses it: I use it in the web servers I write to have the best possible latency. It’s just hard to program for because humans are not used to dealing with asynchronous streams as a concept. Developers that know how to do this get paid way more working in the tech space than gaming.
The server you are watching this video off of is likely also based on ASIO principles because Google/YT has been pushing these kind of event driven paradigms for years.
beyer dynamic DT770 pro, a man of culture I see
I think this is really interesting, and something I have experienced in games myself.
Games like Escape From Tarkov, Insurgency: Sandstorm and CoD MW, uses sound as awareness mechanic, and I find that when playing those games the sound lags behind when you turn sharply.
Would you consider testing some of the games that use sound as an awareness mechanic?
Thanks for an interesting video!
I plan to do Button To Pixel and Button To Audio delay tests in all upcoming "Input Lag" test videos.
@@BattleNonSense wow that is really great of you, I think of supporting you through Patreon. the work and research you put in your videos is amazing
I would second sound tests or more tests in general include COD:MW2019 . Its sound is great quality wise... but is seemingly broken in many ways so it doesn't deliver in Multiplayer how it has in the past (near perfect for player ability to use it as intended as spacial awareness assistance etc).
I've started experiencing audio delay after upgrading to a TCL c635. The delay is only in 60fps games and more noticeable when going from stereo uncompressed to DA for headphones or any of the other Dolby formats. I can't understand what causes it and don't know if getting a standalone audio system will get rid of the audio lag
Its simple you need a RGB Headset with at least 256 different colors to reduce your audio delay.
As someone interested in rythm games, this is very enlightening. I did not expect the usb Sound cards to be that bad in terms of latency but feel reinsured that sticking with 48kHz and pushing the buffer size down in pipewire was the right choice.
Thanks for doing the empirical test and presenting the results in such a nice way!
I knew I wasn't going crazy when some games felt more out of sync than others regarding weapons visually firing and producing the gunfire sounds. Sucks to see there isn't any reasonable solution to reducing it beyond what I'm already doing. Thanks for the video!
My X-Fi Audigy 2 sound card used to have a "creator mode" which seemingly drastically reduced audio latency across the system. Problem is that card came out in 2002 and is PCI (not PCIe) only, and ever since then I've been bothered by audio latency. It really sucks that it's such a hugely overlooked part of the equation, with no good solutions available.. I hope that some good solution comes out of this!
At the end where he's talking about whether people really want lower delay, and mentions footsteps and such, I would just like to mention that rhythm game players have a huge stake in audio delay being lower.
There's a program for windows called REAL (reduce audio latency) that forces the buffer to be lower than default of 10ms, on my usb dac it defaults to 7ms which works well for me. I've tried my motherboard built in audio, but my onboard sound uses 3ms buffer which is broken in every game
Pro tip: If you use the Windows 10/11 "High Definition Audio Device" audio driver, that's the default driver for a fresh install for all chipsets, the latency is insanely low. With Realtek S1200A driver I get 95ms latency. With High Definition Audio Device driver I get between 30 and 38ms. With Astro A40 mixamp I get 110ms.
do you know a way to keep it, because windows instantly update the driver to realtek, i always felt that games run very well after a new windows installation but i didn't know why
@@hulknoob2323 you can just disable driver updates from Windows Update, in advanced system settings
Oh gosh please please please run this test with osu!. Audio delay is very important in rhythm games, and the osu! developer seems to believe that you'll only have a delay >10ms if something is wrong with your setup. It would be amazing to have some concrete evidence to support adding ASIO driver compatibility to the game.
Could you perhaps play around with Voicemeeter Banana in the future? It supports pushing audio through ASIO, as well as has adjustable buffer size both for the sources, and its internal virtual audio cables (via the .exe's contained in program's folder). I'd love to see not only whether and how much Voicemeeter can increase latency, but also whether adjusting its buffer scaling could actually improve latency over standard Windows audio mixing.
Thank you very much for this video! An interesting topic, to be sure.
The increased audio latency he's recording in his testing is not related to Windows audio mixing. It's related to the buffer size every game is set to have. So from my understanding, I would say Voicemeeter Banana can only maintain the same latency or increase it, but definitely not reduce it.
I have done setup guides for VoiceMeeter (Banana) and I'm using it on a daily basis, but I do plan to look into the audio delay in a separate video.
Early testing _(with reduced WDM buffer size as well as ASIO)_ was not promising - though I want to look into it more to make sure I don't miss anything. :)
I second this!
I am pretty confident Voicemeeter always adds input delay
@@BattleNonSense I would guess it would only add latency because the game will still be using the WDM/MME audio output and THEN sending it to the ASIO driver via VoiceMeeter.
I'm so happy that I randomly came across this channel yesterday. Such an amazing content!
Love this content, sir. Keep up the great work!
Audio delay's a huge deal in the rhythm game community
We discovered ASIO a loong time ago and are actively looking to get every single game to support it
Would be interesting to know how the delays are if the sound is transfered through hdmi and the headphones are connected to the monitor or are the sound drivers on the graphic cards exactly the same.
hhhmmmmm that is indeed a very interesting question!
@@BattleNonSense yeah, I use HDMi to AV receiver and I was wondering what is the delay in this case. P.S: thanks for your videos.
@@BattleNonSense I was thinking of this as well, I use external EMU 0404, which has digital input as well as USB connection, I use the USB connection for ASIO, and for other uses than music creation, I route the sound from the integrated soundcard using S/PDIF into my EMU 0404.
Would be interesting to see the different digital options such as digital output from integrated sound card, integrated graphics card and PCI-e graphics card. Also if you do end up testing digital audio, does it make a difference if the output is HDMI or Displayport, and does it make a difference if the DAC is a cheap monitor one or something more expensive like a professional/entertainment soundcard or like Andrej mentioned AV receiver.
@@BattleNonSense I also use a USB microphone, namely the AT2020, so I'm wondering if using heaphones through there is a bad idea. I'd imagine it is considering how Dacs cause delay.
Yea. I use hdmi for audio but I used to use on board audio a couple years ago becuase I noticed more audio lag when using hdmi versus on board audio. My thought process as to why on board audio was less had to do with audio technically coming off the cpu closer to the metal versus hdmi which had to convert the signal again. Also. It is a similar using for external docs that have to process audio a second time.
I used to use hdmi for video and on board audio using analog stereo connections to my headphones or to analog only speakers. If I plugged in an analog stereo plug to certain speakers that accept digital signal certain speakers re-digitize the analog signal before sending it to the speaker.
Also, you didn't test if anything changed from using audio exclusive mode in windows versus non exclusive. I had games that had more audio lag with exclusive on and some games that had more audio lag with exclusive off.
For anyone using Linux: RetroArch has a JACK Audio (the equivalent of ASIO) driver built-in. I've achieved ~7ms of audio latency using it and it feels great.
Hi amazing video (as always) I noticed that they were an audio delay in Overwatch, I hope your video will have an impact on this topic like the one on Gpu bound (you invented Nvidia reflex, congrats). As I have a focusrite sound card I know an issue about ASIO, you need to have all application at same audio frequencies or you'll have heavy sizzling or no sound at all but it's a pretty good idea latency AND quality wise as it bypasses windows. Very good channel you make competitive gaming better.
@ Battle(non)sense
Chances of going back and owning Tarkov netcode again? I know it's rough. I heard there may have been dev drama with you and that's why it hasn't been done? But man it's been over 2 years since your videos AND they haven't done a 'network' patch since then either. Feelsbadman. We NEED you to wake them up again! Love you!
There is a hybrid mode which allows to run both ASIO and WASAPI without taking exclusive control over the audio device (at the cost of it not being a complete ASIO). For example, FL Studio ASIO, included with the latest versions, can go as low as 62 ms, although it's still pretty high.
But it still sucks that there is a fast AF boi called UR12 that can drop the audio delay to fractions of milliseconds if the system can handle it (minimum 4 samples at 44.1kHz per stream window). This mode takes exclusive control of the device though, yet I'd love to pay the price of being able to only hear the game just to experience nearly instant audio response.
EDIT: Also, ASIO reduces CPU time spent on processing audio by a ton. That might be the more important in DAWs, since digital processing can be quite cumbersome even for very fast CPUs.
This video opened my eyes to this issue, but has left me with questions. What about SP/DIF Optical audio output? What about using a dedicated headphone amp connected to the line out on the motherboard?
In theory using line out to an amp should have very little delay added since the amp won't be processing the audio, but only testing will confirm.
SP/DIF I would love to know about as well to know what delay it adds compared to usb and pci-e.
This video makes me hope the topic of audio delay begins to receive the attention it needs in the industry.
I am using a Creative AE-9 PCIe soundcard in "Direct Mode" and your video inspired me to compare the latency to my onboard ALC1220. I am using your technique and am testing in a Modern Warfare (2019) Custom Lobby. I collected ~8 samples from for each soundcard and both are consistently at around 110ms latency. I also tested in normal mode, and it was around ~115ms, but much less consistent.
Same delay onboard and pcie? That conclusion would differ from @battle(non)sense
@@tawl-x Yes, same delay with onboard ALC1220 and PCIe. You have to consider that the Creative AE-9 is a rather expensive high end sound card, maybe they figured out how to do it without delay. Not all sound cards are equal. I dont think battlenonsense's results are false.
Using wasapi give exclusive audio access to the game and usually have waaaay lower latency. However that obviously doesn't play well with discord and other voices apps.
Would still be cool for games to give the option though.
(The only "game" I know that gives that option is the Dolphin Emulator)
This man is a gift, protect him at all costs.
You should have buffer size on the soundcard driver as well. By default this will be set pretty high.
I'm not a competitive gamer (far from it) but this audio delay has been bothering me for years. From the sound confirming a button press in a menu to sound effects in game, it's simply annoying when it doesn't sync up.
Audio processing has a realtime priority in os schedulers for a reason. Was a much bigger problem when cpus were like 1/20 the speed. Honestly don't remember when I last experienced an audio sync issue. Must have been over 4 years ago.
Also something I don't think mr. battlenonsense is aware of is that you need at least some sort of audio delay. If audio cues are instantaneous ie sub 1ms it can be disorienting due to delays in how your brain processes an input with the corresponding sound. It'll seem like the sound occurred a fraction of second before you pressed a button as nothing we do on computers is really capable of doing this. You can of course get used to it though. However with the ~30ms in a best case scenario he measured I don't think that'll be a problem.
This aligns with my experience between a PCIe sound card vs motherboard sound, the sound card had about 10-15ms longer delay, indicated by the need to adjust timing offset in a rhythm game.
Audio and video desync however, especially just delayed audio, is not easy to perceive, and usually won't result in weird experience like input-to-visual lag does. Sound travels slowly, humans are used to hearing sounds delayed. But I can see how reacting to sounds entirely will result in a competitive advantage.
This video covers a topic I’ve always wanted to see!!! Thank you!!!
Excellent research on a largely overlooked topic with suprising conclusions. Great work, thanks!
You are just truly such a nice guy for making videos like this. Just insane! Love it!
if you ever revisit this, could you test some Optical and other digital audio out setups to external DAC's?
The reason the ASUS STX card is laggier is due to PLX converter chip (The DSP used is native PCI inteference). You should test a native PCI-E sound card. Results should be in-between i2c Realtek and USB2.0.
NOTE: Modern Realtek 4080/4082 has been converted to a USB 2.0 interface vs legacy I2C in ALC1220. Latency should be closer to the FiiO and other USB card you tested. Maybe inbetween since its a more direct path to chipset or CPU depending on AMD/INTEL platforms.
I just want people to be aware that the results are likely skewed from what is "possible" on PCI-E.
Old creative cards like Titanium HD (Native PCI-E) were always the fastest for delay back around 2010 era.. IMO. Realtek had terrible drivers in that time frame. (Creative was bad too, but "faster")
Maybe that's changed.. cant tell you since I use a 2.0 interference these days :D
Please do a video on voicemeeter, it’s used by most streamers, and allows for passing game audio via ASIO at lower buffers like 64!
i'd like to recommend audio interfaces like the focusrite scarlet 2i2, behringer umc 204hd, steinberg ur22 or similar ones. those interfaces are something between a external sound card and a small mixing table. you can get really decent ones for about 60-100€.
as guitarist i need a very low audio output and input delay since i play my guitar "through the pc" instead of an amp for recording. any kind of delay between hitting the strings and the processed signal is super obvious, so using ASIO is the only way to go here. but even when not on asio, you can reduce the buffersize in the driver to much lower amounts than on regular sound cards without crackling/ destroying the sound. plus they offer some nice features like different volume controls, headphone amp, mic preamps, outputs and connectivity options.
wäre noch interessant gewesen zu wissen wie es aussieht mit Soundkarten-Optionen wie Virtual Surround, Crystalizer und ähnlichem.
DT 770 Pro gang here, thanks for this vid. I feel sound delay, it's pretty annoying
Man I love these headphones. My Sennheiser Game Zero died recently and I almost bought another set of "gaming" headphones.
The DT 770 *with* a mod mic was not only cheaper, it also sounds sooooooo much better!
@@BattleNonSense When the original pads will start to collapse, check out Dekoni Elite Velour pads! They're made out of memory foam that's muuuuch more comfortable and they boost a bit the sub bass (I needed to actually lower the 30 and 60 frequency bands by -2 and -1 db via my sound card EQ after installing them)
@@BattleNonSense I use audio technica m50xBT with a mod mic. Its just great.
@@BattleNonSense Hey, can you test Linux (ALSA or JACK, PulseAudio has a delay) or MacOS? They should have less audio latency (comparable to ASIO on Windows). CSGO should be a good game for that, it has native support on MacOS an Linux
I have DT 770 Pro 250 ohm with IFI ZEN DAC V2, blue snowball mic. I download APO and PEACE EQUALIZER, with DOLBY ATMOS. Still trying to figure out if HESUVI will be better than ATMOS, or an other one. I dont care to spend more money to get the best VSS. Im playing WARZONE. Somebody have a recommandation? Thanks in advance 🙏 @Battle(non)sense
3:34 the difference between 60fps and 360fps is about 14ms. so that's probably why it is 12ms faster in that case.
yeah. there is 0 gain despite the initial latency gain you always get from higher framerates.
Dude I love every single video you make ... every video you do gives answers to questions I couldnt find answers for anywhere else
Whenever I think that there are many misconceptions, myths and misunderstood concepts in gaming, I only have to remind myself there's always audio stuff ;) So not surprised with the results in the slightest.
Great video! Love to see the proof of what you're saying and not the internet know-it-alls that just say it's a certain way without proof. Keep it up!
I've been curious of lowering audio latency for a while and I mostly agree with your findings. Onboard sound chip have the lowest latency, and external gaming DAC and internal gaming sound card have more latency. This is why I sold my Creative Sound Blaster Omni, and I changed into using regular external DAC (Topping E30) + headphone amp. Its latency is a bit slower than onboard sound chip by few milliseconds, but increasing sample size from 48khz to 384khz did improve latency by around 3ms. It might matches with onboard sound chip but haven't gone into testing it yet.
But my twist is I'm not a competitive gamer, and I prefer using AV receiver + speakers. Audio latency became more complicated as it varies from different brand. For my case, two best solutions for lowest 7.1 sound latency is using HDMI connection (+25ms) or multi channel analog input (+0ms using onboard sound chip).
One final note is bitstreaming audio sucks for latency. I tried Dolby Digital Live and DTS connect in the past and latency is as bad as around +100ms. And I cringe when Dolby Atmos is becoming more of feature, as it is also bitstreaming audio and have poor latency too. I rather use 7.1 audio then use upmixer to 7.1.4, and although upmixer add some latency it is still much better than Dolby Atmos.
I usually play better when I don't use any audio. Never knew why, until now.
Thanks for your work!
Wish we can see how much the audio is delayed by using GoXLR - quite popular among a lot of gamer / streamers.
I have tried some 3rd party sound enhancers to make footsteps and directional sound more easy to point out.
But the audio delay with processing the sound in the software was to great to be useful. This video really shows how important that is.
If anything, I'm glad to know that I'm not crazy about the audio delay and that it being around 100 ms is normal.
I kept noticing that if I slide my headset half-off, I could very obviously hear the difference between the click and the gunshot. I play guitar through my focusrite scarlett interface using the ASIO driver, that's where I have my headphones plugged in, and the latency is completely unnoticeable, so I just assumed it couldn't be my audio hardware. The mouse also felt perfectly fine so it couldn't be my mouse's input latency.
everybody at respawn studios just died laughing
why?
@@gaboqv apex has some of the worst audio queues and audio dumps imaginable lol
What about Creative sound cards?
Are they as bad as the others?
Would be interesting, to see how an external DAC using an optical connection compares to usb connected DAC. I doubt it would make much of a difference, because your still adding an extra processor processor in there, but it would be interesting nonetheless.
One thing to consider, when deciding between using the onboard DAC and an external DAC, is the potential gains in clarity, sound stage, and imaging (where sounds sound like they are coming from), this is also dependent on the quality of the headphones, that you are using.
For reference (Take this with a grain of salt.): I use to run my headphones and my speakers through the onboard audio on my Gigabyte Aorus Gaming 7 (Z370) Mother board. When I upgraded my audio setup I to a SMSL SU-8 (DAC) / THX 789 (AMP) setup and DT1990 headphones (I tried the headphones with the on board audio when I first got them), their was a notical improvement in sound stage, imaging, and clarity, from not only the headphones, but also the improvement in clarity and richness of the sound produced by my Klipsch Promedia 2.1 THX speakers seems vastly improved.
PC -USB> DAC -XLR> AMP -Single ended 1/4> DT1990
PC -USB> DAC -RCA> Promedia 2.1
With that noted, I will gladly give up a little bit of sound latency, (Especially when I don’t perceive the change) for the bump in clarity and sound stage.
Optical would add delay, since there are two opticalelectrical modem devices/operations. What I think you want is coaxial digital output. If that is carefully designed at both ends, then most of the after-buffer latency would depend on your choice of DAC.
RAM buffers and the OS and drivers are the big problems though. Everything else is negligible until those are fixed.
Optical can only be worse, it's adding more steps with encoding and decoding
nice blind A/B testing
Great bit of research, with some unexpected results. Interesting stuff, thanks.
Hi Battle(non)sense! I'm a long time lurker (Now subscriber!)
I've always wondered about the feasibility of "Timer Resolution" namely the use of a program called ISLC to manualy set, yes, Timer Resolution to reduce input delay, which isrelated to your recent scope of topics.
Is ISLC snake oil or is there legitimate differences in hardware latency to really affect perceived FPS smoothness and Input Delay response.
Any chance you could revisit Tarkov? They've had some updates to mitigate issues (according to them). Interested in learning more!
Very high production quality and easy to follow, as always awesome indepth video! I would have never thought that the audio delay was that high.
2 hours ago I was just trying to hear footsteps better in cs. Now I'm an audio engineer
If you would want to test the buffer sizes in another game, you could try WoW or D3. They both have 3 MB size steps for audio buffer size.
Did you test MessageInterrupt priority from normal to high for the audio cards?
Nice video! It would be super interesting how high the audio delay in the rythm game osu! is as that actually benefits from low audio delay a lot.
I will see if I can test that - requires that the game makes a sound on mouse click
@@BattleNonSense You can probably find a test map that would allow you to conduct the testing.
@@BattleNonSense That should be no problem!
Just set music to 0% and you should be able to only hear a click sound when you click a circle.
Keep up the good work my man!
Love the content!
I think you should have pointed out that when using those DACs, they were at default settings? so they would be on the higher end of the buffer setting. Most DACs allow you to change the buffer size to reduce delay. The GO XLR for example, is default at 512 samples, but can be reduced all the way down to 8 samples. Obviously the lowest sample rate may cause issues, but you can usually find the compromise at around 64 samples or 128 samples which is around the 2 to 4 millisecond output delay.
isn't that ASIO
BLUETOOTH headphones add a lot of audio delay too, and are typically the worst culprit. They can add as much as 300ms of delay! Even the special low-latency "Gaming Bluetooth" headphones add upto 150ms of extra delay (kinda unacceptable for gaming).
If you like using wireless headsets, try to make sure they also have an *optional cable jack* so you can plug them in to eliminate the Bluetooth audio-delay when you need to, without needing to keep a 2nd set of WIRED headphones just for gaming.
Very interesting topic, and I'm really surprised about the numbers you get.
I'm not gaming on PC so I have no actual experience or knowledge how PC deals with internal sound cards or external interfaces - but those numbers are really huge compared to what you would get in professional music software/hardware integration.
Here's a comparison (although might be completely obsolete for this kind of situation):
When using Logic Pro with an Universal Apollo interface via Thunderbolt 2 running at 48 kHz sample rate I get the following: 1024 samples buffer size = 49.5 ms delay (roundtrip), 512 = 28.2 ms, 256 = 17.5 ms, 128 = 12.2 ms, 64 samples = 9.5, and 32 samples = 8.2 ms.
Switching to 96 kHz the numbers are: 1024 = 25.1 ms, 512 = 14.4 ms, 256 = 9.1 ms, 128 = 6.4 ms, 64 = 5.1 ms, 32 = 4.4 ms.
At 192 kHz the numbers are: 1024 = 12.9 ms, 512 = 7.5 ms, 256 = 4.5 ms, 128 = 3.5 ms, 64 = 2.9 ms, 32 = 2.8 ms.
You'll probably see similar when you look at your audio settings in Ableton Live?
Conclusion should be that there's tons of delay added by the games themselves?
Yep, added by the game engine, and also standard, non-professional audio drivers. He showed that with the right drivers and buffer settings plus proper audio software the delay can be reduced significantly.
@@shopt6933 This is why I avoid PC unfortunately. Things are so complicated (as a user.)
: )
Would be interesting to see the same test on consoles.
And on PC with “professional grade” audio interfaces perhaps.
I suppose also netcode would be part of the audio latency?
@@cornerliston I suspect that on other platforms you will have similar issues, and just have fewer options to deal with them. Game engine induced delay wont disappear on a console, but driver based delays *might*. Although given how consoles these days keep OSes and notifications services running in the background, I doubt the latency path is magically any shorter there.
@@shopt6933 I think you're right. Without any actual testing but only by judging while playing the console has quite some delay added.
Audio is unfortunately always left behind compared to graphics/visuals. I had never thought that button to audio delay is higher then button to pixel delay.
From what I read, ASIO is not really all that useful these days anymore since we have WASAPI which you can get equally low latency with and you don't need any additional drivers. But as already pointed out by others, exclusive device access is something probably most enthusiasts won't want since you can't output any sound from other applications. And fixing audio delay on that level is probably not something the gamer will be able to fix with a setting when you have a complicated audio processing framework. That's up to the developers.
Though, in the end of the day ~50ms delay more than video delay is probably going to be sufficient in most cases (unless you play a virtual piano / etc.). Veritasium did a video about how much you can tolerate ("What Exactly is the Present?"), highly recommended.
Would like to see is it different in Windows 7 and especially XP, also with a pci sound card and not a pcie. Oh, and another thing - pcie bus frequency overclock may lower the sound latency too.
I believe audio enchantments and sound card features also affect the audio delay. So, if audio enchantments are enabled and the sound card firmware or software does some additional processing it is affecting the audio delay. It can even cause additional input lag if the game itself is badly optimized and the main thread is affected by the audio thread or maybe there is not even an audio thread.
Audio delay sounds like a small problem for most people. But for people who rely on it alot, yes it is actually an annoying significant problem. However, its even more of a problem to rhythm gamers such as myself. I get about 88ms sound delay. 88 ms might not sound alot, but in rhythm games its basically a miss or hit. I hope developers put more effort into sound performance
Great video, been waiting for someone to cover this for quite awhile. Sound and sound cards in general have gotten the short end of the stick for the last decade and a half, basically since EAX was killed off by MS. I also agree about ASIO and not sure exactly why more developers don't care about it and I do believe this is a topic that gamers care about too, they just don't realize they do because they haven't been shown why they should care about it.
As far as USB and PCIE go, PCIE is definitely faster, however, it goes to show you the amount of processing that's happening in the devices and the drivers that matters a lot more then the interface it is using. If you could test some other sound cards in the future, that would also be interesting. With all the gimmicks on the market, I do wonder if Creative is still remotely competitive in this regard or if they're just a shell of their former self.
How about when you hook headphones into an external monitor and run the GPU's audio output from DP or HDMI?
That was my thinking, too. What about putting the sound out from the gpu over hdmi or dp?
I imagine that the delay would be higher bc there is more "stuff" in the way, but it would be good to know
I kept finding myself going back to my onboard sound for gaming often after using various external DACs (audiophile ones) over the last few years, and I think this might be the reason why. I'm very sensitive to latency, and it might have had an impact on my competitive FPS game experience. But I might just be talking rubbish, hindsight and all that.
good luck doing that if you use a proper audiophile headset that requires more power than the motherboard can supply.
@@incediumignis I was willing to make the compromise and go with an easy to power headphone.
@@incediumignis to be honest, for gaming a pair of decent gaming headsets is enough for most people.
For music listening you might need to go shop somewhere else though 😉
@@MLWJ1993 gaming headsets are trash audio wise, gimmicky af. There are plenty audiophile grade headsets for cheap that dont need amps. Personally got an akg mkii k240 for 55 euros. Relatively inexpensive compared to the "premium" corsair,logitech,hyperX and razer counterparts. And for a better audio quality.
Does motherboard optical output perform the same as onboard audio?
Chris, question. What about an interface like goXLR? my understanding is it would be similar to a USB DAC, but its not really the same as others. Any chance this is testable? Both as say a game PC running realtek 3.5mm into line in as well as USB?
I noticed in VoiceMeeter when i have the digital output: SPDIF ( REALTEK Digitial Output) and the Analog output ( Speakers Realtek High Defenition Audio) both enabled ( Voicemeeter combined with my goxlr app alows me to play both simultaneously) i hear a a slight echo. Now the big question is: What do i hear first? The digital output or the Analog output? Cause the one i hear first i wanna play with! I hope you guys can help me out!
its prob the digital one that has the delay. from what i understand from reddit.
For context, you need an audio delay of 10-15 ms max if you're trying to record a performance on a virtual instrument - any higher than that and the musician's rythm is very likely to get worse, as their muscle memory (trying to play in real-time) fights their instinct to adjust for the latency.
This is why rhythm games play pre-determined tracks in the background and rarely involve the player generating sound (and those that do play like garbage).
Amazing video, first time watching your content. You got a new subscriber :)
Nice to see my sound card being tested (Xonar). I didn't get it for latency, but it's still pretty weird to see it slower. Wondering if you were testing the external sound cards with the on-board disabled in BIOS?
Who wound dare to conter argumet you? Its so well researched! Nice job!
Unfortunate... I'll still hopefully be getting a Schitt Modi and Magni to pair with some DT 990s and my HD 280s but I'm sad to see the audio latency go up. Audio is a whole dimension that all too often gets shoved to the way side. Maybe you could compare Windows audio latency w/ and w/o external dacs to MacOS? From my understanding the fundamental way Apple handles audio is much better and why most music producers go to that platform so I'd be curious to see if that leads to significant gains in other use cases.
Thank you for shedding light on this topic!
I'm also using a Modi and now I consider switching to board audio for gaming -.-
Don't worry about that, really. I have an external DAC as well, and the latency is not noticeable
@@KALTBLUTWOLF1 when there's that much latency already slightly more is hardly going to matter if you feel audio quality is lacking.
I first noticed the audio delay in Rainbow 6 Siege, but I couldn't find any reasonable solution for the longest time. After this video, I've basically realized that none exists :/
@@iPlayOnSpica Use onboard sound?
hi @Battle(non)sense
Audio-engineer / studio techinician and musician and allround audio fanatic ( and gamer :P ) here... this video confuses me slightly.
You are saying that games force your audio interface to change its driver set buffersize??
because i use a RME fireface 802 which has an extreme low latency, i can set it as low as 48 samples, which is about 1ms on 44khz but even lower on higher... however its important to realize that setting your audio to 96khz or even higher is also forcing your pc to work much much harder because the data package would be much larger (perhaps i say it weird, i mean that 1 min audio at 16 bit 44khz is much smaller in size compared to a 32bit 96khz wav file ) so i would not recommend that personally, in fact, 48khz which you claim to be default should not be default, it should be 44k, which is standard cd quality (at 16bit)
however, and this is where my confusion comes, when i turn my buffer size down to 48 smp on my RME driver, i do feel the audio delay is smaller (i got a pretty good ear for it)
i just now tested it in the apex legends firerange and once again i do feel the audio is waay snappier on 48 smp, but if i believe your video ( and i always believe your videos) it should not matter because apex would force its own buffer size on my audio card...
I am very curious to your response on this and would be interested to test it further given the right parameters...
I noticed in VoiceMeeter when i have the digital output: SPDIF ( REALTEK Digitial Output) and the Analog output ( Speakers Realtek High Defenition Audio) both enabled ( Voicemeeter combined with my goxlr app alows me to play both simultaneously) i hear a a slight echo. Now the big question is: What do i hear first? The digital output or the Analog output? Cause the one i hear first i wanna play with! I hope you guys can help me out!
I wonder what the differences are between wireless, 3.5mm , usb added to the delay already experienced in game🤔
Great video, man!
You can make any application use ASIO by using a tools like VB-Audio ASIO Bridge. It also supports setting different buffer sizes
The problem is that you cannot affect what the sound renderer inside the game engine does - how large its buffer size is, which driver/API it uses, etc.
By using ASIO for the output device of i.e. VoiceMeeter you can only reduce the delay *added* by VoiceMeeter, not reduce the delay caused by the game engine.
Same applies to the ASIO Bridge.
However I do plan to make a video on how VoiceMeeter affects the audio delay.
Love scientific tests and gathering all the samples to make it repeatable.
Would be interesting to know if and how much the OS matters?
Same sort of tests in linux, since more and more games support linux and you have greater control over all sorts of settings, while probably not for everyone, I'm pretty sure serious gamers would make the jump for better performance.
I care immensely, actually!
Hey, Chris! What is your opininon on Voice Meeter Banana? I'm using that with my GSX1200pro to fine tune my sound.
Banana can reduce buffer sizes also to reduce latency as far as I know. I'm not an expert tho, I'd love to get your opinion on this.
Try to measure VoiceMeeter with and without ASIO as physical output!
Great Video. You should look at Nvidia Broadcast to see if it causes sound delay and how much.
For me, it's more important to hear footsteps, but also not blow my ears out when a grenade goes off, so i use sounds equalization, and a magewell sound card, works really well.
@@wtssmaller i men loudness equalization
Missing variables from this analysis worth a revisit:
Since audio is compute sensitive (with regard to latency), competitive gamers may want to test lowering bitrate and sample size, in addition to number of channels in hopes of striking the perfect balance of latency/player performance
This video while useful does not sacrifice bit depth, sample size, or number of streams and would benefit from analyzing these variables
This explains a lot - i wondered why some games felt like the audio and video felt out of sync - it's because i've ran console ports at 165hz
Now to figure out ASIO4ALL, and how to make games use it...
9:21 Can confirm. Got banned from cs three time because of it
Battlenonsense osu! video soon ? Rhythm gamers are probably the only ones who care about milliseconds of audio delay difference :D
Wondering if streamer devices like GoXLR or even Astro Mixamps for gaming headsets also create that delay as well. Based on your conclusions it seems like they do...interesting. Thanks for creating this!
Yo can you do an escape from tarkov analysis again its been 2 years and its had a whole game engine migration and so much optimization its insane. Id reeally like to know where the netcode is at.
With results like that I feel like this is a very pressing topic.