Audio popping issue

[MiaShino]

It took much time to find simple way of stating this.. maahhh things make so much sense in the mind yet when you attempt putting them to words everything goes fooey.

This audio stuttering issue reminds me of the type of issues experienced during HLE audio emulation on console emulators. It seems Square could be struggling to schedule audio processing requests across multiple cpu cores if any slowdown in the pipeline occurs. This results in the audio engine waiting to process new requests which as you hear causes stuttering effect. This would be especially prominent on FX cpus because of unique nature of core configuration.

Example time! On Bulldozer (FX series) cpu cores do not have dedicated floating point units opting instead to share one across two cores. They are much much of the worse at certain calculations especially ones requiring these floating point units. Vector calculations were and still are quite difficult to work with and as spatial audio requires many of these to be made..

You cannot rely on cpu utilization either. You may see something close to 40% utilization and wonder; but the cpu is not at 100% so I have additional headroom! You may in fact not. If something has stalled thus starving those cpu cores of something to process it will not be processing anything so will read as "not fully saturated". Even worse for FX because if that shared floating point unit is busy doing something then both cpu cores sharing it must also wait.

I may have the number of FP units per core cluster wrong so many apologies again. I hated working with bulldozer and do not do much difficult software development anymore (because it causes insanity..).

Square must increase the audio buffer (this may introduce slight audio delay but is highly preferable to crackling audio) or add option for number of audio channels into the sound options going forward.

For any suffering from this issue who also attempt thousands of erroneous solutions, I would not waste too much time doing this. I.. just.. cannot believe a major developer would make drastic overhaul to their sound engine without seemingly properly testing it across multiple hardware configurations. At the least they should have implemented a legacy fall back should unforeseen problems arise allowing users to temporarily roll back to working solution until problem is resolved. Then I find out they charge for the sound pack that created the issue in first place, seriously? Who does this?

Myself and friends are not encountering the issue yet on older hardware (i7 4790K and Amd Phenom x4).

Edit time! The PS3 cell micro architecture suffered from similar issues when attempting to make those pesky spes do anything because you had to wait for the ppe. It was unholy nightmare for many.

[Maeklos]

Square must increase the audio buffer (this may introduce slight audio delay but is highly preferable to crackling audio) or add option for number of audio channels into the sound options going forward.

Or, just reinstate the sound engine from Shadowbringers and allow that as an optional choice.

[GorthRedsteel]

It took much time to find simple way of stating this.. maahhh things make so much sense in the mind yet when you attempt putting them to words everything goes fooey.

This audio stuttering issue reminds me of the type of issues experienced during HLE audio emulation on console emulators. It seems Square could be struggling to schedule audio processing requests across multiple cpu cores if any slowdown in the pipeline occurs. This results in the audio engine waiting to process new requests which as you hear causes stuttering effect. This would be especially prominent on FX cpus because of unique nature of core configuration.

Example time! On Bulldozer (FX series) cpu cores do not have dedicated floating point units opting instead to share one across two cores. They are much much of the worse at certain calculations especially ones requiring these floating point units. Vector calculations were and still are quite difficult to work with and as spatial audio requires many of these to be made..

You cannot rely on cpu utilization either. You may see something close to 40% utilization and wonder; but the cpu is not at 100% so I have additional headroom! You may in fact not. If something has stalled thus starving those cpu cores of something to process it will not be processing anything so will read as "not fully saturated". Even worse for FX because if that shared floating point unit is busy doing something then both cpu cores sharing it must also wait.

I may have the number of FP units per core cluster wrong so many apologies again. I hated working with bulldozer and do not do much difficult software development anymore (because it causes insanity..).

Square must increase the audio buffer (this may introduce slight audio delay but is highly preferable to crackling audio) or add option for number of audio channels into the sound options going forward.

For any suffering from this issue who also attempt thousands of erroneous solutions, I would not waste too much time doing this. I.. just.. cannot believe a major developer would make drastic overhaul to their sound engine without seemingly properly testing it across multiple hardware configurations. At the least they should have implemented a legacy fall back should unforeseen problems arise allowing users to temporarily roll back to working solution until problem is resolved. Then I find out they charge for the sound pack that created the issue in first place, seriously? Who does this?

Myself and friends are not encountering the issue yet on older hardware (i7 4790K and Amd Phenom x4).

Edit time! The PS3 cell micro architecture suffered from similar issues when attempting to make those pesky spes do anything because you had to wait for the ppe. It was unholy nightmare for many.

Thanks mate for a great explanation! You actually know what you are talking about which makes it a pleasure to read. Now, if only Sicycre could get this post translated and send to the devs, then perhaps we could get a fix out before the end of Endwalker. lol

[MiaShino]

Thanks mate for a great explanation! You actually know what you are talking about which makes it a pleasure to read. Now, if only Sicycre could get this post translated and send to the devs, then perhaps we could get a fix out before the end of Endwalker. lol

Aww many thanks! The at the times crippling imposter syndrome makes me believe otherwise.. Am rusty but all the mention of FX cpus brought back horrible memories of that hot mess of architecture.

If anyone is still curious out there and experiencing problems using the newer Ryzen I have umm, unfortunate news because Ryzen suffers from own unique problem.

Anyone working with Linux or Windows schedulers may know what I am speaking of, it has to do with communicating between core clusters across that "infinity fabric". Optimally you will wish to avoid communicating between primary and secondary blocks of core clusters with there being four cores per main cluster and eight per secondary cluster. Any communication between secondary clusters can suffer from bandwidth constraints.

I should hope they were aware of these things when consulting third party to slap one new audio engine together. You should not do this, save sweeping changes for later patches. If they were somehow not, may it help!

[Maeklos]

I should hope they were aware of these things when consulting third party to slap one new audio engine together. You should not do this, save sweeping changes for later patches. If they were somehow not, may it help!

You know, if they tore apart their testing servers in order to bolster the login servers to help combat the 2002 errors, that would explain why they weren't able to do as thorough of testing on 6.01 as might have otherwise been the case. Especially since they thought that 6.01 would fix the problems introduced in 6.0, only to have them cascade tenfold.

[MiaShino]

You know, if they tore apart their testing servers in order to bolster the login servers to help combat the 2002 errors, that would explain why they weren't able to do as thorough of testing on 6.01 as might have otherwise been the case. Especially since they thought that 6.01 would fix the problems introduced in 6.0, only to have them cascade tenfold.

You would never perform these types of tests upon server hardware though, that is for the dedicated Q&A team. Said team should have dedicated client machines comprised of some range of hardware configurations. The best way to set this up is to gather metrics from your users in same way Steam regularly does via hardware surveys. Some of their programming decisions doth boggle the mind.. like the fifteen minute disconnect that was recently patched.

[raelgun]

It took much time to find simple way of stating this.. maahhh things make so much sense in the mind yet when you attempt putting them to words everything goes fooey.

This audio stuttering issue reminds me of the type of issues experienced during HLE audio emulation on console emulators. It seems Square could be struggling to schedule audio processing requests across multiple cpu cores if any slowdown in the pipeline occurs. This results in the audio engine waiting to process new requests which as you hear causes stuttering effect. This would be especially prominent on FX cpus because of unique nature of core configuration.

Example time! On Bulldozer (FX series) cpu cores do not have dedicated floating point units opting instead to share one across two cores. They are much much of the worse at certain calculations especially ones requiring these floating point units. Vector calculations were and still are quite difficult to work with and as spatial audio requires many of these to be made..

You cannot rely on cpu utilization either. You may see something close to 40% utilization and wonder; but the cpu is not at 100% so I have additional headroom! You may in fact not. If something has stalled thus starving those cpu cores of something to process it will not be processing anything so will read as "not fully saturated". Even worse for FX because if that shared floating point unit is busy doing something then both cpu cores sharing it must also wait.

I may have the number of FP units per core cluster wrong so many apologies again. I hated working with bulldozer and do not do much difficult software development anymore (because it causes insanity..).

Square must increase the audio buffer (this may introduce slight audio delay but is highly preferable to crackling audio) or add option for number of audio channels into the sound options going forward.

For any suffering from this issue who also attempt thousands of erroneous solutions, I would not waste too much time doing this. I.. just.. cannot believe a major developer would make drastic overhaul to their sound engine without seemingly properly testing it across multiple hardware configurations. At the least they should have implemented a legacy fall back should unforeseen problems arise allowing users to temporarily roll back to working solution until problem is resolved. Then I find out they charge for the sound pack that created the issue in first place, seriously? Who does this?

Myself and friends are not encountering the issue yet on older hardware (i7 4790K and Amd Phenom x4).

Edit time! The PS3 cell micro architecture suffered from similar issues when attempting to make those pesky spes do anything because you had to wait for the ppe. It was unholy nightmare for many.

I am also curious to how people are running their audio out as well.

directly from the onboard solution - which I am guessing is most of the case here.

Are people running any USB connected devices such as USB headsets, USB powered headphone amps or a USB audio interface (those running a mixer directly from the onboard is still onboard).

Might be worth if people haven't tried to see if messing with the bit rate and sample rate on their audio settings and see if that alleviates things.

I don't get any popping issues with or without the Embody immersion setting on, I do run the game through a USB audio interface.

For sake of clarity my CPU is a Ryzen 5900X so stupidly high end however looking at reports with people with modern CPUs and I think I've seen some posts with the Ryzen 5600X.

Though my post doesn't provide a solution, it might help with narrowing down further any culprits.

[Grimno]

I am also curious to how people are running their audio out as well.

directly from the onboard solution - which I am guessing is most of the case here.

Are people running any USB connected devices such as USB headsets, USB powered headphone amps or a USB audio interface (those running a mixer directly from the onboard is still onboard).

Might be worth if people haven't tried to see if messing with the bit rate and sample rate on their audio settings and see if that alleviates things.

I don't get any popping issues with or without the Embody immersion setting on, I do run the game through a USB audio interface.

For sake of clarity my CPU is a Ryzen 5900X so stupidly high end however looking at reports with people with modern CPUs and I think I've seen some posts with the Ryzen 5600X.

Though my post doesn't provide a solution, it might help with narrowing down further any culprits.

So my Audio is a bit funky.
I guess something happened with my motherboard so trying to run audio directly from the green port in the back does -not- work. There's no audio at all no matter what I've tried from reinstalling sound drivers and more.
So I had to plug in a Roccat JUKE into the USB in the back, and then plug the audio port into that which allowed me to have audio.
So my audio isn't plugged into the motherboard's sound plug directly, but instead plugged in through the USB port.

[raelgun]

So my Audio is a bit funky.
I guess something happened with my motherboard so trying to run audio directly from the green port in the back does -not- work. There's no audio at all no matter what I've tried from reinstalling sound drivers and more.
So I had to plug in a Roccat JUKE into the USB in the back, and then plug the audio port into that which allowed me to have audio.
So my audio isn't plugged into the motherboard's sound plug directly, but instead plugged in through the USB port.

USB soundcards is something I probably should of added ( I tend to class them as audio interface for some reason ), USB audio devices tend to do more of the onboard processing and ignores more of the CPU or anything onboard for the matter to process the audio signal.

If you play around with the bit rate and sample rate, does that impact anything for you?

[Grimno]

USB soundcards is something I probably should of added ( I tend to class them as audio interface for some reason ), USB audio devices tend to do more of the onboard processing and ignores more of the CPU or anything onboard for the matter to process the audio signal.

If you play around with the bit rate and sample rate, does that impact anything for you?

Nope. I have the Spatial Audio Pack downloaded for the trial and it does help (though combat sounds muffled) but doesn't get rid of the audio crackling 100%
I'll still hear it sometimes in enclosed spaces, or during loading screens between zones. Though I can play Reaper now without crackling. Just ran the Grimlyt Dark dungeon without any crackles as reaper.