archsan wrote:

*that's usually ASIO for Windows, zilch for OSX

OSX doesn't have a low latency direct audio pathway? Nothing like JACK for Linux? OSX is technically a *NIX as it's derived from BSD...

keropi wrote:

That's pure speculation without any evidence. You can't be sure it's not a driver problem or it's not a hardware limitation that […]
Show full quote

That's pure speculation without any evidence.
You can't be sure it's not a driver problem or it's not a hardware limitation that makes audiog behaves like that. If you have some crappy onboard audio device and said device wants to "enrich" sound via a software way , or is simply incapable of doing all sound operations on it's own, then ofcourse it will use cpu time via the audio subsystem.
And define "stock UUA driver" , there is also a "stock SVGA driver" in windows that noone uses or complains about: it's just there to offer basic functionality until a hardware specific driver gets installed.

What hardware devices give you problems and what exactly are you doing when those problems appear? I just checked my work pc with a Q9300 and onboard realtek audio on an ASUS P45 mobo with MS's driver (or whatever driver win7 had for it , I didn't bother to install any driver in my work pc for audio) and audiog is stuck at 0%.

Anything that uses audio causes tiny amounts of audiodg CPU usage. Recording input causes even more. This is with the stock Microsoft UAA driver on RealTek HD Audio codec's. Default settings as well. (Or whatever "settings" that get's set when you click the big "Restore Defaults" button in the audio device page)

Hardware has nothing to do with audiodg, Windows 7 does not do hardware mixing, it just drives a software mixed sample stream to whatever audio device that can accept it through it's generic driver or whatever vendor specific driver you have installed on your system. The stock UAA driver is the "High Definition Audio Device" driver that Microsoft authors, it can be used with any device that complies with Microsoft's UAA specification, yes the X-Fi series cards can be used in this way but you wont beable to use OpenAL with it or anything else for that matter.

When you go to look at sound devices or sound cards, around the time Vista and 7 came out you'd see these stickers for Universal Audio Architecture, telling you the card is effectively "plug and play", because it can work with the built in driver Windows Vista\7 comes with. This is similar to how ALSA works but ALSA has vendor specific drivers for every sort of device you can imagine, not a blanket driver like Microsoft's and that only handles Intel HD Audio specification devices.

For the record this is the driver I'm talking about: http://en.wikipedia.org/wiki/Universal_Audio_Architecture

So when you think about it, this is supposed to be a stock barebones driver. It should just simply provide audio and that's it... Why does even Microsoft's own driver have these stupid software DSP "plugins" associated with it? Why are we not allowed to strip that away so that the audiodg problem goes away for good?

As for devices I tested, my laptop which the sticker on it has since been rubbed off because ASUS has no idea how to design stickers, so I can't identify it for you, my old Gateway OEM which had a RealLek HD Audio, and this new motherboard I'm using in this new system that has a RealTek HD Audio as well. I'm using the UAA driver and I still get audiodg usage with it despite "Disable All Enhancements" being checked... No I am NOT GOING TO GO THROUGH MULTIPLE FRESH INSTALLS, that has done NOTHING to debug this issue. It is a problem with this stupid "Software DSP Plugin" crap you can't permanently disable and unroot from the audio service. If this doesn't count as "evidence" then I don't know what will.

By the way, programs like Skype, Teamspeak, Mumble, etc. Most VoIP programs and also Audacity, will jump audiodg to 5-12% just from making use of your microphone...

I added this to my previous post:

Even on my intel Atom mITX board with crappy onboard realtek chip (but with drivers installed) audiog uses 0%, tried playing 3 yt videos at different volume levels , it brought the machine to a crawl but audiog's max cpu time was 1% for some 0.5 seconds...

I'll try audacity on the atom with a mic and edit again

edit: done, started a recording session with audacity, audiog goes from 0% to momentarily 2% and then back to 0%. The only time audiog gets some serious cpu time is when you plug/unplug mics/speakers and the systems gets updated.
Also audiog without any speakers connected is dormant and consumes ~84kb , once you connect speakers there is a cpu usage up to 40% while the system is working, mem usage goes to ~11MB and it stays there consuming those 11MB and with next to nothing cpu usage. All that with the puny atom.
I just don't get this high cpu usage... also realtek made a gazillion chipsets over the years, it's possible some are simpler than others. Think winmodems.
Did you get these problems with "good" dedicated soundcards like Creative or Asus ones or was it always with onboard audio?

just to be clear, I don't really care to defend win7 when it clearly has a fault, it's just that this specific thing looks like a driver/hardware limitation and not something that win7 is entirely responsible. Sure you may get sound leveling - as per MS's standards that sacrifice speed to looks - but it's up to the drivers to offer a way to opt out of any extra sound processing. Half assed drivers never solved anything and you can't expect MS to bother so much as to make customizeable generic drivers... 🤣

DracoNihil wrote:

OSX doesn't have a low latency direct audio pathway? Nothing like JACK for Linux? OSX is technically a *NIX as it's derived from BSD...

It does, in the name of Core Audio, OSX's low-level audio API that's bla bla ... OK, I'm not that into the technical bits, but I think you'll find this interesting. 😀

So audio interfaces that are Core Audio-compliant don't need any additional driver (I've seen some manufacturers provide ASIO drivers for Mac but not sure if those are used at all -- or maybe they're for OS9) just to provide low latency audio.

Now windows, hmm, looks like the audio architecture was/is designed for the average consumer first, and the pro/enthusiasts second (or last/none -- hence the total bypassing). This is just my big-picture hunch/speculation of course, though I don't follow the audio side that much since the days of KMixer--oh my. So if things like AGC (probably a crude version of it) aka volume leveling takes priority (enabled by default WTF) and then some bad driver-writing is added to the mix (say, "disable" doesn't really mean disable) ... I wouldn't be surprised actually.

So what could you expect from Win9? I say, a better AGC algorithm and probably not much more. After all don't they care more about Skype?

keropi wrote:

Did you get these problems with "good" dedicated soundcards like Creative or Asus ones or was it always with onboard audio?

Oh man don't get me started... When I had my x-fi both xtremegamer and this new Titanium HD PCIe, I've had audiodg use beyond 12% (this is on a quad core CPU and audiodg clearly is not a multi processor process...) CPU usage on recording input as well as just basic audio output. No enhancements enabled. This is on the original Creative drivers, PAX's drivers, and Daniel K.'s drivers. I've never found a solution to this stupid problem and I had to resort to enabling the RealTek HD Audio just to use my microphone with as little CPU usage as possible.

archsan wrote:

It does, in the name of Core Audio, OSX's low-level audio API that's bla bla ... OK, I'm not that into the technical bits, but I think you'll find this interesting. 😀

So audio interfaces that are Core Audio-compliant don't need any additional driver (I've seen some manufacturers provide ASIO drivers for Mac but not sure if those are used at all -- or maybe they're for OS9) just to provide low latency audio.

Now windows, hmm, looks like the audio architecture was/is designed for the average consumer first, and the professionals second (or last/none -- hence the total bypassing). This is just my big-picture hunch/speculation of course, but I don't follow the audio side that much since the days of KMixer--oh my. So if things like AGC (probably a crude version of it) aka volume leveling takes priority (enabled by default WTF) and then some bad driver-writing is added to the mix (say, "disable" doesn't really mean disable) ... I wouldn't be surprised actually.

That link was very interesting, thanks for sharing it.

Also Windows tried to make their own ASIO, the WASAPI "Exclusive Mode" which is just Kernel Streaming (which is what ASIO4ALL does without even using WASAPI Exclusive Mode), though the problem with that is you can't do anything else audio wise.

With my creative card I can use ASIO or OpenAL to bypass Windows Audio without interfering with windows audio in the first place. One of the biggest benefits to having hardware mixing... But I still am plagued by audiodg being a POS. And only Microsoft can fix this problem... Which they wont, they are more than happy to blame sound card driver vendors than fix their own broken code. Likewise, sound card vendors are more than happy to blame Microsoft for hardware faults and driver faults when, again, it is their fault for providing drivers that were coded by inept people just as much as it may be Microsoft's fault for not providing decent documentation. Oh and it's obviously the sound card vendors fault for faulty hardware. They designed the hardware, not Microsoft.

So I ran Process Monitor by Sysinternals and started recording in Audacity. audiodg.exe gave me 0.78% of Total and Kernel CPU, but only for a second. Less than 1%? That's outrages!

BTW, you shouldn't force kill audiodg.exe, you can just disable the Windows Audio service (Audiosrv).

Disabling the service disables all audio (who would of thought), it also makes it virtually impossible to use OpenAL properly.

I do the same thing and it's certainly not less than 1% for me, and not for any of the other systems I have laying around. I've seen people complain about this on technet and other windows related forums and some people come up with these insane registry hacks and stuff to try to disable the "multimedia class scheduler" without disabling the audio service, which for some reason the audio service depends on "multimedia class scheduler".

I really demand to know why getting rid of the software related DSP that hooks into audiodg is so impossible... I actually demand to know why that audiodg thing was thought of a "good idea" for windows in the first place. Software DSP should be done by end user programs, not at a driver and kernel level... ESPECIALLY if you can't disable it entirely.

EDIT: The audiodg thing was also scrutinized as being some sort of secret audio related DRM too, now that I remember...