ASIO, WASAPI, Direct Sound... is there any difference in sound quality?

[morillon]

ASIO doesn't give a shit. RME test files pass the integrity check.

These tests have their limits depending on the circumstances and can be missed in certain cases (has already been seen on asr)....

 

[Pictus]

I am afraid this advice is very dangerous for the majority of setups without battery-backed storage:

If you're running a database, it is dangerous.
I and many others have done crash tests for normal workstation operations
and the damage was more or less the same whether it was ON or OFF.
But yes, UPS and backup are our friends.

 

[phofman]

But yes, UPS and backup are our friends.

That is always true, nevertheless there is no reason to increase the likelihood of data corruption by disabling write cache flushing, especially today with fast SSDs or even NVMes. It will have zero impact on audio or USB performance.

 

[Pictus]

That is always true, nevertheless there is no reason to increase the likelihood of data corruption by disabling write cache flushing, especially today with fast SSDs or even NVMes. It will have zero impact on audio or USB performance.

Check

 

[phofman]

Check

In the transcript I see this:

4:44
there also are even more tweaks that you

4:46
can do for example if you go into the

4:48
device manager and you go into one of

4:51
the ssds you can disable right cash

4:53
buffer flushing that's another example

4:55
of one and so for example down here on

4:57
the nvme and the storage port

5:00
if you enable this for example and do a

5:02
bunch of whole other tweaks as well like

5:04
turning on specific settings and turning

5:06
off specific settings and then you can

5:08
just mess with them inside of the power

5:09
plans and the device manager as well but

5:11
if you go even further you can reduce

5:14
the amount of interrupts in your system

5:17
to 100th of what it normally runs

So he says that "together with bunch of whole other tweaks as well as ....the amount of interrupts is reduced". Still I do not see the effect of this specific setting on the number of interrupts (apart of the interrupts for the actual writing, which are normal in standard operation - it's expected to have the data written on the drive, not kept in memory).

MS says it right away:

Also this is not just in case of power failure, but also in case of kernel crash (which still does happen with windows, like with other OSes).

 

[Aerith Gainsborough]

These tests have their limits depending on the circumstances and can be missed in certain cases (has already been seen on asr)....

That must be some pretty freaky circumstances where they pass the test and you have AUDIBLE changes to the signal.
I believe it when I see it.

 

[morillon]

very audible.......very... ;-)
(could not observe the presence of a powerful drc at -1db to avoid digital clipping due to equalizations in the digital domain etc on a streamer... like everything, only test.... that which it tests.... ;-) )
(Post in thread 'WiiM Pro - Review & Measurements (Streamer)' https://audiosciencereview.com/foru...view-measurements-streamer.42300/post-1499160
test rme..."ok" """"bit perfect""""

)

 

[Pictus]

MS says it right away:
View attachment 388575
Also this is not just in case of power failure, but also in case of kernel crash (which still does happen with windows, like with other OSes).

Over 20 years using this without any issues, even in forced crash tests.
Since fewer interrupts contribute to lower DPC latency, I take the questionable risks to achieve the lowest possible DPC.
Windows 11, Ryzen 5600X + Radeon 5600 XT, browsing (LibreWolf) the forum and VCV Rack playing in the background.

BTW, LatencyMon is not good to measure the effect of Write-buffer flushing, but Windows Xperf is.

 

[Aerith Gainsborough]

very audible.......very... ;-)

I somewhat doubt that -58dB THD is "very audible" in a blind test.
Still, it is interesting that the test signal does not trip the DRC.

That being said: who would expect "bit perfect" in EQ mode anyway. :'D

 

[morillon]

I somewhat doubt that -58dB THD is "very audible" in a blind test.
Still, it is interesting that the test signal does not trip the DRC.

That being said: who would expect "bit perfect" in EQ mode anyway. :'D

a simple two-tone made it even more obvious... it is obviously very audible imagine the impact on musical signals and very complex, very strong drc..
was just there because it did not have (and still do not have) any autogain adapting to the correction...
the rme test was just the argument of some rme-boys who could not admit the presence of this drc which had also been forgotten on the "bit perfect" output... but it was here.

in fact only the story of the "absolute" RME test, "drc" which was not fundamentally interesting...shows the limit of our tests which "ignore" themselves as limited...the foundation of a scientific approach...
,-)

 

[Aerith Gainsborough]

a simple two-tone made it even more obvious... it is obviously very audible imagine the impact on musical signals and very complex, very strong drc..

Distortion like that is A LOT harder to hear in actual program material compared to test tones.
Usually your speakers themselves will have a higher distortion than -60dB on their own already unless you have extremely good and expensive ones.

I'd love to hear from the RME guys why their test does not actually trigger the DRC in this component. I could imagine the spikes of the test signal being too short in duration. IIRC, the VU meter of the RME indicates it hitting 0dB FS but I will look at the waveform when I get home. Also possible that it stays under -1 dB and did not trigger the DRC due to that.

 

[so.music]

Has anyone found any solution to make Firefox play bit perfect audio? I'm not that technical, even though I know my way around my PC, but I could not find any way to direct Firefox towards a WASAPI (exclusive) device.

Thanks

 

[lowdef]

Has anyone found any solution to make Firefox play bit perfect audio? I'm not that technical, even though I know my way around my PC, but I could not find any way to direct Firefox towards a WASAPI (exclusive) device.

You running Windows? Look for EarTrumpet in the MS Store. Free, and you can direct output to any connected device. (Yes, my 1st post - long time lurker).

 

[so.music]

You running Windows? Look for EarTrumpet in the MS Store. Free, and you can direct output to any connected device. (Yes, my 1st post - long time lurker).

Welcome! I'm also new, this is my 3rd post.

As for the EarTrumpet, I can already change the device output from within windows. The problem is that I don't know how to take exclusive access - wasapi or asio - so I can bypass the windows audio service and go directly to kernel. I guess it has to be some setting inside Firefox, to properly ask for exclusive mode, just like Foobar does.

 

[lowdef]

Welcome! I'm also new, this is my 3rd post.

As for the EarTrumpet, I can already change the device output from within windows. The problem is that I don't know how to take exclusive access - wasapi or asio - so I can bypass the windows audio service and go directly to kernel. I guess it has to be some setting inside Firefox, to properly ask for exclusive mode, just like Foobar does.

Quick Google search shows this is missing in both Firefox and Chrome. Apparently not enough demand.

 

[phofman]

The problem is that I don't know how to take exclusive access - wasapi or asio - so I can bypass the windows audio service and go directly to kernel. I guess it has to be some setting inside Firefox, to properly ask for exclusive mode, just like Foobar does.

IMO there is no Wasapi excl support in stock Firefox - AUDCLNT_SHAREMODE_SHARED hardcoded in https://hg.mozilla.org/mozilla-central/file/tip/media/libcubeb/src/cubeb_wasapi.cpp#l1987 , https://hg.mozilla.org/mozilla-central/file/tip/media/libcubeb/src/cubeb_wasapi.cpp#l1987 , https://github.com/mozilla/cubeb/pull/415#issuecomment-1035150562

For chrome https://www.audiosciencereview.com/.../exclusive-mode-from-web-based-players.30876/ and https://audiosciencereview.com/foru...ode-from-web-based-players.30876/post-1736791

 

[Hoods_]

Maybe it has something to do with the ASIO buffer or the latency? The processor executes commands faster. The sound definitely changes even if you turn off all graphics in Windows 11. The system starts working and responding to commands faster and the sound changes dramatically. If you don't believe me, try an experiment, turn off all visual effects in Windows 11 and listen to how your hardware sounds. I guarantee you will be very surprised

View attachment 388459

I created an account just to say YES IT WORKS! A quick and easy option that improves audio. Thanks for sharing it

 

[DVDdoug]

A quick and easy option that improves audio.

I'm always suspicious of vague statements like that. Improved how?

Some options settings add (or remove) EQ (changing) frequency response or other "enhancements" and it's usually those effects/enhancements that account for any real differences. ASIO drivers (if your hardware and software support them) prevent the operating system from altering the digital audio stream (but an EQ built-into the application can still make a difference).

With electronics, "sound quality" is defined only by noise (hum, hiss, or whine in the background), frequency response, and distortion, and you didn't mention any of those. With speakers and room acoustics it's a bit more complicated. (Audiophoolery) ...I once had a soundcard that made noise when the hard drive was accessed.

If you can't describe what you're hearing (or what you think you're hearing), I'll believe a proper blind ABX test.

 

[Hoods_]

I'm always suspicious of vague statements like that. Improved how?

Some options settings add (or remove) EQ (changing) frequency response or other "enhancements" and it's usually those effects/enhancements that account for any real differences. ASIO drivers (if your hardware and software support them) prevent the operating system from altering the digital audio stream (but an EQ built-into the application can still make a difference).

With electronics, "sound quality" is defined only by noise (hum, hiss, or whine in the background), frequency response, and distortion, and you didn't mention any of those. With speakers and room acoustics it's a bit more complicated. (Audiophoolery) ...I once had a soundcard that made noise when the hard drive was accessed.

If you can't describe what you're hearing (or what you think you're hearing), I'll believe a proper blind ABX test.

Oh no, now I see what kind of forum I’ve walked into... (sorry). What I’m describing is a subjective improvement in my listening experience, it simply sounds more pleasant, more relaxed, yet with extra microdetails to my ears.

By toggling those options, I can hear the changes in real time, even with YouTube playing in the background, it’s easy to notice even with my eyes closed and with someone else making the changes. I’ve also tested both ways: keeping the same volume, and starting from zero then raising it again to a comfortable level, in both cases, I still prefer how it sounds with the “better performance” option.

I use Audirvana with headphones plugged into a FiiO DAC/amp via USB2.0, that’s my usual setup.

Since it works for me in some way, I’m just sharing it. You can try it yourself and decide if it’s worth it or not. Of course, I’d also be genuinely interested in any technical explanation behind these changes (I’d guess it has something to do with latency handling or better resource prioritization).

Cheers

 

[respice finem]

Just a thought: If one mode is even slightly louder than the other, it will probably "sound better".
A common catch when comparing "by ear", or after applying PEQ.