"PulseAudio Is Still Awesome"

[mdias]

Citation needed.

What citation? It's pretty common sense if you think about it. At the end of the stack you have a DAC that will have to have a fixed sample rate. Having a fixed sample rate means a digital signal processor can do the job. Hardware has fixed algorithms and cannot improve; software is programmable...

xonar d2X or x-fi xtreme gamer are not able to manage multiple sound mixing!? cmi8788 ca20k1 audio chips should be let hardware mixing. Or not!?

Hardware inferior compared to software!? depends on audio controller...

EDIT: just removed PULSE, browser is faster, avoiding also randomly lags.... 5.1 audio runs well also if sounds are not spreads separately in each different speaker... much better without pulse in my case. So why ot increase performance and abilities of hardware audio acceleration of audio chipset?

Yes, I'm sure most people use xonars......

[Azrael5]

What citation? It's pretty common sense if you think about it. At the end of the stack you have a DAC that will have to have a fixed sample rate. Having a fixed sample rate means a digital signal processor can do the job. Hardware has fixed algorithms and cannot improve; software is programmable...





Yes, I'm sure most people use xonars......

I use both xonar d2x and x-fi xtreme gamer but what the important are the chipset used. Next month I shoould replace my xonar d2x with x-fi xtreme gamer which is excellent on managing both audio and gaming switching from audio controller in windows system.

[magika]

What citation? It's pretty common sense if you think about it. At the end of the stack you have a DAC that will have to have a fixed sample rate. Having a fixed sample rate means a digital signal processor can do the job. Hardware has fixed algorithms and cannot improve; software is programmable...





Yes, I'm sure most people use xonars......

I think about what you said its not a common sense at all. My CA20K2 does its job much better that low quality low-CPU usage software mixing form Pulse. Besides much higher quality of resampling it also supports various sample rates as well.

You are comparing using highly specialized hardware and doing same thing on general purpose CPU, and at the same time CPU implementation is also meant to be light on resources. That is like comparing llvmpipe and hardware with actual driver.

[mdias]

I use both xonar d2x and x-fi xtreme gamer but what the important are the chipset used. Next month I shoould replace my xonar d2x with x-fi xtreme gamer which is excellent on managing both audio and gaming switching from audio controller in windows system.

I hear good things about xonars. But still, the most used chips are onboard stuff like Realtek ALCxxx and things like that.

I think about what you said its not a common sense at all. My CA20K2 does its job much better that low quality low-CPU usage software mixing form Pulse. Besides much higher quality of resampling it also supports various sample rates as well.

You are comparing it with the crappiest setting on pulse, that's not fair. And I like I said, CPUs do a better job quality-wise; it's obvious a fixed-function hardware component will fare better when comparing CPU usage.

You are comparing using highly specialized hardware and doing same thing on general purpose CPU, and at the same time CPU implementation is also meant to be light on resources. That is like comparing llvmpipe and hardware with actual driver.

I am not. You're the one comparing the "trivial" PA resampler to dedicated chips. I only said a CPU is capable of doing a better job than a fixed-function hardware component.
Your comparison is flawed; llvmpipe's purpose is to mimick the abilities of OpenGL (which in turn tries to mirror limited hardware functionality). A better analogy would be comparing what GPUs can do to what offline renderers can do.

You see, with modern CPUs' capabilitites, spending resources on dedicated audio processing hardware makes less and less sense. Our modern CPUs are blazingly fast when doing math, more so if the algorithm is optimized; however resampling can be a tough problem to solve if you want to achieve a certain level of quality. This level of complexity will make specialized hardware components VERY expensive to make, and so people have been shifting this problem to our CPUs.

Normal users don't even notice if the resampling is of somewhat poor quality. After all if you're interested in bit perfect playback you'll make sure all stack plays at the same sample rate and format. Pulse let's you choose between different resamplers that have different quality levels at the expense of different CPU usage levels. If you have performance problems and it's not the resampler, I think it's something worth reporting.

I am not against PA providing options to let the resampling work go through your dedicated soundcards. I think that'd be great, but you have to understand that most people's soundcards only have 1 voice channel, and that's where work must be done first. Projects with limited resources must prioritize things. You are all welcome to contribute though.

[magika]

I hear good things about xonars. But still, the most used chips are onboard stuff like Realtek ALCxxx and things like that.




You are comparing it with the crappiest setting on pulse, that's not fair. And I like I said, CPUs do a better job quality-wise; it's obvious a fixed-function hardware component will fare better when comparing CPU usage.



I am not. You're the one comparing the "trivial" PA resampler to dedicated chips. I only said a CPU is capable of doing a better job than a fixed-function hardware component.
Your comparison is flawed; llvmpipe's purpose is to mimick the abilities of OpenGL (which in turn tries to mirror limited hardware functionality). A better analogy would be comparing what GPUs can do to what offline renderers can do.

You are stating obvious things again: offline renderer provides better quality. But we are talking about realtime, and specialized hardware does better job here.

You see, with modern CPUs' capabilitites, spending resources on dedicated audio processing hardware makes less and less sense. Our modern CPUs are blazingly fast when doing math, more so if the algorithm is optimized; however resampling can be a tough problem to solve if you want to achieve a certain level of quality. This level of complexity will make specialized hardware components VERY expensive to make, and so people have been shifting this problem to our CPUs.

Normal users don't even notice if the resampling is of somewhat poor quality. After all if you're interested in bit perfect playback you'll make sure all stack plays at the same sample rate and format. Pulse let's you choose between different resamplers that have different quality levels at the expense of different CPU usage levels. If you have performance problems and it's not the resampler, I think it's something worth reporting.

I am not against PA providing options to let the resampling work go through your dedicated soundcards. I think that'd be great, but you have to understand that most people's soundcards only have 1 voice channel, and that's where work must be done first. Projects with limited resources must prioritize things. You are all welcome to contribute though.

If you target lowest hardware possible then that is given.

[erendorn]

You are stating obvious things again: offline renderer provides better quality. But we are talking about realtime, and specialized hardware does better job here.

That's just because graphics saturate the CPU, but audio resampling does not, hence you can have real-time for high quality audio mixing. The downside will be the noticeable CPU usage, but you can indeed reach higher quality still well within a desktop CPU computational envelope.

[quandasim]

i use pulse audio with archlinux and i dont havy any problems. from user perspective its much easier to use than alsa. the only exception is recording, which have very bad quality and is desynced with video. So for recording i use alsa.

[mdias]

You are stating obvious things again: offline renderer provides better quality. But we are talking about realtime, and specialized hardware does better job here.

See erendorn's reply. At least for most hardware, the only thing that's better is the speed. Normal users don't care if their sound takes 25ms or 4ms to reach their speakers though...
In any case, you're conceding that it's a matter of processing power to be able to do a better job than specialized hardware. Why is it so hard to believe that in 2015 we already have enough CPU power to do it in realtime?

If you target lowest hardware possible then that is given.

They target first what most people use. And most people don't use xonars or other high-end audio hardware.

[GreatEmerald]

xonar d2X or x-fi xtreme gamer are not able to manage multiple sound mixing!? cmi8788 ca20k1 audio chips should be let hardware mixing. Or not!?

I use both xonar d2x and x-fi xtreme gamer but what the important are the chipset used. Next month I shoould replace my xonar d2x with x-fi xtreme gamer which is excellent on managing both audio and gaming switching from audio controller in windows system.

Haaaa! I have an X-Fi XtremeGamer, and what you said is completely false. One, it does not have hardware mixing, which you would see quite easily by looking at the ALSA wiki: http://www.alsa-project.org/main/ind...-Creative_Labs Two, while its support on Linux is quite poor, it doesn't come close to how utterly terrible their drivers are on Windows. Those cause BSODs, as well as terrible noise while recording unless set in a peculiar way.

I have been hearing good things about Xonars too, and I wish I had known sooner so that I wouldn't have bought the X-Fi...

[duby229]

See erendorn's reply. At least for most hardware, the only thing that's better is the speed. Normal users don't care if their sound takes 25ms or 4ms to reach their speakers though...
In any case, you're conceding that it's a matter of processing power to be able to do a better job than specialized hardware. Why is it so hard to believe that in 2015 we already have enough CPU power to do it in realtime?



They target first what most people use. And most people don't use xonars or other high-end audio hardware.

That's exactly PA's second biggest problem. Normal people don't know that they should care about audio latency. And yet, since they don't know the PA devs use it as an excuse. It's always excuses. The PA devs should care in place of the normal people, because they -do- know that they should care about audio latency.

EDIT: Very few people need highend audio hardware. Less than 100mhz of modern day CPU cycles should be more than plenty.