Subjective comparison of Software Music Player

[lotusaurus]

17 hours ago, lotusaurus said:

While I see that I need only a CPU and a network connection, what sort of CPU do I need? The more powerful the better? Choose speed over cores or vice versa?

17 hours ago, PeterSt said:

Nah, I don't know where you read those "requirements", but you seem to have found the specs of the Mach III Audio PC.

 

I read those "requirements" here:

On 3/22/2021 at 5:45 PM, PeterSt said:

So if not clear already, that also boots the OS from RAM and leaves totally nothing in the PC (no SSD or whatever). Of course you must first have a PC like that; The less you have in there, the better the sound (trust me). So only a CPU and a network connection is what you'd need.

 

But it seems like the comment was meant in a different context. So I'm trying to understand the different situations. Above you say less is better, but then in the bit below you also say beefier is better.

 

17 hours ago, PeterSt said:

Otherwise it would be true that the beefier the processor the better the sound (and still, any normal modern Desktop will do). The more cores the better, with the notice that XXHighEnd will rev the processor down to 500-750 MHz anyway (this is a setting). Think of a Ferrari with over-power.

 

My take is that "less" means less stuff, so no SSD, no special PCI cards for USB, etc. But more cores would be better, and speed of the cores is less important. Which would mean AMD chips with more cores but slower speeds would be better than a faster Intel CPU with fewer cores? Would a GPU be important for processing power?

[PeterSt]

2 hours ago, lotusaurus said:

But more cores would be better, and speed of the cores is less important.

 

Completely correct. It is still the Ferrari example - you can't (are not allowed to) drive over 70Mph / 120Km/h anyway but taking over someone on a one-lane road goes without notice and danger, which is impossible with a 2CV. So it is the over-capacity doing it.

 

2 hours ago, lotusaurus said:

Above you say less is better

 

Yes, but the less is about a different domain: sources which generate interrupts. The less of them, the better again. This is not only theory (empirically determined) but once you get the hang of that, it is all very audible as well.

 

To get the hang of it all somewhat better, try play with the SFS (Split File Size) in XXHighEnd. Set it to small (like 0.8 or so) and know that this is about the smaller read chunks from disk (or memory when really all runs from RAM) or set it to a larger still workable value like 40 and know that the 40 implies ~40 times less read activity, BUT that each activity implies more current spike. Listen to the difference and (thus) get the hang of how all works, but from numerous angles and ditto settings). The lesser SFS implies about continuous current draw which can be determined as noise (the spectrum looks like noise now). Knowing that, again listen and now thus know what it physically does.

All this is not ending up in the audio signal, but indirectly it influences jitter patterns. And it is *that* what you hear.

 

Notice that all is relative to the noisy environment in the first place. Thus, when nothing has been done about it (e.g. Foobar) then the relatively smaller influence of XXHighEnd is noticeable alright, but less. When all is as clean as can be and remaining noise is almost discrete (recognizable source) then the influence becomes apparent and recognizable (because also discrete, like the SFS example I gave).

Regarding this I should emphasize that in MiniOS all happens, because that makes the OS use 55W or so for an e.g. 14 core Xeon, while without that is is 130W (and without license you can't boot into MinOS). This extra Watts come from activity again and this is all transferred to noise via various backdoors. It smooths everything, but meanwhile the resolution (of the sound) is less because of it.

Knowing this, you could compare Normal OS with MinOS and sense the more warmth of Normal OS. You could like it, but it is not the better music reproduction (warmth in this case is "smear".

[kelvinwsy]

I am a bit late to the discussion but here is my xx cents worth.

 

Tried wtfplay over 2 periods. It is command line driven of course and my simple comment on wtfplay is " CD quality on a good player without the digital edge".

Here is where I declare my affliation. 

I am one of the 0.001 % DIY mod audiophiles who use Hqplayer and mod the PC a lot to get the SQ I like.

Hqplayer SQ is something I would say is like playing music with the precision of a brain surgeon. It is precise, incisive and mind boggling in its transparency, immediacy and presence. Everytime I improve on something, I swear that the Sound gets more immediate and clearer at ever Decreasing Volume settings on my HP amp. Now the Abyss AB1266 Phi/CC may not be the latest TOTL HP but it was there 1 generation or so ago. Its freq response is stupendous. On HQplayer with the EC modulators and DSD256, it is just unbelievably good sound

 

This weekend I am trying PGGB's 16fs - 705.6khz music files and I can use JRiver 26 or Hqplayer to play back.

 

What I am trying to say is Software players like some of the members on this thread have said need to be used to their strongest setup. Otherwise its like trying to run the race with one leg tied to a rubber tire. Just won't sound good.

 

Anyway off to listen to some wtfplay tunes now. Yes it is a bit mellow but it sounds good.

Tempted to try XXHighend SW this weekend.

 

Happy Listening.

[pompon]

Try Junilabs player ... it's free and you can optimize the player and specific songs many times.

Mind blowing ...

 

 

[manueljenkin]

The more I use xxhe the more I'm feeling the ui is genius. If only there was a user manual. @PeterSt, would you mind me making one?

 

Played more with the SFS size allocations and got into nice sweetspot. It sounds breathtaking even on my laptop with a u series processor. It is less optimal than a desktop processor owing to power optimization sequences at the expense of switching noises. Apart from processor, laptop Ram, and other components also have low power optimization choices that can have compromises on access noise and periodicity.

 

Version 2.10 sounds great from the get go but version 2.11 takes some 10 mins or so to sound its best in my system (but after 10 mins of break in/warm up, it sounds better than 2.10 as a whole). I assume this is relating to internal schedulers/power optimizations in windows taking time to adapt to the task. Running fantastic on 8x (352.8/384k) on my geek out 450 (TCM filter on the dac) and it sounds amazing. My friend who has an RME adi2 (slow filter on the dac) also finds it amazing running xxhe at 8x. Is 16x exclusive to phasure NOS1, or can other DACs support it as well. He couldn't get it to work at 16x with adi2 yet.

 

I am planning to build a workstation for my compute purposes by the end of this year or early next year, with Radeon pro vii GPU or similar. Haven't decided on the processor yet, likely the new Ryzens. I am curious if I can use the same machine for xxhe audio as well with minimal changed when playing music (just disabling the pcie slots through bios instead of removing the GPU). Any good choices of Linear Power supplies that can supply this GPU and CPU?

 

I would also like to know what are the possibilities of running xxhe on a hyper v windows server running on top of a lean windows server. I might be using linux/bsd for majority of compute task and separate hyper v windows server or normal windows pro (separate from the one running xxhe) for the windows tools (some circuit design tools etc). I haven't checked much if there's noticeable performance hit, anything I should be aware of? Regarding the Second Level Address Translation (SLAT)

[PeterSt]

5 hours ago, manueljenkin said:

I would also like to know what are the possibilities of running xxhe on a hyper v windows server running on top of a lean windows server.

 

I am not sure whether I addressed it already (probably Yes), but the Mach III Audio PC is explicitly based on a server board. So *the* main difference with its predecessor (Mach II) is the server MoBo.

But this is not really what you asked, and the hands-on experience of several months of trialing with virtual machines (the leanest possible) told me that this was a no-go. Theoretically it already won't work well because of too much in between. This is one layer at least and you will notice it in everything (buffers can't be that small etc. etc.).

 

5 hours ago, manueljenkin said:

My friend who has an RME adi2 (slow filter on the dac) also finds it amazing running xxhe at 8x. Is 16x exclusive to phasure NOS1, or can other DACs support it as well

 

Hmm ... it should be able to work; I have one myself but I did not try it so far, but the fact that @manisandher also has one *and* tried it on XXHighEnd (or on his NOS1 DAC only ?) without further complaints, tells me that it just works. Maybe it is the choice between "Needs 24 bits" and "Need 32 bits", although I recall from ever (RME) back that it will do both. Of course the ADI2 is 32 bits inherently ...

 

5 hours ago, manueljenkin said:

Any good choices of Linear Power supplies that can supply this GPU and CPU?

 

Personally I have no other experiences than one customer who applied an external LPSU to his Mach II back at the time, had that LPSU even heavily modded (like making it 3x more expensive by an experienced LPSU EE), now has a Mach III with our own LPSU in there, and likes the latter for the way better. So No, I don't know about third party products, but I do know what we apply to our own (and this is about response speed explicitly - see ATX specs).

 

On 5/4/2021 at 5:47 AM, lotusaurus said:

Which would mean AMD chips with more cores but slower speeds would be better than a faster Intel CPU with fewer cores? Would a GPU be important for processing power?

 

I actually forgot to respond to this ...

I have no hands-on experience, nor feedback from customers about the Ryzen CPU's. They should work out for the better (because more lean), BUT since it also requires a new MoBo obviously, it is too far for me to just try (the development of a new Audio PC around a new MoBo takes many months).

XXHighEnd does not use GPU cores/power, so no, that does not help at all. Contrary, the Audio PC should not contain a GPU in the first place (although Server boards like we use, (always) do BUT not on-board of the CPU, which would be forbidden).

 

Btw, the greatest hidden gem of 2.11 (which came without real release notes) is ending your listening day (Unattended Playback) with Pause (E on the On Screen keyboard pad), which causes all to be really "hot" until the next listening session which you'd start with P(lay). This is way better than a listening session of 3 or so hours, because the other (say) 20 hours are really doing more to your system. I myself have only experience with the NOS1, but I think it will apply to all DACs. Remember that this only works when Unattended is in order (works outside of MinOS the same).

[manueljenkin]

2 minutes ago, PeterSt said:

 

I am not sure whether I addressed it already (probably Yes), but the Mach III Audio PC is explicitly based on a server board. So *the* main difference with its predecessor (Mach II) is the server MoBo.

But this is not really what you asked, and the on experience of several months of trialing with virtual machines (the leanest possible) told me that this was a no-go. Theoretically it already won't work well because of too much in between. This is one layer at least and you will notice it in everything (buffers can't be that small etc. etc.).

 

 

Hmm ... it should be able to work; I have one myself but I did not try it so far, but the fact that @manisandher also has one *and* tried it on XXHighEnd (or on his NOS1 DAC only ?) without further complaints, tells me that it just works. Maybe it is the choice between "Needs 24 bits" and "Need 32 bits", although I recall from ever (RME) back that it will do both. Of course the ADI2 is 32 bits inherently ...

 

 

Personally I have no other experiences than one customer who applied an external LPSU to his Mach II back at the time, had that LPSU even heavily modded (like making it 3x more expensive by an experienced LPSU EE), now has a Mach III with our own LPSU in there, and likes the latter for the way better. So No, I don't know about third party products, but I do know what we apply to our own (and this is about response speed explicitly - see ATX specs).

 

 

I actually forgot to respond to this ...

I have no hands-on experience, nor feedback from customers about the Ryzen CPU's. They should work out for the better (because more lean), BUT since it also requires a new MoBo obviously, it is too far for me to just try (the development of a new Audio PC around a new MoBo takes many months).

XXHighEnd does not use GPU cores/power, so no, that does not help at all. Contrary, the Audio PC should not contain a GPU in the first place (although Server boards like we use, (always) do BUT not on-board of the CPU, which would be forbidden).

 

Btw, the greatest hidden gem of 2.11 (which came without real release notes) is ending your listening day (Unattended Playback) with Pause (E on the On Screen keyboard pad), which causes all to be really "hot" until the next listening session which you'd start with P(lay). This is way better than a listening session of 3 or so hours, because the other (say) 20 hours are really doing more to your system. I myself have only experience with the NOS1, but I think it will apply to all DACs. Remember that this only works when Unattended is in order (works outside of MinOS the same).

Thanks Peter. I'll look into these. I'm more interested into ryzen + Radeon vii pro + of course search of optimal low noise motherboard etc, and disabling vii, either in bios or as a whole since this will be both my number crunching machine for scientific computation and also audio machine. I'll be curious if you ever make a ryzen based one 😬.

 

[manueljenkin]

On 5/4/2021 at 9:37 PM, pompon said:

Try Junilabs player ... it's free and you can optimize the player and specific songs many times.

Mind blowing ...

 

 

Thank you very much. I am using the file optimizer and running the optimized files through xxhighend. It sounds much better than the stock file. Less of the digital haze. I have to make sure only optimized files are loaded in xxhe during playback. If I load a combination of optimized + unoptimized the overall fidelity isn't as good as loading only the optimized file.

 

I also found that in my system I prefer to have everything at 2x or more optimization cycles even if it takes more time. 1x while mostly better feels a little "light" sounding. The problem is it takes a lot of time. Any way to make it a batch operation? I understand batch operation may not be possible in certain scenarios like parallel procs or larger space introducing noise back, but I would be more excited if there's a possibility of batch operation.

 

NOTE: I have checked the files and the data is intact (ie bit perfect). This tool is about reducing the intrinsic noise associated with the physical (electromagnetic) structure of the stored data. It definitely sounds better.

[PeterSt]

7 minutes ago, manueljenkin said:

I'll be curious if you ever make a ryzen based one 😬.

 

I should, because the upgrade path of the Xeon's is a bit dead at the moment. This is BIOS related (super slow boot into/from memory) and Memory related (the required ECC memory sounds like sh*t).

 

7 minutes ago, manueljenkin said:

and the on experience of several months

 

*hands-on (I corrected that in my post by now)

[manueljenkin]

2 minutes ago, PeterSt said:

 

I should, because the upgrade path of the Xeon's is a bit dead at the moment. This is BIOS related (super slow boot) and Memory related (the required ECC memory sounds like sh*t).

 

 

*hands-on (I corrected that in my post by now)

I expected this regarding the ECC memory. It has overhead and additional Circuit to function both of which can introduce noise of their own. Maybe it would be possible to re work/de tune the instruction sets around this but I guess that would be a lot of work.

 

On the other hand ECC is useful for other applications 😅. Not a must for me though.

[seeteeyou]

I saw 19041.1 on the other thread you just started several hours ago, then I recalled something about the "bare minimum" version of that particular Windows build number

 

ReMin - replicated MinWin
https://github.com/replisys/remin-core

 

Remin project notes
https://osg.wiki/books/servicing/page/remin-project-notes

 

FYI - I did a search on Google and found the specific ISO required for ReMin, though it would take quite a few tries to start downloading the file successfully

 

https://pastebin.com/raw/iMGaaCYJ

 

After that, some packages should be necessary for adding audio support back to the OS

 

https://uupdump.net/findfiles.php?id=c8e87bb7-e741-41c1-b164-d97a56d9fd49

 

Someone posted some instructions here

 

https://forums.mydigitallife.net/threads/guide-adding-gui-to-server-core-1-21v.76023/page-9#post-1412792

 

And then this one could also be useful

 

https://cloud.mail.ru/public/qRL8/5ijBydZDU/Packs/MultimediaRestrictedCodecs/10.0.19041/

http://forum.notebookreview.com/threads/help-debloating-windows-10-image-needed.832141/

Quote

+ Updated the Feature "Integrate Multimedia Restricted Codecs" to support Windows Server v2004 Source OS.

 

[manueljenkin]

On 5/3/2021 at 3:32 PM, lotusaurus said:

 

I'm keen to try out XXHE, but the documentation is ... challenging. I've been trying to work my way through that, but the first major obstacle is trying to understand what sort of PC I need. While I see that I need only a CPU and a network connection, what sort of CPU do I need? The more powerful the better? Choose speed over cores or vice versa?

I made a set up tutorial for xxhighend: 

 

Please have a read, you are likely to get most of your questions answered there.

 

XXhighend will work on any x86 pc. In general, Server/multi-core in under-clocked mode are better (the parallelism and higher cache, often relate to lower noise and better determinism on the usb data and ground lines) but it's only a rule of thumb.

[manueljenkin]

4 hours ago, seeteeyou said:

I saw 19041.1 on the other thread you just started several hours ago, then I recalled something about the "bare minimum" version of that particular Windows build number

 

ReMin - replicated MinWin
https://github.com/replisys/remin-core

 

Remin project notes
https://osg.wiki/books/servicing/page/remin-project-notes

 

FYI - I did a search on Google and found the specific ISO required for ReMin, though it would take quite a few tries to start downloading the file successfully

 

https://pastebin.com/raw/iMGaaCYJ

 

After that, some packages should be necessary for adding audio support back to the OS

 

https://uupdump.net/findfiles.php?id=c8e87bb7-e741-41c1-b164-d97a56d9fd49

 

Someone posted some instructions here

 

https://forums.mydigitallife.net/threads/guide-adding-gui-to-server-core-1-21v.76023/page-9#post-1412792

 

And then this one could also be useful

 

https://cloud.mail.ru/public/qRL8/5ijBydZDU/Packs/MultimediaRestrictedCodecs/10.0.19041/

http://forum.notebookreview.com/threads/help-debloating-windows-10-image-needed.832141/

 

Can you guide me through this please. Is this for getting a low footprint windows 10, or windows server 2019 or windows server 2014? I would like to explore this.

[PeterSt]

2 hours ago, manueljenkin said:

Can you guide me through this please. Is this for getting a low footprint windows 10, or windows server 2019 or windows server 2014? I would like to explore this.

 

Please don't do this when using XXHighEnd or everything will be one big out of control mess. This is because XXHighEnd does this herself already (called "MinOS" mode).

 

On another note, XXHighEnd should be used with 14393.0 for best (and super stable) results. This is something for that other thread you created, but it is not easy to incorporate because the thread is already there and it should be in the first post. That means you should be able to edit that (you can ask Chris).

[manueljenkin]

50 minutes ago, PeterSt said:

 

Please don't do this when using XXHighEnd or everything will be one big out of control mess. This is because XXHighEnd does this herself already (called "MinOS" mode).

 

On another note, XXHighEnd should be used with 14393.0 for best (and super stable) results. This is something for that other thread you created, but it is not easy to incorporate because the thread is already there and it should be in the first post. That means you should be able to edit that (you can ask Chris).

Thank you Peter. I'll try to request permission from Chris on that.

 

And don't worry, I won't do any other tweak for the xxhighend installation. The above tweak I'll try independently with other players in a separate installation. As you know I am trying to learn the whole process so I'm trying out all approaches suggested by different people, but I'll make sure one edit doesn't affect another (mostly when a system edit is done, I'll do a clean reinstall of windows before trying another player).

[micr0g]

Hi i think Xmplay is the best option and jriver the worst as it sounds muffly

winyl is ok but not for me as i like an interface and usage like winamp.

Musicbee is great but it has so many options and you can't drag&drop a folder to the "playing now"

Playpcmwin is a great option but it takes to long to read just one track so i'm here to ask you if i can do something to load the track faster or if i can somehow to stop the ram loading.

 

Thanks

[manueljenkin]

10 hours ago, micr0g said:

Hi i think Xmplay is the best option and jriver the worst as it sounds muffly

winyl is ok but not for me as i like an interface and usage like winamp.

Musicbee is great but it has so many options and you can't drag&drop a folder to the "playing now"

Playpcmwin is a great option but it takes to long to read just one track so i'm here to ask you if i can do something to load the track faster or if i can somehow to stop the ram loading.

 

Thanks

Ram playback in general imo sounds better and software devs who design these tools might feel the same. Maybe try Junilabs audio player. It is ram playback but felt snappy for me.

 

Non ram playback, there's a player called Lilith audio player. It also supports vst. Maybe give it a try. I'll try to make a total list of software choices soon.

 

Jriver has settings you can tweak.

[micr0g]

On 5/13/2021 at 5:37 PM, manueljenkin said:

Ram playback in general imo sounds better and software devs who design these tools might feel the same. Maybe try Junilabs audio player. It is ram playback but felt snappy for me.

 

Non ram playback, there's a player called Lilith audio player. It also supports vst. Maybe give it a try. I'll try to make a total list of software choices soon.

 

Jriver has settings you can tweak.

Thanks for the answer

Junilabs looks and sound good except the fast forward option,i tried ulilith at past but xmplay still sound better for me.

What's the setting for Jriver that you're talking about?

[manueljenkin]

I haven't experimented much with Jriver, but it has got options to tweak buffer settings etc. I have similar experience as you, but others have reported good results hence I make a recommendation to try those before coming to conclusion.

[Norton]

On 5/6/2021 at 1:53 AM, manueljenkin said:

Is 16x exclusive to phasure NOS1, or can other DACs support it as well. He couldn't get it to work at 16x with adi2 yet.

Just FYI I have used XXHE 16x into my Oppo205 with no problem.  I would have thought that any DAC that can accept 704/768 would work.

[EdmontonCanuck]

On 5/6/2021 at 12:53 AM, manueljenkin said:

 

NOTE: I have checked the files and the data is intact (ie bit perfect). This tool is about reducing the intrinsic noise associated with the physical (electromagnetic) structure of the stored data. It definitely sounds better.

 

I'm sorry, but please explain how this is even possible. How does a software player manipulate the "physical (electromagnetic) structure of the stored data"? This is an extraordinary claim. 

[manueljenkin]

On 5/17/2021 at 8:15 PM, EdmontonCanuck said:

 

I'm sorry, but please explain how this is even possible. How does a software player manipulate the "physical (electromagnetic) structure of the stored data"? This is an extraordinary claim. 

This was asked and described in a couple of other threads and both the threads were shut down 😅. Hopefully it isn't such a back and forth here. I'll just give my "guess" on how it does this, but it does something for sure.

 

The thing is the data is stored in the cells (assuming SSD) as charges (the microstructure behaves "similar" to capacitors). I don't know how the tool exactly works but the system noise levels (especially ground plane) are different when doing different activity and generally lower when more components are off. When the write happens when the system is at lower noise level, it is likely to do it in a more organized (think of it like charge distribution pattern). Now when you are accessing the data from the cell, you'll be having a noise associated with the access (EM waves and noise is generated when a charge is accelerated). The more organized one are likely to have lesser noise. This whole data/charges goes through a big link of buffers etc before it goes to the usb port and how the specific pattern passes through with its own detectable signature till the DAC is beyond my present comprehension levels.

 

Anyway, could just give a try. And if an objective analysis is needed, feel free to make a thread in objective fi forum.

[March Audio]

7 hours ago, manueljenkin said:

This was asked and described in a couple of other threads and both the threads were shut down 😅. Hopefully it isn't such a back and forth here. I'll just give my "guess" on how it does this, but it does something for sure.

 

The thing is the data is stored in the cells (assuming SSD) as charges (the microstructure behaves "similar" to capacitors). I don't know how the tool exactly works but the system noise levels (especially ground plane) are different when doing different activity and generally lower when more components are off. When the write happens when the system is at lower noise level, it is likely to do it in a more organized (think of it like charge distribution pattern). Now when you are accessing the data from the cell, you'll be having a noise associated with the access (EM waves and noise is generated when a charge is accelerated). The more organized one are likely to have lesser noise. This whole data/charges goes through a big link of buffers etc before it goes to the usb port and how the specific pattern passes through with its own detectable signature till the DAC is beyond my present comprehension levels.

 

Anyway, could just give a try. And if an objective analysis is needed, feel free to make a thread in objective fi forum.

That's OK we can take the discussion and you response here into the objective forum area.

[manueljenkin]

On 5/16/2021 at 2:09 PM, Norton said:

Just FYI I have used XXHE 16x into my Oppo205 with no problem.  I would have thought that any DAC that can accept 704/768 would work.

Thanks. I donot have a DAC that supports 705/768 but my friends have and they have trouble setting the highest sample rate to work. A friend of mine was able to get 705 to work but 768 didn't work strangely enough.

[manueljenkin]

Has anyone here tried the new audirvana with kernel streaming? My trial is over the previous time I tried (older version without kernel streaming) and I am interested to know about the new version changes.