Operating systems and their sound signatures

[PeterSt]

 

On 25-12-2017 at 12:38 AM, sandyk said:

Peter St. has also had a great deal to say about the SQ of different Windows Operating Systems.

 

Yes, but where to begin ... OK, here perhaps :

 

14 hours ago, rando said:

@PeterSt Would it be out of place to ask why you developed XXHighend around W10 instead of Server 2016 (or Server 2012)?  

 

XXHighEnd started out under XP. Of what I recall though, the very first official public version was under Vista and WASAPI (this was in 2007).

Soon we had Windows 2008 Server which also was Vista. Later we had Windows 7 and 2008 R2 Server (which was W7).

 

At this point I should tell you that any server version is essentially the same as its "client"/desktop counterpart. Server versions were and are equally supported. But mind the "supported", which means explicitly recognized - and optimized specifically.

 

Next up was W8 and Server 2012. Regarding this Server version I must be more vague, as I was too long busy with optimizing for W8 itself, and when this finally was done (took me close to a year) there was W8.1. However, W8.1 sounded so poor right out of the box, that I skipped that all together. This is aso why I don't know - or heard much about W2012 R2. That too, is not supported (read : was not optimized) officially. And besides that, W10 was in sight. Server 2012 was just supported but at that time people were not interested in it, because W8 itself was too good (OK, after all the work I put into it).

 

W10, a completely different beast, right out of the box showed much potential. Also it soon appeared that each new Build which came about, sounded totally different. And, because each day new builds emerged, I had to decide which one to support. This started with Build 10074 (this is still a most good example of how an OS can sound devastatingly different) which btw is still supported and also still runs, if you know how to (officially it is expired). The first official W10 for the public, Build 10240, was skipped because of poor sound in the base. A next more formal base point (build) was 10565. Although this was optimized by me, it never touched base really, as it was a more normalized version of 10074 only and let's say that it was not special enough. For 10565 the same counts as for 10074 : although expired it is still supported and still runs if you know how to.

 

With W10 Build 10586 a new phenomenon was introduced : sub-build numbers. So from here on, any .0 version is the most native, in this case thus 10586.0. It does not contain bloat like driver special stuff and each time a new larger formal (ISO downloadable) version comes about, the .0 version must be caught and preserved. Of course the major problem is that your precious .0 version should not be updated (each day again) and a lot of work has gone into this, as officially updates can't be stopped. Anyway, 10586.0 was the first W10 version of which we all could say "ditch W8 NOW and keep 10074 if you like its character" (10074 is far from neutral).

After 10586.0 only one "lean" OS was produced and this is 14393.0. But to be honest, those after that weren't tried because the perception of containing more bloatware only (but with a "game mode", theoretically good for sound, of which I saw no gamer being enthusiastic).

 

Suddenly we have arrived at W2016 Server, because that is 14393.x. Small problem, where I normally got hold of the .0 versions and preserve them, for the Server version I didn't, and I can't provide them either. But anyway, about this latest Server version, look here if you want : 2.08d on WS2016 first impressions.

In this example we can NOT say that W2016 Server is the same as 14393.0 because that Server version is 14393.693 (you can't see that in those postings). Only for W2016 14393.0 it will be the same as the desktop version.

 

The server versions thus are the same, BUT come with different feature sets. In the thread I linked to you can see that the Group Policy Service actually is in issue, because for this Server version it can not be shut off. Solution : let it be. And now it theoretically sounds worse than the normal 14393.0 version. Outside of this difference, nothing is differently optimized for the desktop version vs the server version. There's just nothing to be found which can bring a difference and this has been so for each Server version. And I don't know whether it is in that thread or another recent one, but the optimization goes quite a bit beyond Fidelizer and AO together (if we are allowed to count services and processes and threads and such, it is easy to see for that reason alone).

 

If you only look at this and see how we deal with OSes :

then you can imagine how this is only a start of writing I could do forever. But I will spare you that.

Each of these OSes can be booted in 4 modes and each of them sound totally different. Thus for example, W10 10074 sounds wildly different from anything else but its 4 modes also sound different from each other.

 

Lastly for this little write-up, this is about Operating Systems and how each must be treated differently ("optimized") to bring forward its best potentials for sound. The optimization is done by XXHighEnd, but apart from 5 or 6 or so optimization parameters with some logic (like leave on LAN or kill it al together), this is passive stuff. Crucial, but passive (determined at "Boot into Minimized OS"). Within that though, we have the active XXHighEnd parameters, assumed XXHighEnd is also used for playback. So they too work with the Operating System but within the pre-optimization as per boot session. I am using it myself now for more than 10 years and I (or others) still find new / better settings regularly.

And supposed I get some 16xxx.x W10 version going next month, then the whole feast starts all over again.

 

One more thing about the "server" versions :

XXHighEnd again assumed to be the playback software, notice that one of the key points of all of the OSes is that they are optimized to "core" versions, or IOW, no Windows desktop (GUI) or anything present when audio playback is going on under control of XXHighEnd. Also no XXHighEnd GUI (core does not allow any GUI). This, while in full control of the sound engine (which virtually runs as a service). Therefore the real answer to Rando's question is : it shouldn't be needed to use a Server version because after optimization they end up the same as the optimized desktop versions (but for W2016 a bit worse).

 

Regards,

Peter

[PeterSt]

The OSes from another perspective, subjectively (but judged by 100s of people the very same) and relative to each other (which we can only know in retrospect) and dealt with by XXHighEnd :

 

XP - The beginning of Computer Audio. Mostly not bit perfect and therefore psychologically "wrong".

 

Vista - Great improvement and WASAPI debunked ASIO. From here on all plays bit perfect, if only WASAPI Exclusive Mode is used.

 

W7 RTM - All the timings changed and months of work were necessary to make it sound right.  Somewhere here we stepped away from WASAPI and started to use Kernel Streaming exclusively. WASAPI is not lean as it sits on top of KS.

 

W7 SP1 - Timings changed again and for the worse. SP1 never made it for good sound - everybody could roll back to RTM by trickery means.

 

W8 - Phasing problems for everyone. Here the ultimate optimization was necessary, far beyond W7. Nice thing is that W8 allows for much of it (W7 relatively "nothing"). Took a small year to get it right.

After that, W7 RTM appears to show strange raw bass sound. Rough. Bassy but rough.

The main problem with W8 is that it tends to play the music by robots. Foottapping ratings remain under par. Still better than W7 RTM.

 

W10, 10074 - Very special sound. Most palpable. But not neutral at all. And very odd : today with the Lush USB cable, there is plainly no mid present. None. So with that USB cable, 10074 is rendered useless.

 

W10, 10586.0 - The first OS since possibly Vista which sounds normal again. Neutral. When you're used to this, you can't use W8 any more, because now that is definitely robots playing all over.

 

W10, 14393.0 - The first W10 OS which makes things special again. More color in the highs than 10586.0. From here on W8 is without any color. 14393.0 has the palpability of 10074 (which also was special) but without a sense of not being neutral. 14393.0 is the least digital sounding OS so far and also is open the most to external parameter settings (like buffer sizes in playback software).

[PeterSt]

From again another angle and only mentioning it because I saw it passing by as a subject :

 

Yes, the more powerful the PC, the better the sound will be. This has always been the general consensus, and in 100% of cases where people thought they'd be good with their laptop or other "mini" system, they had to admit they were no good at all, after finally implementing the hefty desktop.

 

General consensus too is the more processor cores the better, and what's fairly standard by now is 20 core hyperthreaded Xeons. But mind you, without anything further in that cabinet, and preferably fully powered by a linear power supply. So no fans, no disks, no SSD, no USB connected stuff and also no SD. Nothing. Only LAN and the music stored somewhere there.

 

And regarding NAS ideas : Might that be beneficial in the first place, then try to envision it is not about any peripheral which doesn't spread noise to your precious Audio PC; it is about the Audio PC being able to run lean. Run evenly (no spiking).

[PeterSt]

On 24-12-2017 at 11:04 PM, ALRAINBOW said:

I have read so much on what sounds good and now after this conclusion I don't know how anyone can post win 10 is good .

 

Hi Al,

 

After better reading your original post, I only now see what it is all about. Should be your question above, right ?

OK. I am not sure whether I answered the question indirectly with my previous posts, but the answer in direct fashion should be : because all is so stripped down that it comes down to a server version. And now both server and normal W10 should sound the same or very similar. However :

 

Do not underestimate those .0 versions I mentioned. But maybe you did take that into account already ?

 

Peter

[PeterSt]

1 hour ago, Lebouwsky said:

 

How is the OS stripped down, is it something you provide as a service?

 

This is part of XXHighEnd (not the Demo version). It is downloadable software. It lets you boot into the stripped down OS.

Or if you refer to those pre-installed OSes (I showed the boot menu of them) - this is available on HDD/SSD.

 

Btw, maybe I should be sorry for posting in this thread. By now I don't understand the questions alrainbow purportedly has/had. Good that I actually responded to SandyK and Rando so it doesn't feel too much as a waste of time.

[PeterSt]

Just now, ALRAINBOW said:

But the stuff that gets posted that is contridicting is.

 

OK Al, thank you. But what is contradicting then ? The only contradictions I see is in your lists(S) of your own observations. At first it looked like questions to me (or asking for a second opinion). But it merely looks like you know all better than any of us. Btw, you don't need to be a high tech person as ears could be enough (no disagreement there).

 

Hey, I count 21 topics in your last list. Of those, there are 13 that I category disagree with. That is quite a lot.

So there's some contradiction indeed. 

 

Anyway, I thought to contribute a little. Now it looks like my opinion against yours or something. That wasn't my intention.

OK. All cleared up now. Everybody can have his opinion. Let's continue ...

[PeterSt]

10 minutes ago, ALRAINBOW said:

Not going to byte on your reply 

 

Al, was that addressed to me ?

 

11 minutes ago, ALRAINBOW said:

so picking on my grammar

 

I sure did not do that anywhere.

 

11 minutes ago, ALRAINBOW said:

how about you post a question or one of my confusing thoughts and let’s reply to it meaning ful just one thought at a time. 

 

That must be addressed to me. But I really don't know what you want now. I don't have a question and don't question anything either !

I only disagree with many things of your list. I myself don't care about it. We all have different opinions and experiences, right ?

But if you like, one example which should be easy to discern (and one of the 13 I disagree with) :

 

Quote

one ssd for os

 

a. SSDs are worse than spinning disks for SQ (my subjective opinion);

b. As I implied, I use nothing in the PC. No, I said that explicitly but maybe it was not clear.

 

I hope we can agree over "nothing is better". And I use nothing (all Operating Systems run purely from RAM).

 

The thing for you to consider is to what degree this small change alone would do to the audibility of the OS itself.

An other thing to consider would be : the more sources of "noise" you eliminate, the more others become profound and the harder it gets to *not* make it worse.

 

Here is one which should be explicitly NOT clear :

 

1 hour ago, ALRAINBOW said:

But allows for AO to do more including tidal. 

 

So you stream from Tidal;

Did it ever slip your mind that if you from one side do all your stinking best to eliminate OS processes but from the other side bring in a multifold of misery by means of streaming, that you put the horse behind the wagon ?

So you see, this is an example of you yourself being contradictionary to yourself.

 

I now realize this was 2 steps in one post, so I will stop for now.

No picking on grammar anywhere. You nay not notice readily, but my own English is worse. So I wouldn't dare to ever blame someone else for his/her language. Never.

OK, Belgians. Haha.

[PeterSt]

1 minute ago, semente said:

Have you ever tried your software with a low-power renderer/buffer akin to HQPlayer's NAA?

 

This question is quite similar to the question whether the Phasure DAC can be provided with I2S input to improve on the USB interface.

 

Low-power is unrelated, unless low-power can be seen as bad in itself (and I am of that stance).

Try to catch the CPU cycles XXHighEnd uses while playing 32/768. You won't. Not even with the processor running at 500MHz (like I use it myself).

The "Audio PC" as Phasure proposes it these days (since 4 years or so) is running nothing but a "sample renderer", all the other work (think FLAC decoding, MQA decoding, uhm ... streaming ) already done ahead of it. Of course you can think that pumping those samples over Ethernet is better for sound than "pumping nothing", but then maybe it is time to reconsider. 

It is a bit old-ish by now, but XXHighEnd ever back started out as a (first) memory player. It still is that for a full 100%. The fun (for me) with it, is that as a user you don't even notice anything of this.

And may it help (to set your mind), Audirvana+ is nothing like this. It looks to be like this, but it is not at all (it is always preparing tracks during playback - in XXHighEnd this is all done ahead of playback (like 2-4 seconds for a complete album)).

 

Btw, others for sure tried it.

[PeterSt]

9 hours ago, bodiebill said:

So what would be more desirable:

      monitor/HDMI but no SSD, no usb, no LAN (wtfplay setup)

or

      SSD, LAN but no HDMI (Windows/XXHighEnd setup)

?

 

Quote

      SSD, LAN but no HDMI (Windows/XXHighEnd setup)

 

Hey bodiebill,

 

If you'd use the RAM-OS feature then no SSD (or anything else) is needed.

When playback has started (for as much as your memory permits, say 4 Redbook albums) you can pull the LAN cable as well. And, because you're "obviously" working headless (no keyboard, mouse and monitor), you'd have to put back in the LAN cable for remote control.

I myself only pull pork but I know of people who pull the USB cable for the USB connected disks (instead of LAN) once playback has started.

 

Regards,

Peter

 

[PeterSt]

Maybe it makes sense to some when I myself add to the OS pile that the most important to SQ degradation is the use of a monitor. No wait, video card. So that should always be the base.

IMO this is not electrical noise related. Just OS related matter.

[PeterSt]

2 hours ago, semente said:

Can't you add a script that turns the Ethernet card on or off?

 

Sure !

But you make us one which turns it on again when we want, OK ? 

[PeterSt]

15 minutes ago, WuNgUn said:

4 hours ago, PeterSt said:

Maybe it makes sense to some when I myself add to the OS pile that the most important to SQ degradation is the use of a monitor. No wait, video card. So that should always be the base.

IMO this is not electrical noise related. Just OS related matter.

Are you talking about RF interference? Part of the reason I'm galvanically isolated (optical).

 

Merely peaky current draw. An SSD is the worst at that.

[PeterSt]

21 minutes ago, Summit said:

Can you explain why RAM-OS is better than SSD-OS? Ram is powered by and goes true the mobo while SSD can be powered by a LPS, so what do the ram memory do better?

 

I have the whole lot powered from a linear power supply. But you won't hear me say it is about that.

You eliminate tons of overhead when there's RAM only (only the disk emulator remains). Easy to see by the vastly increased speed of everything.

 

Think of this too :

Only the sheer fact that a CDRom drive (or DVD Drive etc.) (which is empty !), is connected already makes a very well audible difference.

[PeterSt]

Just now, WuNgUn said:

Or are you saying actual SQ is improved?

 

That.

All matters. I could talk about it for days (or write multi volume books about it).

[PeterSt]

4 minutes ago, WuNgUn said:

What about just disabling these devices in the device manager? Same thing?

 

That indeed matters too. But better is to shut off the device in the BIOS.

 

5 minutes ago, WuNgUn said:

If a DVD drive is plugged in, but not spinning, its are to imagine it effects anything!

 

When active (in BIOS / Device Manager) the OS kind of scans it (not physically, but in software). It goes way more deep because when certain devices are not present, the related services also don't need to run. Example : For USB hot-plug detection a service runs. When you are sure to never have any USB devices connected, that service can be disabled. If you use a USB DAC it is bad luck regarding this of course. In the end this extends to disabling USB in the BIOS as well. Also note that this can be for USB2 only, or for USB3 as well. And, use a USB PCIe card and the whole lot in the BIOS can be shut off (but choose one which is self-contained). This is just an example again.

Did I already say that I can talk about this for days ?

 

[PeterSt]

1 minute ago, rando said:

Peter, what are your thoughts on ECC versus non-ECC RAM's impact on SQ?

 

From (my own) theory only, non-ECC should be the better one.

Could try it with ECC for fun (using Xeon processors), but it is not so funny with the crazy prices of low profile RAM which I need for in the Stealth PC (anyone wants one ?).

 

What is your idea about it ?

[PeterSt]

10 minutes ago, bibo01 said:

In my cMP² days we discovered that motherboard power chips onboard (like audio or hdmi) even if they are disabled in bios.

 

Yes. CICS and me started about the same time, where he was more hardware oriented and I approached it via software. Along the lines this changed as many of my users/customers came up with such tweaks. The CDRom drive was a first as I recall it, and it took a year before I believed so much in that that I started to try such things myself.

In the end I seem to be the keenest in everything (I mean of my user group) - it is all just different thinking (and belief in what's there to accomplish).

 

PS: Especially for you @bibo01 : we just shipped another goody to Italy, so it looks like Italians start to get the hang of hit. Don't shut off your USB !

hahaha

[PeterSt]

25 minutes ago, ALRAINBOW said:

If a server has one drive alone just for OS 

the sound inproves of that drive is not using the mainboard bus power.  

Adding and internal drive for music it also needs its own v buss.  It’s not about power it’s abour noise going back to psu.  

 

Hi Al,

 

Noise is not just noise, it is a residue of using power. Hard to explain. But think of my "peaky current draw" and try to see that those peaks / spikes (of current use) are the noise we talk about. Of course there is also general noise (which can be filtered out) but this is less severe.

The noise also creeps in the the (SATA) data lines. And you can't avoid that ...

 

Quote

Have you tried this concept just to test it. ?

 

No, but many others did and do that.

Please notice : I do not use any drive at all. No SSD/HDD for the OS, no SSD/HDD for the music. NOTHING. The OS and everything runs 100% from RAM (internal computer memory).

So there is not much for me to test. 

 

[PeterSt]

34 minutes ago, ALRAINBOW said:

does any one part of a mainboard and its add ons IE video card , usb card , ssd drive network etc ? 

If yes it’s why it may better not to power the M2

card from Main board.

 

Al, if I understand you correctly, you say that when I don't use the M.2 slot, I better cut the power to it. Is that right ?

I don't use it (no peripheral devices anywhere - only the CPU !) but I now wonder where to be to cut that from the power. I did not investigate it either (yet). Or maybe you mean in the BIOS ?

 

? 

[PeterSt]

15 minutes ago, WuNgUn said:

On the other hand, others might say that modern CPU'S and memory and buses are not bottlenecked in anyway to introduce latency issues with audio data stream...

 

Who is talking about latency issues ?

[PeterSt]

Just now, WuNgUn said:

Well then, what issues are you experiencing with BT and USB adapters connected to your HTPC with regards to audio playback? 

 

 LOL. I am ONLY talking about SQ improvements. 

[PeterSt]

2 minutes ago, WuNgUn said:

How does removing hardware improve SQ?

 

Same as how tearing down the OS helps SQ.

[PeterSt]

27 minutes ago, semente said:

Easy. When the selected track or tracks end, the music player restarts the Ethernet card. 

 

You know, very occasionally something happens which lets me lose the control. Say that my LAN went down.

And do you think I am going to sit out that album I don't like or is too loud (or too soft) without a pre-amp ?

I rather reboot.

 

Not a good solution. 

[PeterSt]

18 minutes ago, seeteeyou said:

during routine "workstation type I/O" is only 220mA.

 

ONLY ?

That is more than my whole DAC uses !

 

I will read into your links tomorrow. 

[PeterSt]

8 hours ago, sandyk said:

16 hours ago, PeterSt said:

Merely peaky current draw. An SSD is the worst at that.

 

 That can be catered for as I have done, by regulating the internal +12V down to +5V, followed by a "shunt"  regulator

 

Alex, yes, smart (if it is a real shunt regulator and not only one between quotes ).