A novel way to massively improve the SQ of computer audio streaming

[lmitche]

4 hours ago, Johnseye said:

 

You're using WINS and not DNS?

Yes, for a samba server. I'm don't known what raatserver uses.

[lmitche]

4 hours ago, afrancois said:

I've been disconnected a few years of all the latest technologies regarding PC's, but AL has somehow retriggered my interest in PC hardware.

I found a Youtube video explaining M.2 NVMe SSD. So I guess if you want to go with an AL server and decide to have your storage in the server and not on a NAS that M.2 NVMe SSD is the way to go. Especially the explanation about less CPU overhead and latency sounded like music to my ears. My apologies if this has been mentioned before, but as in my job, I always inform my colleagues when I have discovered something new. Sometimes, of course, they already knew, but they did not bother to tell me. The latter doesn't seem to be happening here, luckily.

 

Now I'm starting to doubt if a simple NUC, like the NUC7PJYH2 is the best choice as an AL server. I guess as long as you keep all in RAM and don't use any attached drive this cheap NUC is OK.

 

 

 

 

I've run roonserver on three different hardware configs with AL:  Celeron dial core, i7-6700k z270 and i7 quad core NUC. The NUC i7 sounds best here.

[lmitche]

1 hour ago, mourip said:

 

Nope. First that I have heard of it. I believe that CA must be using it.

Yes, i vaguely recall Chris moved CA to cloudflare in the summer.

[lmitche]

30 minutes ago, afrancois said:

The NUC i7 was the only NUC you've tested one? Perhaps the NUCs small form factor is the contributing factor to sound quality.

Has anybody compared a NUC i7 with a NUC with Pentium Silver processor or something similar? As I'm not doing any DSD upsampling so I don't really need the raw power of an i7 processor. I'm just hoping that the same sound quality can be achieved by a "lesser" 

I've tested roonbridge on Celeron, Pentium and the i7 quad core NUCs. Each step is better. 

[lmitche]

29 minutes ago, afrancois said:

The NUC i7 was the only NUC you've tested one?

Sorry the Celeron dual core is a NUC7cjyh. 

[lmitche]

2 hours ago, mourip said:

 

I have scoured my network and only my FIOS router hands out IP addresses and handles DNS. I even added the IP addresses and host names of the three Roon components to the router's DNS but that did not help. I also added the hostnames to the Roon servers "hosts" file. No joy.

 

One more piece of information. If Bridge A is showing in Roon Control and I unplug it's ethernet cable then Bridge B shows up right away and ready to play music.

 

 

Please post the output of "$cat /etc/hostname" for all three NUCS.

[lmitche]

1 hour ago, mourip said:

 

Thanks Larry. After about a dozen hours of tearing apart my system, reinstalling the OS on three PC's, posting many questions here, and mail bombing poor Pietro he came through with the solution. It was not a DNS issue but one of a proprietary unique identifier that Roon is supposed to create upon first install. Even though I did a fresh download for both Bridge installs and used separate USB drives they somehow had the same "unique" identifier. This is why Roon Core would get so confused. Here is Pietro's reply. I hope that someone else can benefit.

 

"RoonBridge generates and writes out a unique id into its data directory on first run and uses that to identify itself to RoonServer. Clear out /var/roon on the second device (or hunt down and delete the RAATServer/Settings/unique_id file) and it will generate itself a new one next time it starts up."

 

I deleted /var/roon/RAATServer/Settings/unique_id and rebooted. I had to log in as root.

 

Now both Bridges show up as expected. It is odd to me that Roon gave me no error regarding a duplicate identifier or that the install could even allow it. Sadly Roon Support never came up with this information when I posted there.

 

Much thanks to all of you who offered your suggestions. It is much appreciated.

 

Paul

Oh wow! Who knew? Thanks so for much sharing.

[lmitche]

10 minutes ago, greenleo said:

Very interesting, the more powerful the endpoint, the better the SQ, contradicting to what we thought previously that higher power implies higher electrical noises.

 

Maybe it's worthwhile if CAers try adding RAM to the same NUC, check the SQ and report.  SQ wise it should not be better, may be equal or worse than the original.  If so, the theory of the electrical noise still works.  Also, the theory of the quicker CPU implies shorter latency still works.  It would be interesting to find the sweet point as well.

All of these NUCs can run powered with a lps1.2, even the i7 set in low power mode with every unused circuit, turbo and hyperthreading disabled in bios and one 4gb stick of ram.

[lmitche]

4 minutes ago, greenleo said:

Thank you Larry for the info.  Any SQ change from high power to low power mode?

Yes, low power sounds better as an endpoint but hard to say if it's the mode switch or power source that's the reason. I don't have a power supply above the lps1.2 current limits that is anywhere near it's quality.

 

That may change soon. Stay tuned.

[lmitche]

2 minutes ago, rettib2001 said:

 

I suppose my real question is whether or not a celeron/pentium nuc as an endpoint powered by an SR4 with a seperete i7 nuc as a server (powered by an El cheapo power supply) both running headless Al in ram (extreme mode) is any better than simply using an i7 nuc powered by the SR4 (again headless Al ram + extreme mode). 

 

Between the two would be a lpsu, ground shunted ethernet switch. 

 

To answer your question we would need to know where the music is stored in the second scenario.

[lmitche]

25 minutes ago, rettib2001 said:

No local storage, I exclusively use tidal hi-fi. 

That's is a good question. Without listening I can't know for sure. I'll add this scenario to the listening test queue. I have roon here and can easily test it in a one and two box setup.

 

While I am at it, is there a preferred tidal app for Linux if any? If so that test may make sense as well.

[lmitche]

9 minutes ago, matthias said:

@lmitche

Hi Larry,

is there consensus about the best NUC solution for streaming from Qobuz and using wifi only (no NAS, no Roon, no wired ethernet).

Thanks

 

Matt

Hi Matt,

 

We don't have qobuz here in the US yet. Nevertheless I would expect the answer to be similar to the tidal SQ testing suggested above.

[lmitche]

2 minutes ago, HeeBroG said:

@Imitche

 

Do you know of a way to use DRC/Parametric Eq with AudioLinux outside of ROON/HQplayer

I don't, but I am the wrong guy to ask. My only experience is with the Roon and Hqplayer solutions.

 

@hifi25nl may know more.

[lmitche]

1 minute ago, tapatrick said:

I only stream from Tidal via Roon, no local music and without a doubt NUC server  plus NUC endpoint is more superior to just NUC endpoint, as good as that is on its own. (See my post above). 

Thanks Patrick, you saved me some time. It seems intuitive that a two box system sounds better with Tidal, but the confirmation is very helpful.

[lmitche]

Could you please be more precise about the definition of a wtf question?

[lmitche]

2 hours ago, Soma said:

Hi all, quick check nuc7pjyh model maximum memory on the intel website says 8GB. Does it mean 8gb for each slot or 8gb for entire system (sorry I am technically disabled). I have ordered 4gb*2 for NUC. If it can take 16gb, I will cancel and buy 8gb*2. Thanks

Well 16 gb worked here. However for an endpoint, less is more.

[lmitche]

1 hour ago, rettib2001 said:

Given the highly subjective nature of our personal findings, this might not necessarily be of interest to many but the nuc7i7bnh (and I assume the more recent i7 nucs) handles upsampling to dsd512 well in terms of playback (smooth/uninterrupted) but the cpu temps are a little worrying when this is done under AL headless extreme. 

 

CPU temps, in this scenario, are hitting 90°.

 

Adding convolution into the mix obviously doesn't do them any favours either. 

 

CPU usage is around 120% when using these 'to the max' settings which suggests the system is being pushed a little beyond what it can comfortably handle, however as I said earlier playback is smooth and as blip/pop/click free as I've heard dsd512. 

 

To my ears the use of the extreme mode offers a sufficiently valuable improvement in sound quality to warrant exploring more creative cooling approaches...

 

 

 

That's interesting. I'm surprised this dual core model is working at dsd512 speeds. Are you thinking about a fanless case?

[lmitche]

1 hour ago, rickca said:

I can't stop thinking ... why didn't we hear about this great NUC/AL combination until recently?  It's unbelievably inexpensive and it's been available for quite some time.  Are we all too preoccupied with ever more expensive sophisticated components and elaborate explanations about their design?

Rick,

 

While I may have been the last kid to the party, I first heard about the headless ramboot version of AL from Piero in mid September. The full SQ impact of what we are now hearing wasn't available, at least to me, until then.

 

So I'd argue this is a recent solution.

 

Larry

[lmitche]

46 minutes ago, mozes said:

Excellent question. This is a great proof that expensive doesn't always mean better SQ 

Or that inexpensive means poor SQ.

[lmitche]

24 minutes ago, Middy said:

8 Gb enough for the server roon core side? 

Yep

[lmitche]

6 minutes ago, Em2016 said:

 Obviously the “xtr-2s” filter is not an option with this NUC at DSD512

Why is xtr-2s not an option? Do you mean xtr non-2s?

[lmitche]

30 minutes ago, Em2016 said:

 

If it wasn't clear, this NUC’s CPU can’t do xtr-2s at DSD512...

 

Thanks Em2016, 

 

I haven't explored hqplayer desktop or server on the NUC as Roon running bitperfect sounds so great there hasn't been a need. Also I could never get Roon and Hqplayer to run without a tolerable, but nevertheless annoying level of glitchyness that is now gone.

 

I'm surprised to see these results. The NUC quad core CPU appears so close to an i7-6700k in benchmarks it didn't occur to me that it was deficient in this area.

 

Thanks,

 

Larry

[lmitche]

2 minutes ago, Whitigir said:

I7-8700K can barely do XTR-2S into DSD512.  Let alone any weak CPU.  Non-2S XTR is defintely too much for I7-8700K

My i7-6700k has no problem with xtr-2s at dsd512 with hqplayer desktop. It hardly breaks a sweat. It will also do xtr non-2s at Dsd256 but the heat builds up for sure.

[lmitche]

13 minutes ago, Em2016 said:

 

CUDA offloading helping?

No, I don't have a graphics card.

[lmitche]

33 minutes ago, Em2016 said:

Side note... I tried AL on my NUC7I7DNHE running HQPe and RoonServer. Heard absolutely no differrence with stock Debian Stretch, playing to my Rendu NAA so I went back to Debian so that I can do other things.

 

I’ve also tried AL as NAA endpoint on my NUC and Jussi recommended Up Board Gateway and thought both sounded a bit brighter than my Rendu, causing slight listening fatigue over longer listening sessions...

 

An ISO REGEN + LPS-1.2 on the output (like Larry prefers) would probably help the NUC and UpBoard Gateway.

What power supplies are you using for your two NUCs?