Jump to content
austinpop

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

Rate this topic

Recommended Posts

6 hours ago, romaz said:

 

I agree.  As I find myself looking for a certain post by a certain individual, leafing through nearly 1,500 posts over 58 pages is no small task.  While I meant it to be a discussion of a novel way to improve either the microRendu or sMS-200, it has unintentionally evolved from this direct connection between server and endpoint via bridged LAN ports to power supplies, clocking, ideal music server configurations, optimized OS solutions, motherboards, ideal storage media, SATA cables, Ethernet cables, endpoints other than the mR or sMS-200, and galvanic isolation.  

 

For those just jumping in, I can imagine that this thread looks like one big disjointed mess but as someone who has been a part of it from the beginning, and speaking only for myself, it has been a wonderful and stimulating think tank where innovative ideas and experiences have been shared, discussed, affirmed and challenged.  Nothing advances progress better than a broad community effort and without a doubt, my server build would look vastly different without the input and inspiration provided by so many.    

 

If all goes well, my reclocked motherboard will arrive from SOtM sometime next week.  At that point, I will report on the impact of removing as many bad clocks as possible from the server and whether this negates or further enhances the impact of endpoints like the mR or sMS-200.  Despite the broad topics that have been discussed, for me, it has still always been about finding ways to improve upon the mR or sMS-200.  Shortly after that, I will be making my exit from this thread (and from posting on forums, in general), at least for the foreseeable future due to other more pressing and time-consuming commitments.  Because we are talking about a few weeks of time, I will likely make no attempt to start any new threads although I certainly won't discourage anyone from breaking this apart into dedicated sub-topics.  It's been a fun ride...

Please say it isn't so. I have not been this enthused about 'reading about computer audio' since the 'back in the day of cics CMP/Cplay massive threads in Audio Asylum. Whatever you do I wish you well but your writings have lifted my audio SQ as well as my hope in independent writing sources about computer audio.

Share this post


Link to post
Share on other sites
11 hours ago, romaz said:

If all goes well, my reclocked motherboard will arrive from SOtM sometime next week.  At that point, I will report on the impact of removing as many bad clocks as possible from the server and whether this negates or further enhances the impact of endpoints like the mR or sMS-200.  Despite the broad topics that have been discussed, for me, it has still always been about finding ways to improve upon the mR or sMS-200.  

 

Just to add my own teaser - I have a similar project underway at SOtM as well, similar to Roy's first sCLK-EX experiments, but using the tX-USBultra rather than the dX-USB HD Ultra. I too am hopeful to receive my gear back in the next week, maybe two. As we get closer, I'll talk about the experiments I have planned.

 

11 hours ago, romaz said:

Shortly after that, I will be making my exit from this thread (and from posting on forums, in general), at least for the foreseeable future due to other more pressing and time-consuming commitments.  Because we are talking about a few weeks of time, I will likely make no attempt to start any new threads although I certainly won't discourage anyone from breaking this apart into dedicated sub-topics.  It's been a fun ride...

 

It takes time and effort to participate in these discussions, and I am profoundly grateful to Roy for sharing so much of his knowledge, insights and findings with us.

Share this post


Link to post
Share on other sites

Although I have barely posted anything on forums, I have been following this thread from the start and have read every post. It has been thoroughly enjoyable and extremely informative. Especially all of @romaz's posts. Thank you it's been fun ?

Share this post


Link to post
Share on other sites
4 hours ago, austinpop said:

It takes time and effort to participate in these discussions, and I am profoundly grateful to Roy for sharing so much of his knowledge, insights and findings with us.

 

+ 1,000,000

 

Thanks @romaz

Edited by kennyb123

Share this post


Link to post
Share on other sites
20 minutes ago, kennyb123 said:

 

+ 1,000,000

 

Thanks @romaz

Yes, Roy's out of the box thinking, his experience, his wriiting skills, and his overall knowledge base are invaluable to veteran CA'ers and newbies alike.  Come back occasionally man!

Share this post


Link to post
Share on other sites

Thank you @romaz for all the knowledge you shared with us. I do hope that we still read some posts from you from time to time. You are a great asset to CA and I have enjoyed reading each one of your posts!

Share this post


Link to post
Share on other sites
48 minutes ago, austinpop said:

Hey hey he hasn't left yet! :D

 

I still want to hear about the mobo clock mods!!

 

Agreed on both counts. 

 

I'm wondering if it's possible to eliminate the need for the endpoint. 

 

Share this post


Link to post
Share on other sites

Hi all!

 

I use Windows 2016 GUI mode as a NAS server for my SMS-200 and I use a network bridge to directly connect this streamer to the NAS.

As you know, network bridge functionality is not available in a core installation of Windows 2016 Server.

 

I have found this utility, which permit to create a network bridge using command line (but I didn't try it yet on a windows 2016 Core installation) https://sourceforge.net/projects/networkbridgingutility/

 

Does someone already tried if this tool helps to create a network bridge on a core installation of Windows 2016 Server?

 

If this is the case, that could be a very good news!

 

CAT

Share this post


Link to post
Share on other sites
3 hours ago, romaz said:

Roy, these motherboards are interesting indeed.  It's also cool that you can supply power via an ATX cable as well, bypassing the onboard DC to DC converters if you choose.  Nevertheless, it's hard to tell which of these has the performance required for Hqplayer upsampling at dsd512 levels, and at these prices it's an expensive experiment.

Edited by lmitche

Share this post


Link to post
Share on other sites

Amazon is selling X10SDV-2C-TP4F (i.e. X10SDV-2C-7TP4F w/o SAS) for 350 bucks, and it's as low as $328.99 somewhere else with 8.8/10 ratings @ resellerratings.com

 

http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-2C-TP4F.cfm

 

However, I checked the pictures of X10SDV-7TP4F and couldn't figure out where the clock(s) could be found

 

Jz0DtZp.jpg

http://i.imgur.com/Jz0DtZp.jpg

https://forums.servethehome.com/index.php?threads/supermicro-x10sdv-7tp4f-xeon-d-1537.9349/

 

One of these guys next to the Mini PCI-E slot and SoC seemed to be close enough but who knows?

 

LkgSMK0.png

Share this post


Link to post
Share on other sites
6 hours ago, romaz said:

I appreciate the warm sentiments!  Thank you.  Like Rajiv said, I haven't gone anywhere yet and will be around until the big audio show in Munich in May.  After the show, my focus will need to be elsewhere and so forgive me if I don't respond publicly or to PMs after that time.

 

On a side note, I came across a Supermicro server-grade motherboard that could be "the one."  If only I wasn't so far into my build, I would seriously consider making a switch but for those of you contemplating your own build, give this motherboard line a look.  It is their X10SDV line and is an SoC (System on a Chip) motherboard.  This means the CPU is built-in.  Fortunately, they utilize the more power efficient 14nm Intel CPUs and range from a 2-core Pentium D1508 (25w TDP) to a 4-core Xeon D-1518 (35w TDP) to an 8-core Xeon D-1537 (35w TDP).  What is interesting is that when a certain model quotes a TDP of 25w, they're not talking about just the CPU but the entire system!  If there is a downside, these CPUs probably can't be overclocked for those who desire this.  Moreover, while they contain healthy amounts of cache, they don't use Intel's "SmartCache" and so each core can only access 1.5MB of cache instead of the whole cache.

 

Here is what makes this line of motherboards so interesting:

 

1.  These boards can be powered by a single 12V rail.  ATX not necessary.

 

2.  All models contain both dual gigabit LAN ports and dual 10G SFP+ (fiber) ports.

 

2.  They contain 2 PCIe 3.0x8 slots, an M.2 PCIe 3.0x4 slot and an mSATA slot and each of these slots have direct access to the CPU.

 

3.  PCH is integrated.  This is perhaps what is most interesting and apparently is typical of all SoC motherboards.  As previously discussed, the disadvantage that many motherboards have is that most buses (i.e. SATA, USB, LAN, video, audio) connect to the CPU via a PCH (Platform Controller Hub).  Back in the day, we used to refer to this as Southbridge although today, PCH is the more correct term.  While the "first class" Northbridge connected directly to the CPU via the front-side bus (FSB), the "coach class" Southbridge (or now, the PCH) connected to the CPU via a slower and more congested Direct Media Interface (DMI) and so having to go through this DMI usually represented an added latency.  Because this PCH generally utilizes a 1.05V rail, in theory anyway, this is probably also a noisier pathway.  

 

If you look at the block diagrams of most modern motherboards, it is generally only RAM and the faster PCIE 3.0 buses that use the more direct FSB while everything else goes through the slower, more congested and noisier PCH.  Here is a typical motherboard block diagram:

 

PCH.thumb.jpg.d93834f4099503105811dd6afe5fb399.jpg

 

Well, as this Supermicro SoC motherboard "integrates" the PCH, there is no more DMI.  It would appear that all important ports (whether it be USB, SATA, copper LAN, fiber LAN, video) now have direct access to the CPU.  Also, gone is the noisier 1.05V bus that is used to power an external PCH.  It appears each component now accesses the CPU via it's native 12V, 5V, 3.3V or 1.8V rail.  At least, this is what I am interpreting.  Here is the block diagram for this mobo line:

 

58f858b4e40a4_ScreenShot2017-04-19at11_43_12PM.thumb.png.1088fe0939078cdb9430476f333df4e2.png

 

And so what this means is that you now have a fairly compact sized motherboard ("flex-ATX" is a little smaller than uATX) that can be powered by a good 12V LPSU with numerous very low latency connections to the CPU including dual fiber LAN ports that can be bridged.  This may be as close to an "ideal" audiophile motherboard as I have seen.

 

While these motherboards aren't cheap, the price includes the CPU, dual copper gigabit and dual fiber 10G LAN ports.  The base model that includes the dual core Pentium sells for about $500.

 

http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-2C-7TP4F.cfm

 

http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-4C-7TP4F.cfm

 

http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-7TP4F.cfm

 

Thanks Roy.

Here's the 8 core Xeon.  Eight 2.3GHz cores at 35w.  This could be a winner for upsampling.

https://ark.intel.com/products/91196/Intel-Xeon-Processor-D-1537-12M-Cache-1_70-GHz

 

The board is about $1k however, depending on who you get it from.

 

If only someone would build an endpoint with a fiber SFP.

 

I'm still hoping Roy's board with upgraded clocks eliminates the need for an endpoint.  This is probably wishful thinking but should be the ultimate goal.

 

Share this post


Link to post
Share on other sites
4 hours ago, Nouchka said:

Hi all!

 

I use Windows 2016 GUI mode as a NAS server for my SMS-200 and I use a network bridge to directly connect this streamer to the NAS.

As you know, network bridge functionality is not available in a core installation of Windows 2016 Server.

 

I have found this utility, which permit to create a network bridge using command line (but I didn't try it yet on a windows 2016 Core installation) https://sourceforge.net/projects/networkbridgingutility/

 

Does someone already tried if this tool helps to create a network bridge on a core installation of Windows 2016 Server?

 

If this is the case, that could be a very good news!

 

CAT

I tried before.  The tool doesn't work.

Share this post


Link to post
Share on other sites
4 hours ago, Nouchka said:

Hi all!

 

I use Windows 2016 GUI mode as a NAS server for my SMS-200 and I use a network bridge to directly connect this streamer to the NAS.

As you know, network bridge functionality is not available in a core installation of Windows 2016 Server.

 

I have found this utility, which permit to create a network bridge using command line (but I didn't try it yet on a windows 2016 Core installation) https://sourceforge.net/projects/networkbridgingutility/

 

Does someone already tried if this tool helps to create a network bridge on a core installation of Windows 2016 Server?

 

If this is the case, that could be a very good news!

 

CAT

Hi CAT,

 

I tried the utility first on W10 : does not work, does not recognize my NICs, does not recognize a bridge (if made under W10).

Under W2016-core : does not start at all (unspecified error).

On Github the author states that the program is for W7 and does not even work for W8. So this is not a solution.....

 

P.S. Greenleo just answered 2 secs before me

Edited by Peter_T

Share this post


Link to post
Share on other sites
38 minutes ago, Johnseye said:

If only someone would build an endpoint with a fiber SFP.

For NAA this is a possibility

 

Share this post


Link to post
Share on other sites
8 hours ago, lmitche said:

Roy, these motherboards are interesting indeed.  It's also cool that you can supply power via an ATX cable as well, bypassing the onboard DC to DC converters if you choose.  Nevertheless, it's hard to tell which of these has the performance required for Hqplayer upsampling at dsd512 levels, and at these prices it's an expensive experiment.

 

 

It's too bad there's no way to swap in a different CPU.  This is this motherboard's biggest drawback.  It's possible you can overclock and go beyond the 35w TDP rating and maintain stability provided that your PSU is up to snuff (which yours is) but overclocking Xeons are tricky (I'm not sure it's even possible).  Do you have a sense of the minimum hardware requirement to upsample to DSD512?

Share this post


Link to post
Share on other sites
6 hours ago, seeteeyou said:

Amazon is selling X10SDV-2C-TP4F (i.e. X10SDV-2C-7TP4F w/o SAS) for 350 bucks, and it's as low as $328.99 somewhere else with 8.8/10 ratings @ resellerratings.com

 

http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-2C-TP4F.cfm

 

However, I checked the pictures of X10SDV-7TP4F and couldn't figure out where the clock(s) could be found

 

Jz0DtZp.jpg

http://i.imgur.com/Jz0DtZp.jpg

https://forums.servethehome.com/index.php?threads/supermicro-x10sdv-7tp4f-xeon-d-1537.9349/

 

One of these guys next to the Mini PCI-E slot and SoC seemed to be close enough but who knows?

 

LkgSMK0.png

 

 

You're right, the model you are quoting is the actual model I had initially looked at because I was never interested in the SAS option and so thanks for bringing that up.  $330 including the CPU is much easier to swallow.

 

Clocks are not always easy to see.  I believe what you highlighted are clocks.  On close up view, if they say something like "25MHz", then they are definitely clocks.  In my ASRock motherboard, I had assumed there were 4 clocks that could be swapped out but once SOtM received it, they told me there are actually 6 clocks including the main system clock at 25MHz, 2 LAN clocks at 25MHz each, and 3 server management CPU clocks at 25, 48 and 50MHz.  Possibly for space reasons, ASRock chose to encase 2 of the clocks in much smaller enclosures.  Same thing could be happening with this Supermicro board.

 

As I reviewed Intel's white papers on this SoC chipset, it relies on an outboard 25Mhz crystal clock for its timing and then provides subtimings (via subclocks) to its various buses (PCIe, SATA, USB, video, etc).  Ideally, a clock is placed as close as possible to whatever it is clocking and so that may provide a clue.

Share this post


Link to post
Share on other sites
4 hours ago, ted_b said:

For NAA this is a possibility

 

 

 

Thanks, Ted.  I've been following this thread also.  It looks to be an SoC device also that uses an ARM chip.  Some SoC boards supposedly use only integrated clocks (within the integrated PCH) and it is unclear if those clocks can be accessed and replaced.  Otherwise, this board that he custom specified could be an ideal endpoint.

 

Alternatively, you could always get a TP Link FMC as a receiving module that can also be used as a reclocking device.  From there, you can connect to an mR or sMS-200 via copper.  I opened my FMC up and it has one clock that can be replaced along with several switching regulators that can be upgraded.  

Share this post


Link to post
Share on other sites
1 hour ago, romaz said:

 

Thanks, Ted.  I've been following this thread also.  It looks to be an SoC device also that uses an ARM chip.  Some SoC boards supposedly use only integrated clocks (within the integrated PCH) and it is unclear if those clocks can be accessed and replaced.  Otherwise, this board that he custom specified could be an ideal endpoint.

 

Alternatively, you could always get a TP Link FMC as a receiving module that can also be used as a reclocking device.  From there, you can connect to an mR or sMS-200 via copper.  I opened my FMC up and it has one clock that can be replaced along with several switching regulators that can be upgraded.  

Or just get one of these from Adnaco:

 

adnaco_rax_5gt_s_remote_small_form_factor_expansion_adapters.jpg

Share this post


Link to post
Share on other sites

I apologize if this has been asked before, but now I have a PC bridging the network between the router (which has the SonicOrbiter I5 plugged into with my HD attached to that) and the mRendu. Okay, so if I run a DSP engine on the PC will it pass on the digital processing of the 1s and 0s coming out of the PC to the mR, which in turn will apply DSP the music? Mind you, this is using Roon. 

 

Hmmm... Thinking about it, wouldn't I have to point the signal to Roon somehow on the PC, in which case this wouldn't work if I wanted to control Room remotely? I think so, but I could be wrong here. 

Share this post


Link to post
Share on other sites
On 4/23/2017 at 5:53 PM, EVOLVIST said:

I apologize if this has been asked before, but now I have a PC bridging the network between the router (which has the SonicOrbiter I5 plugged into with my HD attached to that) and the mRendu. Okay, so if I run a DSP engine on the PC will it pass on the digital processing of the 1s and 0s coming out of the PC to the mR, which in turn will apply DSP the music? Mind you, this is using Roon. 

 

Hmmm... Thinking about it, wouldn't I have to point the signal to Roon somehow on the PC, in which case this wouldn't work if I wanted to control Room remotely? I think so, but I could be wrong here. 

 

 

From my experience, all of Roon's functionality remains intact using this "direct connection."  Please report back if you find that this isn't the case but I would be surprised if it isn't.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×