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

[lmitche]

You may want to run traceroute on the daphile machine with both the microRendu and sms-200. You may find the microRendu packets take a different path.

[gadgetman]

You may want to run traceroute on the daphile machine with both the microRendu and sms-200. You may find the microRendu packets take a different path.

Well, as expected, both mR & sMS-200 go to Daphile directly.

 

Does your mR stop by any place that it shouldn't.

[lmitche]

Well, as expected, both mR & sMS-200 go to Daphile directly.

 

Does your mR stop by any place that it shouldn't.

That's good to know. I don't have have a microRendu, but whenever I read about latency and dropout issues I think it worthwhile to check the routing. I have had both Windows and Linux machines choose to route via routers several hops away even though they are directly adjacent.

 

Unexpected this is tough to check from the microRendu as you can't SSH directly into the box. Is there an SSH door on the sms200?

[gadgetman]

That's good to know. I don't have have a microRendu, but whenever I read about latency and dropout issues I think it worthwhile to check the routing. I have had both Windows and Linux machines choose to route via routers several hops away even though they are directly adjacent.

 

Unexpected this is tough to check from the microRendu as you can't SSH directly into the box. Is there an SSH door on the sms200?

Yes, both mR and sMS-200 support ssh. I do have login info of sMS-200 but not mR. It's provided by May@SOtM. But she didn't want me to spread it. It's better you ask her in PM if you need it.

 

Or, you think of any further tests like me to do? Please feel free to tell me. I'd be glad to do it.

[tboooe]

You may want to run traceroute on the daphile machine with both the microRendu and sms-200. You may find the microRendu packets take a different path.

 

The problem with traceroute is that it doesnt show if the path goes through a switch. It only shows if the data hops through a router. I tested this playing music local on my server machine then disconnecting it from my switch. If the music still plays then the server is communicating directly with the downstream device.

Well, as expected, both mR & sMS-200 go to Daphile directly.

 

Does your mR stop by any place that it shouldn't.

Your results are really interesting. I would expect any directly connected device to have a very short ping time. Even on my el cheapo i3 pc I use an NAA I get <1ms ping times, though for some reason, every 4th or 5th ping I get much higher times (4-5ms). My unsupported opinion is that there are major differences in hardware between the microrendu and sms200. I believe the usb implementation alone has to be very different because the microrendu uses the regen technology and power topology.

[lmitche]

Yes, both mR and sMS-200 support ssh. I do have login info of sMS-200 but not mR. It's provided by May@SOtM. But she didn't want me to spread it. It's better you ask her in PM if you need it.

 

Or, you think of any further tests like me to do? Please feel free to tell me. I'd be glad to do it.

It would be interesting to hear the traceroute results to your daphile box from the sms-200. If nothing else it would confirm no router interaction.

[gadgetman]

Your results are really interesting. I would expect any directly connected device to have a very short ping time. Even on my el cheapo i3 pc I use an NAA I get <1ms ping times, though for some reason, every 4th or 5th ping I get much higher times (4-5ms). My unsupported opinion is that there are major differences in hardware between the microrendu and sms200. I believe the usb implementation alone has to be very different because the microrendu uses the regen technology and power topology.

I agree there's some difference in h/w implementations. But I suspect it's in networking design. Sonore mR uses Solid-Run SOM which is based on i.MX6. The network performance is limited to 480Mbps max because of internal architecture. I suspect SOtM uses external Ethernet chip to achieve better effiency. May be @May could tell us more.

[gadgetman]

It would be interesting to hear the traceroute results to your daphile box from the sms-200. If nothing else it would confirm no router interaction.

OK. I'll do it sometime tomorrow. It's kind of late here now.

[Cooler]

Guys, can you please help me get working roon + hqplayer on one PC(Win10, two ethernet PCI-E devices) with direct connect of mR.

I tried many times, the same problem, PC doesnt see mR in NAA mode. In RoonReady it works fine, roon sees it without a problem.

The best result i got, when i add static IP to bridge with subnet different to gateway. Gateway is 192.168.0.x, ip of bridge 192.168.1.xx. Then i see mR in NAA on iMac, connected to router via wifi and to this router connected PC via ethernet cable. But then PC doesnt have internet connection and of course it doesnt see mR

 

As i understood i have to give to mR ip from another subnet during the bridge mode, but its impossible, as windows gives an option to adjust IP of a bridge and do other magic by itself.

[greenleo]

Guys, can you please help me get working roon + hqplayer on one PC(Win10, two ethernet PCI-E devices) with direct connect of mR.

I tried many times, the same problem, PC doesnt see mR in NAA mode. In RoonReady it works fine, roon sees it without a problem.

...

As i understood i have to give to mR ip from another subnet during the bridge mode, but its impossible, as windows gives an option to adjust IP of a bridge and do other magic by itself.

 

I'm using Win 10 as well. No need to set separate subnet.

Onboard LAN to router (DHCP)

PCIe LAN to mR (DHCP) then add both of this to form a network bridge (DHCP)

Attach a WiFi dongle then add it to the bridge.

Then HQPlayer should be able to find mR in NAA mode and play.

You may now eject the dongle.

 

Never try roon. Hope this method helps. I tried to used DHCP server in Win10 but didn't work out.

[gadgetman]

It would be interesting to hear the traceroute results to your daphile box from the sms-200. If nothing else it would confirm no router interaction.

The tracepath (name of traceroute in Fedora) shows only one IP address which is Daphile's.

 

Here is the ping results from sMS-200 to Daphile.

 

With switch

100 t, 100 r, 0 loss, 98999ms, rtt min/avg/max/Mdev = 0.355/0.419/0.486/0.035 ms

 

Without switch (direct connection)

192.168.128.1 (Ethernet port connected directly to sMS-200)

100 t, 100 r, 0 loss, 99000ms, rtt min/avg/max/Mdev = 0.295/0.338/0.424/0.031 ms

192.168.2.70 (Ethernet port connected to Internet)

100 t, 100 r, 0 loss, 99000ms, rtt min/avg/max/Mdev = 0.286/0.334/0.454/0.034 ms

[Cooler]

I'm using Win 10 as well. No need to set separate subnet.

Onboard LAN to router (DHCP)

PCIe LAN to mR (DHCP) then add both of this to form a network bridge (DHCP)

Attach a WiFi dongle then add it to the bridge.

Then HQPlayer should be able to find mR in NAA mode and play.

You may now eject the dongle.

 

Never try roon. Hope this method helps. I tried to used DHCP server in Win10 but didn't work out.

Ok, can you write more detailed?

So i connected mR and router to PC, created bridge, assigned static IP to this bridge, then added wifi dongle, connected it to router and added to bridge. Are these steps correct? As before on another computer i HQPlayer sees mR, but PC with mR doesnt see it.

Did you assign static IPs to all of them separately? Should there be some rule?

[Nouchka]

Hi,

 

I was using my sotm sms-200 directly connected to my win 2012 r2 server with a bridge connection without any issues for more than a month. Yesterday I did a clean install of win 2016 server, but I cannot make the sotm sms-200 working with a direct connection with a network bridge on this new version of windows.

I would like to know if others peoples had the same issues... Is there something different in the configuration of a network bridge with windows 2016 server?

 

Thanks in advance,

[Peter_T]

Hi,

 

I was using my sotm sms-200 directly connected to my win 2012 r2 server with a bridge connection without any issues for more than a month. Yesterday I did a clean install of win 2016 server, but I cannot make the sotm sms-200 working with a direct connection with a network bridge on this new version of windows.

I would like to know if others peoples had the same issues... Is there something different in the configuration of a network bridge with windows 2016 server?

 

Thanks in advance,

 

Bridge is not possible in core (w2012 & w2016) but is possible in W2012-GUI and minimal server mode and should be possible in 2016-GUI (I did not test that in 2016-GUI) see this thread

[Nouchka]

Bridge is not possible in core (w2012 & w2016) but is possible in W2012-GUI and minimal server mode and should be possible in 2016-GUI (I did not test that in 2016-GUI) see this thread

 

I have a GUI installation of Windows 2016. I have tried all differents ways possible but couldn't make it working ... In fact I see the problem: when I create the bridge, the new connection doesn't receive an IP Address from the DHCP. I can manually configure the IP Address but that doesn't work either. If I do exactly the same on Win2012 or Win10 on the same computer, that works directly. So, that could be related to a driver problem of my Intel server mainboard. Does someone have similar problems when trying to create a network bridge in Win 2016?

 

Thanks in advance,

[Cooler]

Big thanks to clipper, now my PC sees directly connected mR in NAA mode!

So what i did.

I plugged mR to PC, PC to router via ethernet, i also plugged wifi usb adapter to PC and connected PC with router via wifi (installed all drivers and checked connection). Then chose all of them and created bridge, rebooted PC and mR, then added static IP to bridge.

Rebooted again. Now i see mR on PC.

But i realized, if i remove wifi adaptor, no longer HQPLayer sees mR, but if i turn off wifi in menu, music still works, so wifi adaptor will be connected there forever!

[dman]

I have a GUI installation of Windows 2016. I have tried all differents ways possible but couldn't make it working ... In fact I see the problem: when I create the bridge, the new connection doesn't receive an IP Address from the DHCP. I can manually configure the IP Address but that doesn't work either. If I do exactly the same on Win2012 or Win10 on the same computer, that works directly. So, that could be related to a driver problem of my Intel server mainboard. Does someone have similar problems when trying to create a network bridge in Win 2016?

 

Thanks in advance,

 

I was able to establish a network bridge in Win Server 2016, the exact same way as in windows 10. Using DHCP for all ports. I am using the onboard ethernet port and a PCI E ethernet card for the bridge. I know you were asking if someone had similar problems, but I don't remember anybody saying that they definitely got it working with Server 2016. So it should be possible. Sorry, I do not know how help. Hopefully, someone else will have some ideas.

 

 

Sent from my iPad using Computer Audiophile

[greenleo]

Big thanks to clipper, now my PC sees directly connected mR in NAA mode!

So what i did.

I plugged mR to PC, PC to router via ethernet, i also plugged wifi usb adapter to PC and connected PC with router via wifi (installed all drivers and checked connection). Then chose all of them and created bridge, rebooted PC and mR, then added static IP to bridge.

Rebooted again. Now i see mR on PC.

But i realized, if i remove wifi adaptor, no longer HQPLayer sees mR, but if i turn off wifi in menu, music still works, so wifi adaptor will be connected there forever!

 

Glad to see that your system is now working. Clipper helped me as well.

 

You may even eject and remove the WiFi adapter. HQP still works. You may check if this sounds better :-)

You may disable the LAN that connected to the router as well. HQP still works for at least two hours (see my earlier post) and a total isolation from the router and the WiFi dongle is possible. You may check if this sound better :-)

This is fun :-)

But if your system restarts, you need to setup the bridge again :-( but it just take 2 minutes. What I did was to enable the LAN before shutting down Win10 and reinsert the WiFi adapter just before starting up Win10. The bridge will be re-built automatically so far.

 

Cheers.

 

Enjoy.

[greenleo]

I have a GUI installation of Windows 2016. I have tried all differents ways possible but couldn't make it working ... In fact I see the problem: when I create the bridge, the new connection doesn't receive an IP Address from the DHCP. I can manually configure the IP Address but that doesn't work either. If I do exactly the same on Win2012 or Win10 on the same computer, that works directly. So, that could be related to a driver problem of my Intel server mainboard. Does someone have similar problems when trying to create a network bridge in Win 2016?

 

Thanks in advance,

 

Was the bridge for HQPlayer or what?

 

For the bridge that you've set up, have you used ipconfig to check the IP addresses of the tow LANs and then use ipconfig to check the IP address of the bridge formed? I will use DHCP for both of the LANs initially to check if the bridge have been setup correctly.

 

Hope this helps.

[WynnHiFi]

I have been following this thread for months trying different combinations of optical and Ethernet isolation devices with bridge and direct connections and was unsatisfied. My Oppo 105 is already bright, then add a pair of FMCs and it’s too bright or sibilant. I returned to my previous setup because the music was too harsh or fatiguing.

Yesterday I created a bootable SD card with OS and JRiver. Then I cracked open the MacMini to removed the Airport card, disconnected the HDD, and installed an UpTone Audio fan kit with JS-2 linear power supply. Next I Booted up and bridged 2 Apple Thunderbolt Ethernet adapters for the router and NAS, then direct connected the MacMini Ethernet port to the microRendu, and the music is now warmer, open, and more detailed. Hate to say it, but I’ve been a couch potato all day enjoying my music library favorites.

 

 

[romaz]

Perhaps it's not enough to make any conclusion. But the most interesting part is that mR's network latency is much higher than sMS-200's. And the differences of min and max of mR is way too large. On the other side, the sMS-200 did have very consistent and small latency.

I appreciate this post! As you've stated, this probably doesn't fully explain the differences between the mR and sMS-200 but I believe latency is an important parameter and one worth trying to optimize with any music server. You mentioned that the mR's network performance is limited to 480Mb/s max because of internal architecture and John Swenson has confirmed this in his previous posts but I don't think this is the issue as I understand that SOtM went out of their way to purposely limit the sMS-200's network performance to 100Mb/s. In fact, their earliest firmware could only communicate with 100Mb/s switches.

 

I view low latency in the same way as I do low impedance as ultimately, they impact the responsiveness of a component and a component's ability to resolve the micro details and the micro dynamics within a passage such as when steel brushes are used on a drum or when ride cymbals are played low. When you can hear the resonances from a clarinet's reed (an instrument that I play) and can feel the texture from a cello's strings as it is bowed or can distinguish the subtle differences in reverberation between 2 violins, then the engagement factor for me goes way up. In my view, it is the ability to resolve the infinitesimal details that separates extraordinary digital from common every day digital. Low noise is important for the sake of clarity and lack of harshness but it is the responsiveness of a system that is responsible for manifesting the timing brought about by a good clock.

 

As I reflect on why the mR and sMS-200 are both so good, I think the folks at Roon have summed it up very well. In the past, Roon was known more for it's library skinning features than for ultimate SQ as it was considered a "heavy" music player compared to a "light" music player like Bug Head Emperor. Where Roon's focus was on a rich overall listening experience that included convenience, ease of use and a pretty interface, Bug Head had only a singular focus and that was ultimate SQ. The problem with Roon's approach is that skinning, sorting or searching through a library of thousands of albums and presenting them in a pretty GUI requires "heavy" and noisy hardware and software resources whereas the best SQ generally comes from "light" resources that are much less noisy and more agile and responsive. Here is how Roon has proposed to resolve this dilemma:

 

Roon is heavier than Other Player

 

Many Audiophile grade media players focus on being extremely lightweight and doing nothing but audio playback. This is a tradeoff--they get to offer a simpler approach to sound quality, but they can't offer a rich user experience.

At Roon Labs, we aren't willing to sacrifice user experience or functionality at the altar of sound quality--we are determined to provide both at once.

At the same time, this is a bit of an oxymoron. A fully featured media server doing everything that we want it to do will never be as lightweight as a standalone, single-zone player application. So instead of trying to do the impossible, we've addressed this problem by changing the rules a little bit.

Instead of trying to pack a lightweight media player, and heavyweight media server into a computer in your listening room, we provide the tools you need to put space between those pieces--because they really don't belong together anyways.

This is, we feel, the best of both worlds--

 

• Our lightweight playback components (Roon Bridge, Roon Ready devices, and networked endpoints) are much lighter than any media player software out there because all they do is copy audio from a network interface to an audio device. They don't even have to do work to decode your media files!
  
• Our media server can afford to do things that no lightweight player app could do without losing its lightweight status--like automatically fetching and updating metadata for your library in the background, and supporting audiophile-quality playback in the listening room while other members of your household import new content and play audio in other rooms simultaneously.
  

As for their recommendations:

 

1. Core and output on separate devices

2. Control and output on separate devices

3. Keep an eye on the signal path

4. Don't underspec the server

5. Use ethernet between core and output

 

While these recommendations are made with Roon in mind, I think they can make sense in other situations and in fact, it is the pioneering efforts of the AO/JPLAY dual-box crowd that deserves credit here more than Roon.

 

Where the mR and sMS-200 excel vs the typical output PCs used in typical dual box AO/JPLAY setups is in their much simpler form factors. As opposed to off-the-shelf motherboards, these endpoints use custom spec'd motherboards free of the unnecessary and noise creating ICs that are unavoidable with off-the-shelf motherboards. They use audiophile-grade clocks fed by ultra low noise regulators and they have very short signal paths leading to very low impedance. Of course, they also incorporate lightweight OS's that further contribute to low latency and are run off of a small microSD card with a minimal electrical footprint. Having built more than a dozen music servers over the years, I have yet to build anything that has come close to sounding as good as these two endpoints and for so little money.

 

Regarding recommendation #5 (Use ethernet between core and output), I believe ethernet was viewed as an ideal connection method as opposed to USB between a noisy core PC and a quiet output PC due to its inherent galvanic isolation and error-correcting protocol but as those of us who feel we have benefited from this "direct connection" have realized, ethernet is not such a benign connection. Mark Jenkins of Antipodes Audio who is responsible for the very highly regarded DX Gen 2 ($6.5k) believes streaming via ethernet (ie NAS) definitely has its problems resulting in a compromise in resolution, clarity and immediacy compared to local playback in his machine while the enterprising folks at Sound Galleries believe their $16k SGM 2015 sounds better without an NAA.

 

What I do believe this "direct connection" has brought besides increased resolution, clarity and immediacy that Mark Jenkins talks about is transparency to upstream components which is both a good thing and a bad thing. For those of us who apply proper care and attention to our upstream components, we are rewarded for our efforts and for those of us who don't, there are sonic penalties.

 

As I embark on building the best music server I know how to as a feeder for either of these fine endpoints, if there is one observation I have made with all of my testing, it is the incontrovertible truth that it is always better to prevent a problem or to address it at the source than to have to fix it downstream because by then, irreversible damage has been done. If this is indeed true and I am successful with this build, will there still be a need for either the mR or sMS-200? I guess we'll see but my feeling is that any music server that is required to do heavy lifting can probably never be as quiet or as low impedance as either the mR or sMS-200.

 

As for the components I have thus far chosen for my build, it seems my decisions have come down to a balance between low noise and low latency. Sometimes it possible to have both but more often than not, it's trying to find the lesser of the evils and this isn't always intuitively obvious. Addressing latency, for example, is a matter of knowing where the bottlenecks are and addressing those so that other components in the chain aren't having to wait longer than necessary which can lead to not just timing errors but missed hand offs and even dropped bits. Of course, low latency at all costs can result in such noise that it isn't always worth it.

 

This is what I have come up with so far. For those contemplating their own build, hopefully, this can serve as another perspective and perhaps you would be willing to share your own. Nothing is completely set in stone and so I would appreciate the opinions of others regarding alternative options that are based on personal experience and insight. For those that believe bits are bits, this is the wrong thread for you because my experience has shown me that the transfer of digital very much has analog qualities. I prefer to be guided by science and common sense but like with most things regarding digital these days, I have found an open minded "empirical" approach to be more effective since there seems to still be much with digital that is not well understood.

 

Case - Streamcom FC8 Alpha Fanless ($180)

For NUCs, I have a bias for Akasa and for anything larger, I like the Streacoms both for aesthetic and functional reasons. The FC8 Alpha accommodates mini-ITX motherboards perfectly and are easy to work in. They use Streacom's thicker 4mm 6063 aluminum alloy which I have found to be fairly resistant to outside resonances and effective in blocking airborne RF.

 

Motherboard - Gigabyte GA-H110TN-GSM Plus ($110)

The smaller and the simpler, the better. The thin mini-ITX form factors can be powered directly from a 12V source via a convenient 2.5x5.5mm barrel connector which allows me to use my Paul Hynes SR7 which is a must. I like ASUS also but this Gigabyte model uses 2x the copper (2oz) in their PCB compared with other motherboards which was designed for overclockers as a means of better heat dissipation but also has the added benefit of lowering impedance by half. This model happens to have dual integrated Intel gigabyte NICs (Intel I219V and I211) which are not server grade. As this motherboard will be sent to SOtM to have its key clocks replaced, I noticed 5 clocks including a 24MHz system clock and dual 25MHz clocks for each LAN port and so I figured I would replace these 3 clocks. As I approached SOtM about replacing the linear regulators that feed each Intel LAN port, here is what Lee (one of SOtM's lead engineers) had to say:

 

"It's very difficult to replace switching regulator in mother board with linear regulator because mother board is very complex. Almost mother board use 6 or 8 layer PCB, it means maybe we can't see power signal lines. And maybe power signal lines share with other device...It makes difficult to replace."

 

Bummer.

 

LAN Adapter Card - Intel E1G42ETBLK Dual Port PCI-Express Server LAN adapter ($145)

So my solution to Lee's statement was to bypass the dual Intel LAN ports built into the Gigabyte motherboard with this card. They incorporate Intel's slightly older but highly regarded server-grade LAN chipset, the E1G42ET. In contrast to their newer I350-T2 chipset, this one consumes less power (2.8w vs >4w) but is also devoid of Intel's integrated power saving feature. Because it is PCIE, it has a lower latency pathway to the CPU and each LAN port has it's own independent path. Moreover, instead of 2 separate clocks for each LAN port on the Gigabyte motherboard, this card uses only a single clock which leaves me with 2 free clocks. To my delight, Lee has told me he can also replace the switching regulators on this card with SOtM's ultra low noise voltage regulators and further add capacitors. This would be as close to an audiophile LAN card as I can think of and may possibly allow me to remove my reclocking switch from the direct path and reposition it between my router and music server.

 

CPU - Intel 7700T (7th generation Kaby Lake) ($330)

This is a quad-core, 8-threaded CPU based on Intel's 14nm lithography. It has a base frequency of 2.9GHz with turbo speeds up to 3.8GHz and a TDP of 35w. I could easily have gone for the 6th generation 6700T but the newer 7700T is essentially the same price. As I have zero plans to overlock the CPU and will not be using HQPlayer for oversampling, I could also easily get by with an i5 but chose the i7 for its larger 8MB cache. As opposed to older generation Intel CPUs where the cache was equally distributed among the cores (usually 2MB per core), their new SmartCache has the option of utilizing the entire cache for a single core which is important since Roon isn't a multi-threaded application (from what I understand). According to Roon, the size of the cache makes a difference which is why they recommend against Celerons and their measly caches. Since the CPU is often the bottleneck, a large cache can keep latencies to a minimum.

 

Of interest and speaking of latencies, I expect to do something counterintuitive with this CPU that I hope will result in better SQ and so this is why the largest cache possible may become important. As opposed to some who are interested in overclocking to improve HQPlayer's performance, I intend to underclock to well below 15w TDP. I used to own a CAD CAT which had a dual core i7 in it with a TDP of 35 watts that Scott Berry, CAD's creator, dethrottled to 800MHz. He even dethrottled his RAM down to this same speed and so the OS experience was quite slow. As I started playing around with CPU and RAM timings, for sure, I found a direct correlation between slower speeds and improved SQ. This has been one of those situations where taking a hit on latency resulted in lower noise and better overall SQ. Fortunately, straight PCM or native DSD playback isn't CPU intensive.

 

RAM - Either 4GB of Kingston HyperX DDR3L 1600MHz CL9 SODIMM at 1.35V or Kingston HyperX DDR4 2133MHz CL13 at 1.2V (about $40 for either). Roon uses fixed buffers to allow memory playback totaling 1-2 tracks and so it doesn't benefit from large amounts of RAM. I figured this out by disconnecting my ethernet cable from my Mac Mini and my Mac Pro while streaming from Tidal and I noticed that music continued to play for a total of about 2 tracks regardless of whether my server had 4GB or 64GB. Of interest, when I disconnect the ethernet cable between my server and sMS-200, memory playback continues for only 5 seconds and so its memory buffer is quite small. Anyway, regarding DDR4 vs DDR3, DDR4 is more power efficient at 1.2V (and can be dethrottled down to 1.05V) vs 1.35V for DDR3L. DDR4 has faster read times but it turns out DDR3L even with its slower clock speeds has faster write times and lower latency (not just lower reported CAS latencies but actual lower measured latencies in both light and heavy environments) and so I need to hear for myself which results in better SQ and what further dethrottling may lead to. The folks at Sound Galleries have reported that with their highly overclocked server, they are getting better SQ with DDR3L.

 

OS Drive - I am finding the OS drive to be more important than the music storage drive with respect to SQ but to not have SSDs at all in the system is very beneficial with regards to less harshness. As to what I'll actually end up using is up in the air but I will be testing a few options and I already know one thing, I won't be going with a PCIE SSD in NVMe mode. PCIE NVMe SSDs have extremely low latency compared against SATA SSD and spinning hard drives and indeed, in my comparisons, there is better perceived detail resolution with NVMe but it comes with a huge compromise that is no longer tolerable and that is high frequency noise. As I posted previously, with my previous Mac Mini setup where I ran the OS (El Capitan) off a 64GB PCIE SD card, there was this wonderful relaxed, calm and natural presentation. Despite hours of listening, I never fatigued. Unfortunately, it isn't possible to run Windows off an SD card on a Mac Mini which forced me to install a small 128GB PCIE SSD and run Windows off of that. While I reported that the presentation is more "in your face" and fatiguing, with W2012R2 + AO + Process Lasso, it resulted in a return of some of that calm and quiet that I enjoyed with the SD card with the further upside of better precision and imaging and so I found the overall improvements to be worth the compromise. Or so I thought.

 

With the addition of SOtM's superclock to my sMS-200, switch and dX-USB HD (USB to SPDIF converter), what I have noticed with more time is that these components (especially the switch) are now even MORE transparent to both the good and bad qualities of my upstream server. Clarity, detail, timing, dynamics and depth are all much improved but so is the etch and glare that this PCIE SSD brings with it, so much so that I removed my new reclocking switch from the signal path due to irritation and fatigue.

 

As I decided to move away from the PCIE SSD and installed my OS onto a SATA SSD which I have connected via my Pachanko UPOCC copper SATA Reference cable and powered by my LPS-1, the SQ is MUCHO improved! Based on @lmitche's recommendation, I also compared my Samsung 850EVO SSD against a Seagate 2.5in 500MB hard drive also connected via this SATA cable and powered by an LPS-1 and I have to agree with him, there is even less HF noise that is to my liking. Thank you, Larry!

 

But this was not my best overall SQ, however. I went back to El Capitan on the SD card yesterday and without question, right now, this is where the magic is at. That relaxed, natural and effortless presentation that made the Mac Mini so magical when I first heard it was back and gone is the HF noise that even my hard drive seemed to still have some of. Where OSX lacks in precision and imaging compared against my optimized Windows setup, SOtM's better clocks seem to more than make up for it, especially now that I am able to reintroduce my switch back into my chain without irritation. While I still believe optimized Windows is better than Mac OS, the secret is to be able to run the OS off a flash card of some sort.

 

My Lexar 64GB SDXC card consumes no more than 0.6 watts max and while they are definitely slower than even a spinning hard drive, the newer UHS-I drives have good enough random disk I/O capability where the OS experience with El Capitan isn't painful and since I run my Roon remote software off my Mac Pro or iPad, it hasn't mattered. Of interest, based on the experience of Raspberry Pi users who run their OS off a microSD card, it has been found that certain microSD cards have even better random I/O performance (specifically the Samsung EVO+ which run about $30) and so that is what I will be testing next.

 

As to how to run W2012R2 + AO off an SD card, here is what I have found and placed an order for. It is coming in from China and who knows if it will be good but it accepts SDXC cards and connects to the motherboard via SATA. It is supposed to be bootable and the motherboard is supposed to see the SD card as any other SATA drive without the need for drivers. Of interest, this card has a clock and if it sounds good, I will likely replace that clock as well. I will also have the option of powering it with my LPS-1 if necessary which is something that is not possible with my Mac Mini:

 

Sintech SD/SDHC/SDXC/MMC Flash Memory card to SATA Adapter as 2.5" SATA SSD | eBay (about $20)

 

If it is as good as hoped, I would have the option of using others for music storage as 1TB SD cards have already been announced by SanDisk. Presently, 512GB SDXC cards run as low as $200 and when used for music storage and playback, they impart that same relaxed quiet quality minus any harshness.

 

Sorry for the horribly long post but this is where I'm at at this time and I would appreciate any constructive input others may have on how to improve it.

[lmitche]

Hi romaz,

 

Great to hear you were able to duplicate the SQ benefits gained from replacing an SSD with a hard drive. I am sure your SD card experiments will prove to be even better and I look forward to reading about your results.

 

The rest of your build looks great. I would just say that it may make sense to keep the pcie x4 slot open rather then use it for a network card. My experience here with the Adnaco fiberUSB solution powered by the LPS-1 has been very successful. With the Adnaco, if you store your music locally, the network is only used for control signals, and you achieve a first level of GI without the need for the NAA.

 

I am anxious to try a second level of GI and reclocking with the ISO REGEN.

 

Without the need for network drives, more windows processes can be turned off, and the SQ increases again. I'm running Windows 10 with 21 processes including Roon and HQplayer.

 

Having only run I7-6700ks here, I am not sure that the lower TDP CPUs will crank dsd512 in HQplayer. DSD512 is a big step up, so I'd try to ensure you can handle it in your new build. I think the 65 watt parts like the i7-7700 are your best bet with a turbo clock speed at 4.2 ghz.

 

Also populating the two closest ram slots to the CPU of the 4 DDR4 slots on my motherboard sounds better, even though it is said to break the dual channel capability.

 

Anyway my 2 cents FWIW,

 

Larry

[austinpop]

Sorry for the horribly long post but this is where I'm at at this time and I would appreciate any constructive input others may have on how to improve it.

 

 

Good luck on this journey, Roy! I am eager to hear what you learn.

[kennyb123]

What I do believe this "direct connection" has brought besides increased resolution, clarity and immediacy that Mark Jenkins talks about is transparency to upstream components which is both a good thing and a bad thing. For those of us who apply proper care and attention to our upstream components, we are rewarded for our efforts and for those of us who don't, there are sonic penalties.

 

Exactly my experience. I reported the other night substituting a Mac Mini/MMK/JS-2 for a 2008 Mac Pro. Subsequent listening had me thinking this was among my most significant upgrades. That Mac Pro had been doing much harm - and holding everything else back.

 

But this was not my best overall SQ, however. I went back to El Capitan on the SD card yesterday and without question, right now, this is where the magic is at.

 

I had thought I installed Sierra on my SD card, but this morning there was a notification that encouraged upgrading to Sierra. Turns out I inadvertently installed El Capitan. :-)

 

Sorry for the horribly long post but this is where I'm at at this time and I would appreciate any constructive input others may have on how to improve it.

 

Apologize - are you kidding? It was an awesome post. Thanks so much for putting all that effort into it.

 

 

Sent from my iPad using Computer Audiophile

[romaz]

Hi romaz,

 

Great to hear you were able to duplicate the SQ benefits gained from replacing an SSD with a hard drive. I am sure your SD card experiments will prove to be even better and I look forward to reading about your results.

 

The rest of your build looks great. I would just say that it may make sense to keep the pcie x4 slot open rather then use it for a network card. My experience here with the Adnaco fiberUSB solution powered by the LPS-1 has been very successful. With the Adnaco, if you store your music locally, the network is only used for control signals, and you achieve a first level of GI without the need for the NAA.

 

I am anxious to try a second level of GI and reclocking with the ISO REGEN.

 

Without the need for network drives, more windows processes can be turned off, and the SQ increases again. I'm running Windows 10 with 21 processes including Roon and HQplayer.

 

Having only run I7-6700ks here, I am not sure that the lower TDP CPUs will crank dsd512 in HQplayer. DSD512 is a big step up, so I'd try to ensure you can handle it in your new build. I think the 65 watt parts like the i7-7700 are your best bet with a turbo clock speed at 4.2 ghz.

 

Also populating the two closest ram slots to the CPU of the 4 DDR4 slots on my motherboard sounds better, even though it is said to break the dual channel capability.

 

Anyway my 2 cents FWIW,

 

Larry

Thanks, Larry. Very helpful and I appreciate the tip on the Adnaco USB unit. I am not familiar with this unit and I wonder how it might compare to the Acousense AFI USB modules from Germany which I haven't tried either. The Adnaco looks simpler and to be able to power it with an LPS-1 would make it a better choice probably. To be honest, I am skeptical that the optical isolation portion would benefit me since my DAC's USB input is already galvanically isolated but you never know what complete optical isolation might provide and it would appear the Adnaco would have 2 clocks that could be replaced for the better and this dual layer of reclocking could allow for a single box solution instead of server + NAA. It would help if you had experience with an mR or sMS-200 and could say that this Adnaco solution was better but as it stands, I am bowled over by my new sMS-200 Ultra. If I can secure a Adnaco unit to try, I will definitely do so.

 

Regarding the 6700K (or 7600K), I don't use HQP and have no interest in overclocking and so I am content to be able to use a low power CPU.

 

Thanks for sharing that you like how your DDR4 sounds. I have no experience with DDR4.

 

I prefer not to use network drives either as in my system, local playback for sure sounds a bit better although I maintain a NAS for comparison testing and for backup but I also stream from Tidal and so I can't shut down my network completely. As I plan to upgrade the "incoming" LAN port's clock on my new server and with a Paul Hynes SR7 powering my modem/router, I am hoping to bring Tidal streaming to nearly the same level of SQ as local playback. As it stands, network streaming doesn't sound horrible at all. If I am successful in elevating network playback close to the SQ level of my current local playback, it may encourage me to pursue building an audiophile class modem/router with better clocks and regulators and attempt to integrate larger capacity music storage into the router itself. I find the idea of an audiophile class modem/router/NAS that I can power with a single lead from my SR7 to be an attractive scenario just so I can provide a high quality stream to other areas of my home.

Anyway, interesting times ahead. Thanks again.

 

Roy