Music in the Round #79
Korean manufacturer SOtM, Inc. describes it on their website as a "music server based on Windows Server OS besides the original Linux [Vortexbox] OS based sMS-1000SQ." I'd describe it as a Windows-based PC that's designed and optimized to manage a database of music files and stream the music to local or networked DACs, and that supports multiple options for file management, playback, and target devices. (Hmmm: that's not much better, is it?)
With its stylish, compact chassis (14" wide by 2.7" high by 9.4" deep, weighing 8.8 lbs), the sMS-1000SQ WE looks more like an audio component than a computer. But this server is actually a PC with an Intel Atom CPU N2800 (1.86GHz) processor, 4GB RAM, and 32GB of SSD storage, running Windows Server 2012 R2 Essentials (64-bit). It comes in three different versions: with an analog output, with an S/PDIF digital output, or with a USB digital output, the port of which is based on SOtM's tX-USBexp output card; SOtM sent me the USB version ($3500), to which they had added their Ultra Low Noise Jitter Clock for USB outputs ($500). In addition, SOtM sent me their sPS-1000 DC power supply ($1000); I used two of its three independent outputs for the sMS-1000SQ and the tX-USBexp, the latter of which normally works off the server's internal power supply. To complete the suite, SOtM sent an iSO-CAT6 LAN isolation transformer ($350), which I inserted in the Ethernet link to my home network.
Clearly, with its limited 32GB storage for OS, programs, and data, the sMS-1000SQ WE is intended to be used with music files stored on attached or remote drives. (I found that it worked easily with both.) Similarly, because the sMS-1000SQ WE runs under Windows Server, it has limited display capabilities: it's designed to be controlled, via network, by a tablet, a smartphone, or another computer. Indeed, after installation and setup, there's no need for any physical interaction with the SOtM server. This is underscored by the fact that SOtM installs on the sMS-1000SQ WE a software package called AudiophileOptimizer, by the Swiss company Highend-AudioPC, which works with Windows Server to optimize audio quality. AudiophileOptimizer further limits user interaction with the SOtM server by disabling all functions not required forand that might interfere with or compromisethe management and playback of music. As we'll see, this is an important feature of the sMS-1000SQ WE.
Finally, the sMS-1000SQ WE comes loaded with apps, including Foobar2000, JRiver Media Center, Qobuz, Roon, and Tidal. The default is Roon, which as yet doesn't do multichannelbut Foobar2000 and JRiver do, and that, my friends, is why we are here.
Physical setup was trivial. DC power from the sPS-1000 supply was via provided cables, network connection was via the iSO-CAT6 to my local Ethernet switch, and USB output was from the tX-USBexp output to my exaSound e28 DAC, via an AudioQuest Carbon USB link. Just to get going, I plugged a 1TB USB drive full of files into one of the other USB connectors on the sMS-1000SQ WE. After power-up, the SOtM server and power supply were dead silent but accessible, via a remote-desktop app, from my iPad or my main PC. Roon being already installed, I downloaded the RoonRemote app and pointed it to my USB drive; it quickly added the files on the USB drive to the SOtM server's library, ready for play.
However, the exaSound DAC, like so many high-performance devices, requires the installation of its proprietary ASIO driver, and that installation is more complicated for the sMS-1000SQ WE: in its zeal to disable all functions not strictly necessary for music playback, AudiophileOptimizer (AO) disables the ability to install new applications. I had to use a remote-desktop app to disable AO, reboot the sMS-1000SQ WE, install the ASIO driver, and reboot to re-enable AO. This was a bit of a pain, as I needed to download and print the instructions for it (sMS-1000SQ Windows Edition Quick Guide); these instructions are not included in the manual, which is printed in painfully tiny type. AO provides additional documentation, but it is almost encyclopedic in scope. Despite my kvetching, the entire installation, physical and logical, took me about two hours, and then I was listening to stereo, including DSD, with Roon. Very nice sound in every way, but around here . . . mere stereo? So what?
The promise of the sMS-1000SQ WE is that it makes almost anything possible because it's a network-connected Windows PC, as comfortable with installed programs such as Foobar2000 and JRiverwhich support multichannelas with Roon. That characteristic also means that it can be used to send as well as render music via DLNA. Using JRiver, I pointed the SOtM to the attached USB drive and my NAS, the latter containing some 6TB worth of files. It took JRiver an hour or so to process all the files, but then it was old home week, running the sMS-1000SQ WE from JRemote or via remote desktop. All my files, PCM and DSD, stereo and multichannel, even DXD and ISO, were available because JRiver can play them. It ain't the hardware that limits servers and renderers to stereo, it's the software . . . or is it?
I began with Willie Nelson's Night and Day (DVD-A, Surrounded-By SBE-1001-9, out of print) because it's assertively immersive, and I love it. Sure enough, this 24-bit/96kHz, 5.1-channel FLAC file sounded great, each instrument in its place, depicted with clarity, presence, and weight, and yours truly sitting at the center of it all. From there I climbed the resolution scale, first to 24/192 and then to 24/352.8: Beethoven's Piano Concertos 1 and 2, with pianist Hannes Minnaar, and Jan Willem de Vriend conducting the Netherlands Symphony Orchestra (SACD/CD, Challenge Classics CC72712D), was presented with a delightful balance between piano and orchestra in a warm acoustic. Playback of all high-resolution PCM multichannel files by the sMS-1000SQ WE was unperturbed, uninterrupted, and excellent.
I switched to DSF files and was just as impressed. The delightful Bailes, Tonadas & Cachuasmusic from the 18th-century Codex Trujillo del Peru, with Adri†n Rodriguez Van der Spoel directing M£sica Temprana (Cobra 0036)is infectious and spiced with piquant percussion, which the sMS-1000SQ WE and the exaSound reproduced with aplomb. Pushing the SOtM even further, I tried a few 5.0-channel DSD256 files obtained from some friendly recording engineers under promise of confidentiality, and boy, were they lovely! Clearly, the sMS-1000SQ WE, partnered with a capable DAC, could play almost anything and realize the complete potential of the recording.
As I listened to these tracks I often browsed my files, looking for what to sample next. When I did this in JRiver, it was usually no problem. However, browsing with Windows Explorer produced an occasional stutter in playback that was correlated with my file scrolling. This suggested that the Atom processor was running near the limit of its processing ability. The Windows Resource Monitor revealed that the CPU was running at about 20% with nothing playing. When playing hi-def multichannel files, PCM or DSD, it jumped to over 50%, and responded to file-scrolling by peaking at close to 100%. I know I wasn't supposed to be fiddling around while the music was playing, but this demanded investigation.
First, I turned off AudiophileOptimizer. As expected, the CPU load increased and the stability of playback worsened, so I turned it back on. I then tried several other maneuvers to see what else might be a problem. The 5.0 DSD256 tracks played beautifully as DSD direct to the exaSound, but the sMS-1000SQ WE ran out of steam when asked to convert them, on the fly, to PCM. That happened with all multichannel DSD files, nor would the SOtM play ISO files: they just buffered . . . buffered . . . buffered. Interminably (footnote 1). I didn't even think about installing Dirac Live EQ, which uses more resources and requires PCM conversion.
Bottom line: The Atom CPU is the sMS-1000SQ WE's limiting factor, though AudiophileOptimizer helped in getting a lot out of it. That said, the SOtM sMS-1000SQ Windows Edition sounded just wonderful playing all music files, from 16/44.1 up to 24/352.8 PCM and DSD256, in their native formatswhich basically fulfills its mandate. It does stereo as well as any of the dedicated servers, and it also does multichannel. But it needs more processing horsepower.
What's Up with Multichannel Servers?
Playing and enjoying multichannel music is easy in theory, but there are some frustrating impediments. Most mainstream servers and audio components ignore multichannel. For example, a glance at Native DSD Music's NativeDSD Database (www.nativedsd.com/database) reveals that, of over 400 DSD-capable components, only 20 or so are multichannel devices, and of those, half are disc players; only three are DACs. (Add a handful more with the Mytek stereo DACs, which can be stacked for multichannel use.) I can only presume that a wider survey, of non-DSD products, would return an even smaller proportion of multichannel-capable devices, especially if one included mobile listening via headphones. In the audio market, multichannel is small potatoes.
At the 2015 Consumer Electronics Show, I asked a manufacturer (I won't name him; he's no guiltier than most others) whether the USB audio input on his new multichannel product would support the playback of multichannel files. He didn't know. One of his tech guys said that it wouldn't, but that he'd patched his own sample so that it would.
"Why isn't it in the product?"
"It wasn't in the design spec."
I'm sure that the exclusion from the design spec of multichannel via USB was based on market analysis that showed little evidence of consumer interest. On the other hand, this product would have been the first of its kind with such a featureeven today, it would still be the only such device made. Such thinking only perpetuates the difficulty of exposing users to multichannel music streaming and continues to ensure low market interest. Ditto all those servers and renderers listed in the NativeDSD Database.
Think about it. Excluding DACswhich, obviously, must have the requisite number of hardware channelsservers and renderers are data processors that access local or networked data files and, via software, translate the contents of those files into an audio format acceptable to the DAC(s). Simple mathematics tells us that six channels present a CPU with three times the work of the two channels of stereobut all sorts of general-purpose computers are up to the job, given enough CPU and RAM. At audio shows, one can't ignore the ubiquity of the laptops sitting atop component racks. The servers and renderers in the NativeDSD Database are among some of the most sophisticated and powerful audio-processing components; the inclusion of multichannel capability should require only changes in the software.
Of course, one can argue whether or not such a general-purpose computer is optimal; indeed, making something that works and sounds better than an ordinary device is the high-end audio industry's raison d'etre. But in a generic laptop or custom processor, it's the software that determines what it does and how well it does it. To the programmer, the number of channels is just one of myriad variables. It should be no surprise that there are probably more music-player programs that can play more than two channels than there are hardware-based streamers/renderers that can. My four-year-old, base-model Mac mini handles it all with aplomb. Why won't they?
Footnote 1: I found similar but more severe limitations with the DigiBit Aria music server (see my March 2015 column), which used the earlier Atom N2600 processor.