Audiophiles looking for a music server can use a home computer, a music server made by a traditional audiophile component manufacturer, or do it themselves. Typically, the DIY approach is for the very computer savvy, and some custom-designed servers are very good indeed. They can include specially made or modified audiophile parts -- USB cards, clock modules, solid-state drives, SATA cables -- and can even be on the technological cutting edge.
But if you lack the knowledge, time, or inclination to build a custom server, there’s a fourth way: go to a company that will design and build one for you. One such company is England’s Hifidelit, whose server I used to test the software that is the subject of this review.
Unless an audiophile chooses a server from a traditional component manufacturer, he or she must also select a computer operating system (OS). It’s commonly accepted that the OS can affect sound quality. However, as with many things audio, there’s no agreement on which OS sounds best.
Enter Phil Hobi, owner of Switzerland’s Highend-AudioPC, who usually goes by the moniker AudioPhil. A senior systems engineer on the server-engineering team of a prestigious financial institution, AudioPhil holds over 30 technical certifications from companies such as Microsoft, VMware, and Citrix. Phil -- who, not surprisingly, went the DIY route for his own music server -- argues that because “client-side” operating systems, such as Microsoft’s Windows 8.1 and Apple’s OS X Yosemite, are designed to be all things to all people and run an enormous variety of programs and applications, they can’t provide the best possible sound.
According to Phil, Windows Server 2012 R2, or WS 2012 R2 (approximately $285-$750 USD, depending on edition), which is primarily used with corporate enterprise servers, offers the best sound for computer audio because it’s more robust, and less bloated with unnecessary, jitter-causing features than are client-side OSes. Phil notes that, for applications in which maximum performance is demanded and microseconds matter, WS 2012 R2 can operate in something called Core mode, known in enterprise parlance as Server Core. This mode permits the removal of the OS’s graphical user interface (GUI), including the desktop, which is not needed for sound reproduction.
However, Phil wasn’t content to merely download WS 2012 R2 to his server and call it a day. Instead, he designed software, which he later named AudiophileOptimizer (€119, or about $133 at press time), which works with WS 2012 R2 and, he claims, further improves its sound quality, for the reasons discussed below.
One thing is certain: mission-critical applications such as Microsoft’s Azure cloud service, which must reliably host millions of the virtual servers used by major corporations to reduce the number of physical servers those companies need, don’t run the same OS you have on your home PC. To deliver the best performance possible, they use WS 2012 R2 in Core mode. If we’re looking for the best OS for computer audio, perhaps, as Phil suggests, we should avoid systems that are designed to do everything and anything, for everyone: send e-mail and instant messages, watch dancing-bear videos on YouTube, create documents, and order dinner from the local sushi bar.
Features: less is more
When running a client-side OS, a computer’s central processing unit (CPU) simultaneously runs approximately 100 services and 50 to 100 processes. In simplistic terms, a process is an abstraction of a program in execution. A service is a process that runs in the background without user intervention. While these definitions are mostly unilluminating to all but computer geeks, the bottom line is that running lots of processes and services taxes CPU resources. This, according to Phil, causes jitter, and thus poor audio performance. Phil states that, due to its robustness and lack of unnecessary jitter-inducing features, WS 2012 R2 is excellent for the reproduction of audio files.
However, as good as WS 2012 R2 is alleged to be, Phil had an idea for making it even better. He began designing software to further reduce the number of services and processes WS 2012 R2 had to run, thus theoretically creating less jitter and resulting in even better sound. As envisioned by Phil, his software, when completed, would accomplish this by, among other things, turning off the OS’s visual effects, taskbar, and icons, and modifying its registry. It would also optimize the server’s hard drive, and adjust its file, memory, CPU, and power settings.
There was just one problem. Although WS 2012 R2 offers the choice of a number of operating modes, Phil was convinced that Core mode would have, by far, the best sound quality. But that mode has no GUI, and it’s so stripped down that it contains no audio drivers. It can’t play sound files.
Phil got around the problem by reinstalling the original kernel-streaming drivers into WS 2012 R2, and modifying its registry to permit those drivers to work in Core mode, which allows the OS to play music in that mode. Though kernel streaming employs older technology, it’s still supported by many digital-to-analog converters, and is the preferred driver for many computer-audio enthusiasts. For those needing the more modern WASAPI drivers, another WS 2012 R2 mode, Minimal Server, supports these drivers and provides nearly the performance of Core mode. ASIO drivers work in several WS 2012 R2 modes, including Core.
Phil’s software progressed to where it could eventually disable some 150 unnecessary features of WS 2012 R2, leaving only 5 to 7 services and 15 to 20 processes running at any given time. When Phil downloaded the software to his music server and did some listening, he couldn’t believe his ears, he said. Virtually every aspect of sound had dramatically improved, and his system was free of digital harshness.
As he’d suspected, Phil got the best sound in WS 2012 R2’s Core mode. However, despite Core’s minimalism, Phil found that he gave up little in convenience. His software allowed him to manage his music library through a media player's or streaming service’s GUI, and thus access player settings, view album art, change player skins, etc. However, WS 2012 R2’s lack of a GUI meant that he had to use a few shell (formerly called DOS) commands when otherwise operating the server. This wasn’t a problem for Phil, and it shouldn’t be difficult for the average audiophile to master.
Phil offered his software to a few computer-audio enthusiasts through an Internet forum. After word of its sound quality spread, he was flooded with e-mails from around the world asking where it could be purchased. So he offered a beta version of the software, and in October 2013 released the first finished version, which he called AudiophileOptimizer. So far, he’s sold more than 1000 copies of AO in more than 60 countries.
Like WS 2012 R2, AO can be run in one of several modes. Express mode results in a somewhat limited optimization of the OS, installs without any user input, and offers the least customization. For example, AO includes Sound Signature and Digital Filter adjustments (see below). For these, Express mode automatically selects settings that Phil has found are preferred by 80% of AO users. Express mode is said to be an improvement over the stock sound, but is still far from the best AO can offer.
Advanced mode, recommended for advanced users, offers the ability to customize some of AO’s features and settings, including Sound Signatures and Digital Filters. Advanced mode allows for the deactivation of more Windows features than does Express, and, Phil claims, offers better sound quality.
Ultimate mode offers the greatest degree of customization, and lets the user deactivate even more Windows features. It is said to provide the best sound, and it is for only very advanced users who have considerable knowledge of Windows systems and AO.
AO offers four Sound Signatures. Signature 1 is the default setting. There is also Signature X, which is the Windows default. There are four Digital Filter settings, A through D. Filter B is the default setting, and Filter X is the Windows default.
Because AO operates outside the signal chain, changing any of these settings doesn’t alter the source signal. Rather, the settings affect only the way the OS prioritizes the running processes and threads in light of the CPU’s operational state. Thus, according to Phil, the results of any of these adjustments are “bit-perfect.”
WS 2012 R2 comes in a number of versions, including the relatively inexpensive WS 2012 R2 Essentials ($285-$389, depending on edition). Phil states that AO works with all versions of WS 2012 R2, as well as the older WS 2012. For purposes of audio quality, however, Phil says that there’s no need to buy one of the more expensive versions. However, AO won’t work with any OS other than WS 2012 and WS 2012 R2.
Phil also says that AO and WS 2012 R2 both work extremely well with a variety of media players and streaming services, and for those that don’t work out of the box, there’s always a workaround.
Setup, part 1: basics
The first step in setup is to select a computer. After some back-and-forth with Phil, we selected a music server (£1119, or about $1250 at press time without an OS) from Hifidelit, a British maker of custom servers. To power the server, Phil asked Vinnie Rossi, of Red Wine Audio, to send me one of his Black Lightning battery power supplies ($1295 as configured).
I left it to Phil to select the server’s internal parts: HDPlex fanless case, Intel Quad Core i7 processor, 8GB memory, Intel DH87RL Media Series motherboard, and SanDisk solid-state drive. The list of possible hardware upgrades for the Hifidelit is virtually endless. Nonetheless, the server was certainly capable of producing very good sound, and demonstrating the effects, if any, of AudiophileOptimizer.
Note that for ultimate performance, many who build or buy a custom server end up with a two-box configuration, exclusive of power supply. You can keep things somewhat simple by running only one box in Core mode, and leaving the GUI intact on the other box. However, any two-box system requires some sort of networking so that the boxes can communicate with each other, which complicates things. Via e-mail, Phil recommended that I “START THE HELL WITH A SINGLE BOX.” So that’s what I did.
Once the hardware is ready to go, you connect to the server a mouse, keyboard, and monitor. You also connect the server to the Internet, wirelessly or via Ethernet.
You’re now ready to make some preliminary Basic Input/Output System (BIOS) adjustments. Doing so will disable a number of unnecessary features and devices: energy-saving modes, serial ports, Intel’s Hyper Threading (which allows the CPU to efficiently handle multiple tasks), and the onboard soundcard. Don’t worry -- this step is optional, and only for those who are comfortable making BIOS adjustments.
After that, you install WS 2012 R2.
You then install the drivers for your DAC (in my case, an Esoteric K-01), configure the media player (in my case, JRiver Media Center), and create a library within the media player for an external drive containing digital music files (in my case, a Western Digital 4TB drive).
Now, from Highend-AudioPC’s website, you can download AudiophileOptimizer to your server.
The next step is to download a remote-control app, so that you can wirelessly access the server from a tablet or mobile phone. I used JRemote, an iPad app from JRiver ($9.99 from the iTunes App Store).
The final step is to run AO. If you don’t choose Express mode, AO asks you questions concerning the features to be deactivated, the Sound Signatures, the Digital Filters, and other matters.
Setup, part 2: going headless
From here, many users will choose to go headless, ditching the mouse, keyboard, and monitor. These are needed only to make changes in the Windows or AO settings, and are not used to manage music playback. Or you can sidestep decapitation, and use the monitor to display track information, album art, visualizations, etc., most of which will also appear on your phone or tablet.
Phil made it easy for me. He partitioned the server’s hard drive and, in addition to WS 2012 R2, installed Windows 8.1. As a result, I could compare not only the server’s sound with and without AO, but also any differences in the sounds of the two OSes. He also preinstalled AO and JRiver. Still, I insisted that I set up AO on my own, so that I’d know what the typical user is in for. Although several times I lost my way for a bit, I was able to install the software with very little help from Phil.
Each of these setup steps is laid out in AO’s detailed and well-written manual, which can be downloaded from Highend-AudioPC’s website (where you’ll also find helpful videos). If you can install a Windows OS, copy folders, select drives, and run programs, and can become familiar with a handful of shell commands, you’ll have no problem setting up AO. Help is also available on a number of computer-audio Internet forums, and Highend-AudioPC offers premium e-mail support.
When I was ready to fire things up, I connected the server to my Esoteric K-01 SACD/CD player-DAC with a Synergistic Research Galileo LE USB cable. I placed the server on three Synergistic Research MIG footers.
As for the Sound Signature and Digital Filter adjustments, I’ve been in this game long enough not to predict which will sound best to anyone but me. Over time, my preferences have changed; when it comes to digital filtering and upsampling, I generally prefer the accuracy, honesty, and clean leading edges typically provided by the unadulterated signal.
The only thing that seemed indisputable about AO’s digital settings was that Filter A provided a fast, transparent, detailed, solid-state sort of sound. Filter D was more tube-like, providing more body, and was harmonically and texturally superior to Filter A. Of course, the “best” settings will be determined based on user preference, associated equipment, and recording played.
Performance: fully optimized
Regardless of mode, WS 2012 R2 sounded better than Windows 8.1. On any song from Peter Gabriel’s Scratch My Back (16-bit/44.1kHz AIFF, Real World Productions), the string instruments of the London Scratch Orchestra, such as double bass, cello, viola and violin, sounded fuller, with better leading-edge transients, than with Windows 8.1. In “The Power of the Heart,” Jason Rebello’s piano notes were denser and more distinct than with Windows 8.1. These improvements were similar in magnitude to what you get when you upgrade a power cord. I didn’t want to go back to Windows 8.1, but I could if I had to.
Keeping WS 2012 R2 in Core mode, and now activating AudiophileOptimizer in its Ultimate mode, brought much more dramatic improvements. The quantitative differences from the stock OS were like replacing several power cords and upgrading my server or throwing in a good power conditioner. From here, there was no going back -- not for $133, and not for $1330.
AO significantly reduced harshness and glare. It also instilled in the music a calm and an ease that were particularly welcome with singers. In “Let’s Do It,” from Ella Fitzgerald Sings the Cole Porter Songbook (16/44.1 AIFF, Verve), Fitzgerald’s voice was clearer and silkier than with the stock OS. At the same time, the sound of her mouth closing at the ends of phrases was easier to discern. In fact, voices of all types sounded richer, more solid with AO activated.
To anthropomorphize: Good analog components -- and, today, the best digital ones -- reproduce so much harmonic texture and richness that they permit instruments to breathe. I’m tempted to agree with a statement on Highend-AudioPC’s website: that AO makes “analog-like” sound. But even with AO fully implemented, the Hifidelit server didn’t allow instruments to breathe the way that very good tubed components do. As discussed below, this might have to do with the modest nature of the server. There’s no doubt that AO revealed considerable amounts of texture and richness, and improved performance in other ways typically associated with analog playback. For example, it significantly reduced glare, and created a much more fluid and organic sound.
Another change wrought by AO was that everything became much tighter, more focused, more weighty. In fact, it made the unaltered WS 2012 R2 sound downright wimpy. For example, in “In My Place,” from Coldplay’s A Rush of Blood to the Head (16/44.1 AIFF, Parlophone), drums had more authority and transient cut. But this was no parlor trick -- the faster leading edges weren’t imbued with displeasingly brittle sharpness.
AO also significantly lowered the noise floor, exposing or better portraying sounds that were absent or muted with the stock OS. I noticed this not only in Ella Fitzgerald’s voice, but in a wide variety of music, including Yes’s Tales from Topographic Oceans (24/192 FLAC, Rhino Atlantic). In “The Ancient: Giants Under the Sun,” the articulation of cymbals was clearer and more pronounced, with a less glassy and more tonally accurate sheen, and boasted a more detailed and extended decay curve than with the stock OS.
Also notable was that, with AO, the soundstage was now deeper and wider than with the stock OS. Soundstage size is crucial to “Last Train Home,” from the Pat Metheny Group’s Still Life (Talking) (16/44.1 AIFF, Geffen), in which the sitar-like sound of Metheny’s guitar synth effortlessly floats over a vast, open soundscape while Paul Wertico’s wire brushes on snare drum re-create the sound of a steam locomotive, and Lyle Mays’s keyboards contribute additional train sounds. The larger soundstage and greater focus wrought by AO kicked up the dramatic effect of “Last Train Home” at least one notch, and maybe two. More generally, AO permitted the Hifidelit server to less obviously restrict the sizes of images created by my Esoteric components, themselves known for creating soundstages of generous proportions.
As much as AO transformed the sound of my system, it didn’t deliver better sound from the Hifidelit server than I hear when spinning CDs in my $20,000 Esoteric K-01 transport-DAC. Even with “Red Book” discs, the K-01 produced warm yet transparent and holographic sound that was nothing short of intoxicating. However, the comparison with the K-01 isn’t exactly fair. At £1119/$1250, the Hifidelit is extremely modest in cost, and contains none of the high-performance audiophile parts that can lift a DIY or custom-designed server to the state of the art.
Moreover, there is no dispute that AO delivered the proverbial night-and-day difference when I compared it to the stock version of WS 2012 R2, itself an improvement over Windows 8.1. Without a doubt, AO caused the humble Hifidelit server to sound as good as or better than a number of much more expensive servers I’ve had in-house.
It would be interesting to know what the results would have been, and how much more the music might have been able to breathe, had I chosen one of Hifidelit’s more tricked-out servers, or even experimented with a two-box server. My guess is that, when used with top-shelf hardware, AO might well serve as a key part of a state-of-the-art digital front end. It’s that special.
In a world in which $133 buys you just a handful of hi-rez downloads, Highend-AudioPC’s AudiophileOptimizer is a stone-cold bargain -- and that would still be the case even if AudioPhil charged a lot more for it. If you’re an audiophile looking for a music server, you’ll want to strongly consider one that can run WS 2012 R2, and thus AO. If you want the level of sound quality AO can deliver but go any other route, be prepared to spend a pretty penny. And if you’re a manufacturer of music servers that use Linux, or any other OS that’s not WS 2012 R2 or WS 2012, take notice: Your competition has upped its game.
. . . Howard Kneller
- Amplifier -- Esoteric A-03
- Preamplifier -- Esoteric C-02
- Sources/DAC -- Esoteric K-01 SACD/CD player-DAC, Hifidelit music server, Windows 7 laptop running JRiver Media Center 17
- Speakers -- YG Audio Kipod II Signature, JL Audio E-Sub e112 subwoofers (2)
- Interconnects -- Synergistic Research Element Copper-Tungsten-Silver (components, bass modules of active speakers)
- Digital cables -- Synergistic Research Galileo LE USB, JPlay JCAT USB
- Speaker cables -- Synergistic Research Element Copper-Tungsten-Silver (tweeter) and Element Copper-Tungsten (midrange)
- Power cords -- Synergistic Research Element Copper-Tungsten-Silver Analog (amplifier, preamplifier) and Copper-Tungsten-Silver Digital (disc player, DAC), Tesla Precision AC SE (speakers), Element Copper-Tungsten (Powercell power conditioners), Element Copper-Tungsten-Silver Analog and Digital (Enigma power supply fed by two power cords), Tesla Hologram A (QLS Lines strips with Galileo MPCs)
- Power conditioners and distribution -- Synergistic Research Powercell 6 SE (digital only) daisy-chained to Powercell 10SE Mk.II
- Isolation devices -- Symposium Acoustics Osiris Racks and RollerBlock Series 2+ Equipment Support System, Synergistic Research Tranquility Bases and MIG footers, Custom Isolation Products amp stand, Silent Running Audio VR fp Isobase
- Room treatments and correction -- Synergistic Research Acoustic Art System, HFT and FEQ room-treatment devices, XOT Crossover Transducer; DSPeaker Antimode 8033 Subwoofer Equalizers with Channel Island Audio linear power supplies
- Misc. -- Black Discus Audio System Enhancer, Synergistic Research Galileo Universal interconnect and speaker-cable cells
Highend-AudioPC AudiophileOptimizer Computer Audio Software
Price: $133 USD.
Warranty: 14-day money-back guarantee.