Media Server Reviews

The Innuos ZENith Next-Gen ($20,700 as equipped) does what streamer-servers do: store music files, read them into memory, and send them on to a D/A converter to make music. In Innuos's complex (yet logical) lineup of streamers and streamer-servers, the ZENith Next-Gen sits just below the flagship Statement and above the ZENith Mk.3 (footnote 1). The ZENith Mk.3 remains in the Innuos lineup for now but will be replaced at AXPONA shortly after this issue hits mailboxes and newsstands.

Though similar in many respects, with a very similar appearance, these two products—the ZENith Next-Gen and the ZENith Mk.3—are very different beasts. There is one rather obvious difference: a CD slot on the Mk.3 (with, of course, a CD drive inside), which makes it easy to rip CDs to the server's internal memory. This feature is absent from the more purist ZENith Next-Gen. But with the Next-Gen you can have your cake and eat it: Attach any USB CD ripper to one of the USB ports, and it will work just the same as the built-in ripper on the Mk.3. The other differences between the Mk.3 and the Next-Gen are less obvious, but those differences go much deeper; see the Details section, below.

If you want to, you can leave off the internal storage in the ZENith Next-Gen and use it as a pure streamer. In addition to the network connection intended to send data to a streaming DAC, there's another you can connect directly to a network switch, providing access to streaming services and music data stored on your local network.

All this is tied together by Innuos Sense. Currently on revision 3.2, this is Innuos's proprietary app, which manages music-data access as well as (via the associated player) music playback. Sense is a downloadable app that runs on iOS and Android tablets and phones; it's also built in to the ZENith so that you can use it with a web browser by typing in an IP address or my.innuos.com. During my audition, I used Sense on an M4 iPad Pro, but most often I accessed it, with perfect reliability, on my laptop.

Using the Sense app, you can stream music from Deezer, Idagio, Qobuz, and Tidal. Also built into Sense is access, without a subscription, to a large number of internet radio stations (which despite the severe lossy compression employed by most, I've come to enjoy and value). You can also use Tidal Connect to play music directly from the Tidal app; you'll soon be able to do the same thing with Qobuz Connect (footnote 2).

What's more, if you prefer Roon's information-rich environment to Innuos Sense, the ZENith Next-Gen can serve that up, too. To make this as plain as possible: You have your choice of Roon (with a subscription) and Innuos Sense. You can use either or both or a combination; more on this below.

Details
The following comes from Innuos founder and R&D Director Nuno Vittorino, with whom I've had past conversations about digital audio and sonics, at hi-fi shows, by internet, and in my New York City apartment. "As you know, we've been investigating for years factors within a digital source that affect various aspects of sound, such as PRaT, soundstage width and depth, as well as detail and realism of instruments and voices. There were three technical variables that we've been investigating in the past years: Processing latency, power supply impedance (AC stage) and resistance (DC stage)," and "further minimizing electrical noise beyond the power supplies." To reduce latency, the research and design team works to shorten processing times at various stages and to shorten the signal path; Innuos has found that such changes can improve sonics. Reducing PS impedance/resistance determines "how fast the power supply can provide current in a clean way, helping again with lowering processing latency." And then there's the issue of electronic noise—EMI—which, Innuos has found, is one of the chief contributors to sonic degradation prior to D/A conversion.

How does this translate into hardware and software? Vitorino was remarkably forthcoming. "Starting with the aim of reducing processing latency," Vitorino said, "we developed the following technologies:"

Real-Time Operating System. "We adapted our own SenseOS to use a Linux Real-time kernel (footnote 3), which allows much better control over process latency." The Linux real-time kernel was developed to allow Linux to handle time-critical operations. Linux has been around forever, but the real-time kernel was finished only in late 2024. At least one other server/streamer manufacturer is utilizing it.

Audiocore. "Using the new Real-Time kernel now allows us to control processes so that we can dedicate specific processor cores to specific audio tasks, from USB/Ethernet interrupt requests to audio decoding. This minimizes interruptions to the sound-processing chain. For a real-world analogy, think of crossing New York in a car. We'd ensure you'd get green lights at every intersection."

More Processor Cores. "In order to make the most out of this technology, we needed new processors with more cores, but we needed to retain low power consumption. ... [T]he ZENith Next-Gen uses eight real cores."

The Power Supply. For as long as I've known about the company, Innuos has focused on the quality of its power supplies. The power supply on the Next-Gen is totally different from the one on the ZENith Mk.3. It starts with a 300VA toroidal transformer. "We had started with the Statement Next-Gen to work with active rectification," Vitorino said, "which is a much more efficient and fast way to convert AC to DC. ... We used the same active rectification technology"—they call it ARC (footnote 4)—on the NEXith Next-Gen as the Statement Next-Gen. "This is all isolated in a dedicated, EMI-shielded enclosure to prevent both noise transmission and absorption.

"Given the requirements on the new processors, we developed a dual-rail regulator based on gallium nitride MOSFETs that we call NGaN." The photo of the ZENith's insides shows the separate NGaN module. "The advantage of this method is that the regulator works at a much faster switching frequency, which can be precisely adjusted without loss." Plus, the higher frequency makes filtering the resulting noise easier. "NGaN is extremely efficient in speed whilst also running very cool, with no heatsink requirement," Vitorino continued. When paired with the toroidal transformer, "this greatly reduces electrical impedance, resulting in more rapid power distribution, which helps to give a more precise sound. Win-win."

The Mainboard. "For the purpose of minimizing electrical noise, we focused on the mainboard. Developed together with our motherboard manufacturer, the PreciseAudio mainboard is a highly customized motherboard that goes well beyond anything we have done before." "PreciseAudio" is an Innuos trade name. "Not only did we remove unnecessary, EMI-prone components, ... we have customized the switching regulators with the manufacturer for further noise reduction. Additionally, we have for the first time been provided access to low-level control of the mainboard, so we can customize power aspects for the processor, the RAM, clocks, and many other areas. We can now tune every aspect of I/O, from memory to processors to the PCIe bus, in order to optimize it for lower latency, power noise, and EMI."

All that mainboard customization, in turn, allows the use of "more precise power supplies with faster CPUs that would normally require a noisier power source." The result is "even faster performance with an even lower noisefloor." Faster CPUs can be used while producing less heat. It's a virtuous circle. In conjunction with real-time Linux, which allows the unmodified operating system to respond promptly to time-sensitive commands, system latency is "dramatically" reduced, Vitorino said. "The ZENith NextGen is the first Innuos system that incorporates all these technologies in one system," he wrote in an email. In this respect, not even the Statement can match it.

Flexibility. "We decided to make it a more flexible product," Vitorino told me. You can now buy the ZENith Next-Gen with no internal storage, which makes it a pure streamer. Or "storage can be factory-fitted or user-fitted via a M.2 SSD, resulting in up to 16TB storage." That's two separate SSDs, one factory-installed, one accesible to the user. The factory-fitted SSD is connected directly to the mainboard, reducing latency and the EMI that can result from a wired connection. But the storage modules don't act like they're separate. "Our software team developed the Expandable Storage Management System (XSM) in order to add storage seamlessly as one single pool with storage 100% usable, yet ensuring that in the case one SSD fails, only what's contained on that SSD is affected and not the whole storage." The SSD that runs the OS has its own separate power supply and runs at very low power (a third of a watt), producing less noise and heat.

Actually there's a third SSD, the one used to store and run the operating system—the specialized Linux and the rest of the Innuos software. For this SSD, Innuos uses the SSD in a mode known as "pSLC SSD," short for "pseudo single-level cell," which means that only every second or third bit is used for storage—just one bit per storage cell. In a normal SSD, each cell holds up to four bits, depending on the SSD type. pSLC makes storage more expensive (though less expensive than "true" SLC, which has one bit per cell natively), faster, and less prone to failure. It's an approach used in cost-constrained mission-critical applications like data centers.

As previously mentioned, "the ZENith Next-Gen also includes replaceable Digital Output modules." Current choices are the PhoenixUSB Lite reclocker module—a modular version of the freestanding Innuos PhoenixUSB—a similar module for I2S, and a module combining S/PDIF on RCA, AES3, and TosLink. "Other modules are planned in the future, including dedicated output modules to specific DAC protocols."

A final difference between the Next-Gen and the Mk.3: The Next-Gen's chassis is more advanced than the one wrapping the Mk.3, with a "10mm-thick CNC-machined, bead-blasted and anodized" aluminum enclosure and a new vibration-dampened platform that "helps to isolate the toroidal transformer." On the EMI front, the Next-Gen has a chassis-ground connection so that you can ground the Next-Gen chassis as you would that of a phono preamp.

So, while the Mk.3 was the starting point for the Next-Gen, the Next-Gen is a much finer-tuned, more sophisticated machine. Mercedes fans might think of it as the AMG version of the Innuos ZENith.

Innuos Sense
Jason Victor Serinus's auditions of Innuos servers have documented well the development of the Innuos control app, starting in 2020 (footnote 5). In 2020, Jason found the software "frustrating," as he wrote in his review of the first version of the Innuos Statement. Some things just didn't work. Innuos was not alone in this; server makers were struggling with becoming software companies. Innuos started out as a server company, but it takes time to build a flexible, user-friendly music-management app.

I've been using Innuos Sense for close to two years (footnote 6), during which time the software has auto-updated many times; I'm using version 3.2.1. It isn't perfect—some needed features are missing, and usability is not quite optimized—but outright errors are rare. Consequently, I'm ready now to call Innuos Sense mature.

Sense (together with the hardware) accomplishes its core mission—keeping track of, serving up, and playing back music—very well. On my M4 iPad Pro in particular it is instantly responsive. The design is logical, elegant, and in some respects (on the iPad's gorgeous Retina display) visually stunning.

Setup
The setup part of the software is nothing special, but it's utilitarian and easy to use. It's logical. Once I added the ZENith to my system, connected it, and turned it on, the software found the ZENith right away. I selected it. DACs on the same network are detected automatically. There are only a few key settings, which largely depend on the D/A converter in use. You must tell the ZENith how your DAC handles DSD—no support, via DoP (DSD over PCM), or native—and whether to enable MQA software decoding or allow MQA to pass through to be decoded by hardware. Another choice: which USB latency mode should you use, Normal or Low. Normal is the more error-free option, it is said, but assuming it works, Low sounds better. I can confirm: It's a small difference, but it's real. If you need guidance with setup, there's an excellent knowledge base and a community of supportive users (footnote 7).

---

Footnote 1: Another legitimate comparison would be with the ZEN Next-Gen, which has less processing power but shares many of the ZENith's Next-Gen's purist/perfectionist features.

Footnote 2: In February 2025, Qobuz was beta-testing Qobuz Connect, which allows you to deliver music directly to your UPnP-enabled device, as with Tidal Connect. Innuos confirms that its products will be ready when Qobuz Connect is available

Footnote 3: Real-Time Linux was first included in Linux kernel v6.12, which was released on November 17, 2024.

Footnote 4: A couple of years ago at High End Munich, I first witnessed an Innuos demonstration of the improvement a better power supply can make in the music resulting from an all-digital server. I can't explain it, but I heard what I heard.

Footnote 5: See stereophile.com/content/innuos-statement-music-server. I recommend reading this review for more insight into how Innuos fights noise and other issues that can affect digital sound.

Footnote 6: I first started using Innuos Sense with the Innuos Statement, which has been in my system since mid-2023.

Footnote 7: See innuos.com/support and community.innuos.com.