Lay of the LAN. Before we talk sonic COS and effect, some words on my network status. As a classically trained former musician now web content creator, I prefer paying proper coin for the music I consume. That used to be CD at ~€15/pop. Today it's their digital folders on my SSD. Buying those, more of my cash ends up with the actual performers than with cloud streaming. On sonics too my local files still beat it. LHY Audio's SW-8 $600 network switch narrowed the gap but didn't close it. Recently I replaced a +9-year 27" iMac with a late year 2020 NOS version. I installed an extra 32GB of RAM and Audirvana Origin, then offboarded my library to a 4TB SSD via USB3. That gave local files a new bigger lead. If I streamed more cloud, I'd have bought Audirvana's Studio license instead. Like Roon it integrates local/offshore files to route both through its sound optimizer engine. That also defeats background computing threads. Most my streaming occurs on the desktop to hunt for music I want to own. So I shared an Origin license between office/music computers to focus on local files. Both stations exploit Audirvana's r8brain 4 x upsampler. As such cloud files entering my iMac over 20m CAT8a off our fiber-optic modem didn't see Audirvana. Instead they saw Apple's Core Audio embedded in macOS Ventura 13.4. Was that an unfair advantage for my local yokels? The thing about unfair advantages is being daft if we don't press 'em. But I subsequently wanted to know how much Audirvana pips Core Audio. So I bought Audirvana's Studio license. With my existing Origin account that came to just €40/yr. Now iMac and HP Workstation both host Origin and Studio. I can still run Qobuz Sublime direct into my macOS or Win 10/64 sound engines; or through Studio. The latter gives cloud files the same massage as the locals enjoy. That means Audirvana's r8brain upsampler and direct/extreme mode to hog computing resources by shutting out competing processes.
I quickly learnt that without changing any hardware, better signal-routing software made a noticeable improvement to my cloud streaming. This reiterated a far earlier conclusion when I first started PCfi on an iMac. Then I had quickly circumvented iTunes with PureMusic later Audirvana. Both were significantly better than iTunes; yet between each other didn't sound alike. The takeaway is simple and not costly. A good bit of sonic improvement headroom hides inside player software, be it HQPlayer or Roon, Euphony or JPlay. And with a PC/Mac's processing powers dwarfing a DAC's FPGA, one can exploit complex 'convoluted' up/resampling algorithms embedded in such player software or added as plug-in. Often those do a demonstrably better job than a DAC's. In short, if your streaming budget can't stomach a hardware upgrade like today's, don't overlook a far cheaper software license. Back on my lay of the LAN, a fiber-optic modem/router from our local IT provider hardwires to a node on a mast outside my office. From there 20m of industrial CAT8a wire run along floor boards to the SW-8 switch in my music room. A 1m CAT8a lead then connects from that to my new iMac. That's because our household is WiFi allergic—fuzzy brain then headaches—so all audio must hardwire. It's also why we still run a legacy landline and only use a smartphone on the road.
As my very basic diagram shows, multiple computers on our home network not only increase the load on the router/modem and its bandwidth. They also tie together a plethora of components which connect to said computers. Those might be USB peripherals like mice, keyboards, monitors, printers/scanners or DACs, active speakers and other hifi kit routinely plugged into different spurs of dissimilar ground potential. It's intuitive that a network switch optimized for audio gear should block some of the noise that distributes across the Ethernet cables connecting the lot.
Asking Stephen Gong whether any of their ports have priority, "the switch is like any post office. It relays information packets from one port to another based on MAC addresses. All ports are supposed to be the same but of course aren't. The four near the SFP are from one chip, the other four from another. I personally prefer my DAC connected to the latter so away from the fiber-optic ports, the router in the other quad. If only a few ports are used, we recommend spacing them as far as possible. The unused copper ports should be closed since they're never truly idle when no cable plugs in. The Ethernet standard demands that unused ports constantly transmit fast link pulses. Those cause noise to compromise sonics. The enable/disable dip switches are read just once after power on reset. After each change you must power cycle so the change sets. Then give the S10 at least three seconds to discharge its bulk power caps before turning it back on." Following our man's advice, my incoming CAT8a cable thus seated in the right-most port, the outgoing shorty in the most leftie. When I fired up the S10 the port LEDs lit up just once in amber/green to show 100Mbps transmission from our router/modem. Then they turned off as the dip switch instructed. The S10's small frontal power indicator lit up an attractive white.
Don't share resources. It seems key to high-end hifi. Circuits turn gloriously selfish bastards. They refuse to share a joint power supply. They demand their own down to an OXCO; and soon a complete chassis and all that goes with it. On my Win10/64 workstation and 27" iMac, once Audirvana Studio hogged playback resources by shutting out competing programs and processes, the sound of Qobuz Sublime coming in over the network got clearly richer and smoother. One not very long session had my answer on how much Audirvana pipped the generics. True, only its French coders know exactly how their software enacts its piggish resource greed. Regardless, the very same general-purpose computers designed for maximal parallel processing performed better when they focused down harder on the one task I inspected. To hell with multi tasking that said. All we're doing is listening to music whilst involving a computer. With that settled, let's segue into S10 sonics; and difference margins.
