I've recently moved on from Win 7 to Linux and will release a new machine based on a super-slim Linux version in late Dec. Our philosophy is a minimalist-approach, based on the CMP2 project. BTW, the CMP2 site is my most recommended source for DIYers who want to delve into building their own machine and the tutorial on the Genesis Loudspeaker website is also worthwhile.
Many servers out there try to be all things to all people, which is fine if you’re building a general-purpose multi-media server, but not when you’re talking about connecting it to a world-class audiophile system. In Windows, I can detect a change in audio quality just by turning off the Intel HD graphics driver and reducing the color depth to 8-bit; think about that if you’ve got $50k worth of gear hooked up to a server with HD graphics, network capability, iPad control, virus scanner, etc.
The last server, which I had in our reference system for a while, has a lightly tweaked Win 7 OS alongside a highly tweaked Win 7 audio OS. The normal boot section can connect to the Internet and is set up for Blu-ray and other video. Latency measurements in the audio boot section are less than half that of the regular section and the difference in sound quality is quite surprising, given that both OSs are booting on the same fan-less hardware.
As operating systems go for audio, a default installation in OSX (which is based on Unix) will beat a default installation in Windows (which is based on DOS) however, Windows is much more amenable to tweaking and you can disable substantial portions of it in the back end. A friend from my audio club and I compared the tweaked Windows/Foobar setup to his Mac Mini, with Pure Audio and Amarra and the Windows setup came out ahead with either player.
However, I believe Linux (which is based on Unix) presents by far the most elegant and efficient platform for audio. Unfortunately, the learning curve on Linux is quite high. When I started it took me 2 weeks to figure out how to install it and another 2 weeks to get the sound to work (this was after already knowing Windows like the back of my hand). Moving to a more compact and resource-efficient version of Linux presented a second learning curve. There are also potential hardware compatibility issues with Linux for certain sound cards or DAC with custom USB Drivers (we’ll offer the tweaked Windows audio OS as an additional option), but it’s not nearly as bad as you might think. Any DAC that doesn’t have custom drivers will work with Linux and many of those can still use the other drivers (ASIO4All works for my Wyred4Sound DAC, in Win 7, and sounds a little better than the custom driver.). My current audio OS in Linux uses approximately 20% of the memory resources as my Windows audio boot section and I still have a majority of the tweaking left to do (I think this might cut it by at least 60%).
BIOS optimization shouldn’t be overlooked either. A BIOS with default settings can have dramatically higher latency readings in the OS than with optimized settings. If you run a super-tweaked audio OS on the machine by itself you can go a little further in the BIOS in trimming the resources, but most people find it a little more convenient to have a 2nd OS to rip CDs on the same machine.
Anyway, there’s no shortage of DAC designers who will tell you that the server is irrelevant with their DAC or their async driver, etc. Pure BS in my opinion. On a highly-resolving system you will hear the hardware (motherboard, fans, drives, power supply, power cable), the OS (kernel, interfering processes and overhead, memory usage, drivers) and the player.