Quote:
I believe the improvement in sound quality from the 1st setup mentioned above could be from any number of factors, but has nothing to do with how many tasks are running on either of the devices. Some of these factors that could be making your first system sound better could be the digital media player that it uses, the bit/sample-rate at which it's playing, the fact that you are using both Wyrd and U12, and even the volume which being louder can actually make something sound better.
In all my approx. 20 years of listening to music with computers, I have never perceived one slightest inkling of difference on sound quality whether the computer is at an extremely low vs. extremely high CPU usage - except of course when the CPU / RAM gets full enough to the point that it causes audible stuttering or in-out signal loss. Otherwise, ALL modern computers use (vague approximation) about 30-50 processes / tasks, and around 2-3x more service running simultaneously - even the PC that you thought had only the digital media player running must have had at least half that many running concurrently. Modern computers are designed to be elaborate multi-tasking machines, and a simple digital audio player may only account for less than 1% of your computers' total processing power.
I will call this one a hypothesis, unless serious scientific testing is done with data show otherwise; -AND- if that difference is audible in a blind test.
Totally off-topic, but I wanted to point out that your approximation of 30-50 processes and 2-3x more services running is almost exactly spot on for a machine operating with ~8 applications open at once. On a Mac (or Linux, or nearly any *nix system), you can get an exact number by opening up a terminal and running "ps aux | wc -l". Currently, my machine has around 230 processes running at once.
I'm starting to write my own audio engine, so I've recently had some experience with possible factors that can influence playback. It's worth noting that audio playback itself is not a single process, but actually at least three simultaneous processes (although it may appear in programs like the task manager as a single application). One process* reads in your media (say FLAC) from your filesystem and pushes it to the decoder process. The decoder process decodes that FLAC file into PCM audio (similar to WAV format), and sends it to a mixer process. The mixer process accepts the PCM-encoded audio and sends to the hardware mixer, in this case over USB. If you have a spectral analyzer or a VU meter or another visualizer, that might be another link in the chain, or it might be computed in parallel.
Since processors are tasked with running over 200 simultaneous tasks at once, but generally only have 1 - 8 cores, the operating system will do something called "context switching." Essentially, the operating system will, thousands of times per second (potentially), "evict" a process from the CPU and swap in a process that has been waiting. By doing this extremely quickly, the operating system creates an illusion that all 200+ tasks are running at the same time.
So, if one is streaming 192Hz audio, which samples 192 times per second, couldn't such context switching be a problem if any of the three required processes aren't allowed to cycle at least 192 times per second? Absolutely. But, most good audio players tell the scheduler "hey, I'm a real-time media application playing at sample rate X" and the scheduler says "OK, I'll make sure you get enough time." And generally, the scheduler does this pretty well (it's worth noting that the scheduler used in the Linux kernel -- a multi-level queue with some advanced features -- is *substantially* better at doing this than the Windows scheduler). As DecentLevi noted, sometimes the computer gets so bogged down that the scheduler can't make its commitment, which generally happens when the computer experiences significant loads.
The Atom processor is a special case. It is designed to use extremely low power and is not very good at multi-tasking. Even running the scheduling algorithm itself can be stressful to these lightweight cores. They also tend to work with much lower voltage gaps (the amount of measured difference in a transistor to decide if a bit is a 0 or a 1), which makes them more susceptible to noise. This could potentially be a problem in an audio scenario, although I highly doubt it. Even these processes are executing cycles in gigahertz -- well above any sampling rate a human can discern. My only point here is that while a scheduler on a "full" CPU is almost certainly never going to give a media process less cycles than it needs unless the CPU is about to keel over. An Atom CPU, on the other hand... you can still be almost entirely certain, but not to the same extent.
Now, what kinds of degradation are actually possible? CPUs and RAM have ECC (error correcting codes) and a lot of fault-detection methods. Essentially, when a fault is detected, the required value is recomputed. That means that noise in an Atom system can result in a lot of redundant cycles, and a possible gap in the music. It is absolutely absurd to imply that the bits coming out of the USB could be any different. So the degradation from load on a machine (or noise in the power supply) is only going to show itself as either "skips" or "gaps" in the music, or as sub-discrete noise in the USB connection itself (which your Wyrd will take care of).
GioF71, it seems like you know what's going on computer wise (MPD + Linux) -- I'd love to get your opinion. And of course anyone else in the thread as well.
* I say process here, but more than likely they are implemented as threads.
Hello, while I cannot give a final answer, I can share my current experience with the Gustard X12 DAC.
The DAC is currently connected to 2 distinct computers:
1) Dedicated music player (ubuntu server + mpd + upmpdcli) with a dedicated Linear PSU, FT 146 from italian brand ZetaGI
Atom PC with ubuntu server/mpd/upmpdcli -> AudioQuest Cinnamon USB 0.75m -> Schiit Wyrd -> AudioQuest Cinnamon USB 0.75m -> Gustard U12 -> Ricable 0.5m HDMI Cable (italian brand again) -> Gustard X12
2) General purpose PC (i5 3300K IIRC)
PC -> AQVox linear USB PSU -> Audioquest Cinnamon USB 1.5m -> Gustard X12.
The DAC is then connected (via XLR) to a Gustard H10 which I use to drive my current best headphones, a pair of Sennheiser HD650. All listening impressions are done with these headphones.
Also, a desktop amplifier (SMSL SA50) is connected via RCA to the X12. I am using JBL Control One speakers.
Well, the first chain sounds quite a lot better than the second.
As you can see, there are more differences than the simple usb vs hdmi connection, last but not least the fact the pc is dedicated to playing music. No other tasks are running on that pc.
I have found that using a general purpose PC for playing music is very detrimental to the listening experience, especially if you do stuff (browsing, but especially, in my case, heavy file transfering/manipulation).
I don't consider using software like fidelify or similar because I think they simply defeat the idea of using a general purpouse pc, so I set-up this dedicated linux machine instead.
Anyway for clarification, I can tell you the first chain sounds better than the second even when nothing but foobar is running on the desktop PC. The difference is more subtle, but it's still there.
My .02€
EDIT: some corrections
Alright... I have to ask about the cinnamon cables. Bits are bits. I thought the AudioQuest argument was that traditional USB cables pushed additional noise into the DAC via sub-discrete variations in the single. If that's the case, then why run Computer -> Expensive USB -> Wyrd -> Expensive USB -> Gustard when doing Computer -> Regular USB -> Wyrd -> Expensive USB -> Gustard would produce the same results? Can you tell the difference?
Dedicated linear PSU and the MPD setup is sweet. I've been thinking of building something like that for a long time now in a MicroITX (MiniITX? I can never remember...) case, but I would grab a more powerful CPU.
There are surely more factors differentiating the two setups, but I can definitely hear differences (for worse) depending on cpu load. This may not be due to cpu load, but from more than normal noise generated by ethernet adapter, hard disks (though there are SSDs on the PC) and whatever. I just don't know what is the exact cause.
But the result is a distinct degradation of the sound quality.
The player does work under load. Jitter and transmission errors probably rise in such conditions.
Maybe some expert can give us a better explaination.
I simply do not believe in A/B comparisons.
I know I can clearly spot when my system is playing mp3 instead of (real) lossless. My wife always ends up choosing some crappy mp3s and I immediately recognize how bad they sound.
But under A/B test conditions if A is very close to B I end up trying to focus on some details, so I tend to get it wrong after a few iterations.
So I think it's a waste of time. At least for me
You're going to have to explain the A/B testing bit to me. I understand trying to "overfocus" and getting the wrong answer, but it still seems like some kind of blind testing is required to differentiate between confirmation bias ("I feel good about my setup, so now I will hear better things." Kind of like how food you cook yourself tastes better regardless...).
Have you tried A/B/X testing?
Well the dedicated computer is an Intel Atom D510, the other is an Intel I5 3300K.
The difference is in what the two computers are doing, in my opinion.
Recording the output is a good idea. Unfortunately I do not have a dedicated recorder but I could use a third computer with a Asus Xonar Essence STX which I already used to record from analog outputs. Let's hope the STX is neutral when recording.
Will try to do this in the next few weeks.
Can you connect your i5 box to the same chain as the atom box and compare?
Also, general question about the Wyrd -- seems like BS to me. Reclocking USB? Maybe that could matter if you had a really crappy USB controller on your mobo. Or some really bad cables. Or a terrible PSU. But in normal conditions... can you describe how it changes the sound w.r.t the Gustard?
