Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Matt]

We often hear claims that shutting down system services or other programs improves sound quality.

I think intuitively people believe this frees up the computer to "focus" on the audio.  What I know about computers tells me they do not work this way.

But it's important to test claims, even if you can't understand them.  So I've tried to test this claim many times.  Let me describe a recent test.

I tested using analog output, which is far more susceptible to electrical noise than USB, S/PDIF, etc.

I've got an X-Fi connected to a power amplifier. At full volume, the system is extremely loud (far past any threshold of comfort).

There's a small amount of shot noise in the system audible when your ear is within a foot of a speaker.

The computer is reasonably fast (overclocked i7) so it's possible to consume a lot of power when pushing it hard.

So this setup is ideal for testing for noise, because the high gain and ability to sit much closer to a speaker than normal greatly amplify any effects. And the ability to use a lot of power make it easy to test what CPU usage might do.

So with my ear a few inches from a speaker, set to full volume, playing silence or a low-level 1 kHz tone, I started and stopped Prime95 many times (Prime95 is a stress tester that pushes all cores to their max, so CPU usage was switching between 0.5% and 100%, with a swing of over 100W).

I can discern no audible change in the volume, color, or any other characteristic of the shot noise or the 1 kHz tone in this test.

So even with normal listening levels magnified by 50dB or more, in an extreme CPU usage example, using the most susceptible output type, there's no discernible difference of CPU usage on audio.

[Skogkatt]

Hi Matt,

I'm glad you started this thread because I could report something quite different from your findings.

After some time spent lusting over various tablet models, I took the plunge and bought one. The main application for it is remotely control my HTPC having the possibility to see on its screen what I would normally see on the monitor (I want to see all).

I initially used the tablet's pre-configured RDC based on Splashtop which requires to install a resident program in the HTPC to function. I started to play music and immediately felt something different, a sort of harshness that wasn't there before, not nice. The day after the nasty sensation was still there! Checking various settings I discovered that JRiver convolver speed was lower than usual (around 78X instead 90X on a i7 3770k 16GB). Turning off the RDC the sound returned to be what it should be.

Later on I decided to test different RDC programs on my tablet and found some that do not require a "server" installation on the HTPC: needless to say that they neither reduce the convolver performance nor they affect in an audible way the sound.

This isn't the only example I can tell, it is just the last experience occurred to me with CPU consuming tasks, not even intensive tasks.

My interface is an Asus ST modified to be clocked with an external oscillator having very low phase noise. The audio card feeds through a balanced line (AES-EBU) my DAC.

I'm afraid your test methodology wouldn't reveal much unless you start measuring the output with a spectrum analyzer and you are able to measure jitter on S/PDIF.

My humble suggestion would be to try again your test with music and a system that is transparent enough to reveal differences like changing interconnect cables.

I know this can open an endless debate on the audibility of different signal cables but in my experience they do sound different... a lot different.  

What's important is knowing very well how your system sounds and playing a well known music: one of the most revealing is instrumental music like baroque or classical.

[Ancient_Audiophile]

The flaw here is the concept that the ability to play a 1khz tone is equivalent to the ability to reproduce a music waveform over an extended period of time.

[Mike48]

Hi Matt,

I have, as yet, no opinion on the matter, as I haven't investigated it myself. It's worth noting that in the 1970s a bunch of horrible sounding solid-state gear was sold, based on very low distortion with steady-state tones.  Later, the phenomenon of TIM (transient intermodulation) distortion was discovered and explained why this gear measured great but sounded bad -- to those who really listened -- with music. Since then, audio manufacturers have reduced the amount of negative feedback used, especially global feedback, and many, probably most, listeners and critics approve.

The explanation I have heard for CPU-loading effects is that heavier CPU load is somehow jitter inducing. Why this should make a difference with modern DACs, which usually have methods to defeat jitter, I don't know. However, we must remember that the history of audio, as well as being littered with green pens and magic pucks, is also marked by statements by golden ears that something wasn't right -- though it measured fine -- and later confirmation from the engineering community. Indeed, my first CD player was just terrible sounding, though critics loved it. The gear I have now, almost 30 yr later, is tremendously more engaging. That was achieved through engineering, not magic, once the engineers figured out what the problem was and how to measure it.

I do think you are measuring the wrong thing in this case, but I don't think we yet know what to measure to quantify this effect. It may turn out that nothing measure it, as it's some kind of mass delusion; or it may be some kind of jitter modulation that is not well removed by the methods now in use. Or something else! It's great that you are trying to figure this out. You're not there yet!