CPU Load and Sound Quality

[Summit]

4 hours ago, Blackmorec said:

Do you know what I find most amusing about this thread? The fact that certain forum members constantly accuse subjectivists of confirmation bias, while being themselves completely delusional when it comes to hi-fi.  Friends and I have fooled around with hi-fi for over 45 years and if there’s 1 lesson we’ve learned it’s that set-up is by far the most important single aspect of getting the best sound out of any kit.  You can buy the best amp, the best speakers and the best servers but if you neglect to optimise its set up all you’ve done is waste an awful lot of money. 

 

This is my understanding end experience as well.

[Summit]

53 minutes ago, charlesphoto said:

I never realized there are so many ten year olds into digital audio, and so many that can go on and on about it in thousands of posts. It gets really really old sometimes. Children are great - I have two mod them - but we're also teaching them the responsibilities that come with maturing, like when and where to keep your opinions to yourself. 

 

Time to grow the f*%k up a little around here - as I say often, it's only STUFF. Don't take it so damn personally, and don't take it out on another person because you think they're wrong and you are right. And if all you can add to a conversation is thousands of one lined barbs or huge diatribes about being poor me and those internet bullies, then maybe it's time to find a new hobby/life. 

 

Age is no guaranty for good behavior or wisdom – just look what the (mostly 50+) decision makers does with our planet.   

 

[John Dyson]

40 minutes ago, Summit said:

 

Yes it could have been an interesting topic to discuss that I like to know more about.

 

My hypotheses is that a higher CPU load can have a small impact on SQ. My assumption is based the below 3 aspects.

 

1.  Higher CPU load means more digital processing which I believe (everything else equal) will generate more EMI and RF noise. 

 

2. Higher CPU load generate more heat which can inflict negatively on other components on the MB and how they operate, but it will depend on design and topology of the PC/server.

 

3. Higher CPU load will need more power and more power (everything else equal) generate more HF noise.

 

The question for me is how much higher CPU load dose it takes before we can perceive and measure any difference. Is it 1, 5 or 50 %?

Also is it better to get a bigger CPU that can do the same operations with lower CPU load or will the result be the same in the end?

When people talk about the CPUs causing noise -- well, there is a reason for all of those capacitors inside of the CPU packages -- they are mega noise generators -- pulsing current in the 100 Amp range.  (Well, ARM, not as much -- Intel/AMD, yes.)  Changing current through wires, like changing voltage across a boundry of conductors -- each can transmit noise.  The big bugaboo is when it happens that the impedance of the connections match the current & voltage (simplified), then you get radiation, and that goes a long way -  not quite as intense as, say a couple of conductors nestled near each other though.

 

John

 

[marce]

Those capacitors are a critical part of the power delivery system (without them its going to be noisier), the power is generally distributed by planes in mother boards. That said commercial motherboards are lacking on ground planes, often being 6-8 layers, having a dedicated GND plane for every signal and power layer layer provides a solid return path and minimises the loop area, greatly reducing emissions. All those little transistors switching of course will create SSN, spread spectrum clocking can help level the peaks.

Don't understand what you mean by the connector impedance matching the voltage & current, creating noise... The signals are designed to match the interface requirements (PCIe, DDR memory), the power planes should be designed to have the lowest impedance at the required operating frequency.

A standard PCB is best in a nice metal case of course.

[lmitche]

1 hour ago, John Dyson said:

When people talk about the CPUs causing noise -- well, there is a reason for all of those capacitors inside of the CPU packages -- they are mega noise generators -- pulsing current in the 100 Amp range.  (Well, ARM, not as much -- Intel/AMD, yes.)  Changing current through wires, like changing voltage across a boundry of conductors -- each can transmit noise.  The big bugaboo is when it happens that the impedance of the connections match the current & voltage (simplified), then you get radiation, and that goes a long way -  not quite as intense as, say a couple of conductors nestled near each other though.

 

John

 

It is terrific to see someone discuss what is observed here, that SQ largely correlates with the dynamics of supply and demand for current, and perhaps not just the magnitude of average current demand.

 

Put simply the lower the variance of current demand, the better the SQ. Variance can be controlled by lowering clock speeds, disabling energy saving features and tuning memory timings. Likewise stopping unnecessary processing and disablement of unnecessary circuits can often be accomplished through bios or software settings.

 

On the supply side, a power supply that can keep up with the volatility of demand for current is key, and hence the reason some supplies sound better then others.

 

Lastly, it is much more difficult to build a fast, high current DC power supply vs. a lower current supply. Hence low current demand devices like NUCs are well matched with low current, fast supplies like the Paul Hynes SR4. High current fast power supplies are few and far between, are either unobtainable or stupidly expensive.

[Ralf11]

6 hours ago, mansr said:

Where did anyone express such a belief?

 

No one did - it's a strawman like his other guff

 

I've been involved with HiFi for longer than 45 years, so that claim is also made up

[Ralf11]

if CPU load affects SQ, you have done something wrong; same for the 'dynamics' of the PS

[daverich4]

2 hours ago, mansr said:

Another way of describing the process is as a (weak) high-pass filter.

That’s correct. In fact, another way of creating the effect is by using a high-pass filter on a duplicate layer of the background with the blend mode set to Overlay. The amount is controlled by changing the layer opacity. 

[mansr]

2 hours ago, Summit said:

Yes it could have been an interesting topic to discuss that I like to know more about.

 

My hypotheses is that a higher CPU load can have a small impact on SQ. My assumption is based the below 3 aspects.

 

1.  Higher CPU load means more digital processing which I believe (everything else equal) will generate more EMI and RF noise. 

True. The question is, however, how much of that noise escapes from the computer box, how much of that then gets into the DAC box, and what effect it has there. It is not a given that there has to be any difference at all in the actual analogue output from the DAC.

 

2 hours ago, Summit said:

2. Higher CPU load generate more heat which can inflict negatively on other components on the MB and how they operate, but it will depend on design and topology of the PC/server.

If things get too hot, they will indeed malfunction. Assuming the computer is not that hot, what heat it does generate will not adversely affect its operation. That's pretty much by definition. If analogue circuits get warmer, they also get noisier while otherwise operating correctly (at some point they will fail, of course). It is probably best not to put the DAC chip or line drivers right beside a red-hot CPU. That said, if the DAC is being heated to any appreciable extent by the CPU, the bigger problem by far is going to be interference, not thermal noise. Unless the temperature of the entire room is raised significantly, a DAC in a separate box is not going to suffer from heat generated by the CPU.

 

2 hours ago, Summit said:

3. Higher CPU load will need more power and more power (everything else equal) generate more HF noise.

Power in and of itself does not create noise. For noise to be created, there must be some kind of variation. This point is thus actually the same as #1.

 

2 hours ago, Summit said:

The question for me is how much higher CPU load dose it takes before we can perceive and measure any difference. Is it 1, 5 or 50 %?

That depends entirely on the specifics of each system. I have seen on-board sound cards pick up all manner of noise from within the computer. On the other end of the spectrum, a well made DAC in a separate enclosure should be pretty much immune to the goings on in the computer. Even internal sound cards can be very quiet if designed well.

 

2 hours ago, Summit said:

Also is it better to get a bigger CPU that can do the same operations with lower CPU load or will the result be the same in the end?

The obvious answer is whichever CPU gets the job done while producing the least amount of noise. I suspect this will often be one with only a little (10-20%) more computing power than necessary. With the variation in available CPU (and system) designs, a simple answer is probably not possible. Better to make sure the DAC is properly shielded from interference and worry about something else.

[John Dyson]

55 minutes ago, marce said:

Those capacitors are a critical part of the power delivery system (without them its going to be noisier), the power is generally distributed by planes in mother boards. That said commercial motherboards are lacking on ground planes, often being 6-8 layers, having a dedicated GND plane for every signal and power layer layer provides a solid return path and minimises the loop area, greatly reducing emissions. All those little transistors switching of course will create SSN, spread spectrum clocking can help level the peaks.

Don't understand what you mean by the connector impedance matching the voltage & current, creating noise... The signals are designed to match the interface requirements (PCIe, DDR memory), the power planes should be designed to have the lowest impedance at the required operating frequency.

A standard PCB is best in a nice metal case of course.

I don't understand the quandry -- L * di/dt will create magnetic fields that can be picked up as currents in other conductors.  Equivalent for capancitors, but voltage instead.  When both are existent and are more or less matched by whatever antenna/wire/etc  to the 377 ohm impedance of space, then you get radiation.  Simple EE/Physics stuff..   Radiation can be received over longer distances, and has a different propagation than the simple coupling as I described..

 

John

[John Dyson]

32 minutes ago, Ralf11 said:

if CPU load affects SQ, you have done something wrong; same for the 'dynamics' of the PS

That is true, but there very often something measurably 'wrong'.  Whether or not the 'wrongness' is important, that is in the eye of the beholder.

 

[Ralf11]

in the ear, really

 

 

I'd need to see some numbers on the radiation getting out of one sealed box, a foot or more of travel, and into another sealed box...

[John Dyson]

29 minutes ago, Ralf11 said:

in the ear, really

 

 

I'd need to see some numbers on the radiation getting out of one sealed box, a foot or more of travel, and into another sealed box...

Most of the problems are NOT radiation so much as coupling.   Radiation is the free space thing, while coupling is the mutual current flow in wires or the voltage transmission between conductors (it is a bit tricky to describe what is happening without writing some equations -- a capacitor doesn't really transmit voltage, but the voltage changes on both sides tend to equalize -- ithe connection is actually current or electrons..)  Radiation has a different way of decaying vs distance, where the coupling is more intense, it falls off very quickly vs distance.

 

Radiation can be a problem -- mostly at higher RF frequencies, coupling is more of a problem at lower frequencies.  Low frequncies dont radiate very well without very large 'antennas'.  You can look at an antenna as a matching device which matches the signal to free space.

 

One problem is when higher frequencies do leak into audio circuitry -- they sometimes do not deal with the high frequency disruptions very well, and you can easily get something commonly called 'detection', but is really a manifestation of the circuit nonlinearity.

 

 

John

 

[fas42]

1 hour ago, Ralf11 said:

if CPU load affects SQ, you have done something wrong; same for the 'dynamics' of the PS

 

What has been "done wrong" is to assume that the design of all the areas of an audio chain is good enough - by the industry as a whole. If a system only performs at a relatively low standard, compared to what's possible, all of this fiddly stuff doesn't matter. If one aims for a higher SQ, then it becomes a bit of a nightmare - because effectively one has to dive in and re-engineer so many things; or build workarounds on the outside, as many do using added 'gizmos'. Blackmorec points out that there is light at the end of the tunnel if one perseveres, John is being thoughtful about some of the mechanisms that can be relevant; it's all part of a single story: normal audio engineering is just not good enough to deliver the best SQ.

 

To throw in the requisite car analogy, if one only uses a high performance vehicle to go shopping in, any old rubbishy rubber is good enough to use on one's wheels; but if you desire to find the limits of what you car can do, then the tyres you fit will be critical to achieving the best that can be reached.

[sandyk]

4 hours ago, daverich4 said:

 

Alex,

As much as you might know about Audio, Digital Photography seems to be somewhat outside of your area of knowledge. What you are describing is correctly identified as "Acutance" which Wikipedia defines as "a subjective perception of sharpness that is related to the edge contrast of an image". Photoshop and other image programs simulate this with a techniques called Unsharp Masking which is the digital version of a technique used in Analog Photography whereby a slightly soft mask was sandwiched in register with a negative or transparency to increase contrast along edges of detail. It does not actually increase sharpness, it just appears sharper to the eye. In Digital Photography this technique is accomplished by adding dark pixels on one side of a line and light pixels on the other side of it. The width of these added pixels is varied depending on resolution of the image and a number of other factors. There is absolutely no possibility that the USB Regen could do anything that would correctly add those extra pixels to your images. Absolutely NONE.

Dave

 The USB Regen has NOT added extra pixels to the image as the files are still BIT PERFECT.

 There is far more to this area of digital processing than is currently realised. The USB Regen does exactly the same to both  Digital Audio and Video as has already been confirmed, first in the series of 6 separate DBTs with Audio, and more recently with Music Video clips where  several members confirmed marked differences with both A and V.

In fact , with DTV you can do exactly the same with saved files on an STB, but instead of using a USB Regen, powering the USB memory stick via a separate Linear PSU instead of the noisy +5V SMPS from the STB, in this case it simply involves making the PSU to have a markedly lower output impedance to >100 kHz.

 About 12 months ago,the local HDTV transmitters were changed from mpeg 2 to .mp4 with .aac audio. However, initially one broadcaster used way too low a Bit Rate, and the transmission were markedly softer looking than the original mpeg 2 high bit rate transmissions, and not much better than SD TV. By doing this, the saved programs when played back looked very close to that of the original .mpeg2 transmissions, and were considerably improved over the direct transmissions. Several months later, probably due to viewer complaints , they increased the bit rate of the .mp4 HDTV transmissions and the linear PSU for the USB memory stick was no longer needed.

 

Alex 

[daverich4]

5 minutes ago, sandyk said:

Dave

 The USB Regen has NOT added extra pixels to the image as the files are still BIT PERFECT.

 There is far more to this area of digital processing than is currently realised. The USB Regen does exactly the same to both  Digital Audio and Video as has already been confirmed, first in the series of 6 separate DBTs with Audio, and more recently with Music Video clips where  several members confirmed marked differences with both A and V.

In fact , with DTV you can do exactly the same with saved files on an STB, but instead of using a USB Regen, powering the USB memory stick via a separate Linear PSU instead of the noisy +5V SMPS from the STB, in this case it simply involves making the PSU to have a markedly lower output impedance to >100 kHz.

 About 12 months ago,the local HDTV transmitters were changed from mpeg 2 to .mp4 with .aac audio. However, initially one broadcaster used way too low a Bit Rate, and the transmission were markedly softer looking than the original mpeg 2 high bit rate transmissions, and not much better than SD TV. By doing this, the saved programs when played back looked very close to that of the original .mpeg2 transmissions, and were considerably improved over the direct transmissions. Several months later, probably due to viewer complaints , they increased the bit rate of the .mp4 HDTV transmissions and the linear PSU for the USB memory stick was no longer needed.

 

Alex 

Alex, we were talking about digital photography and you responded with a long unrelated post about digital video and audio files. So I will repeat again, the USB Regen did absolutely ZERO to alter the apparent sharpness of your images. 

[sandyk]

1 hour ago, John Dyson said:

Most of the problems are NOT radiation so much as coupling.   Radiation is the free space thing, while coupling is the mutual current flow in wires or the voltage transmission between conductors (it is a bit tricky to describe what is happening without writing some equations -- a capacitor doesn't really transmit voltage, but the voltage changes on both sides tend to equalize -- ithe connection is actually current or electrons..)  Radiation has a different way of decaying vs distance, where the coupling is more intense, it falls off very quickly vs distance.

 

Radiation can be a problem -- mostly at higher RF frequencies, coupling is more of a problem at lower frequencies.  Low frequncies dont radiate very well without very large 'antennas'.  You can look at an antenna as a matching device which matches the signal to free space.

 

One problem is when higher frequencies do leak into audio circuitry -- they sometimes do not deal with the high frequency disruptions very well, and you can easily get something commonly called 'detection', but is really a manifestation of the circuit nonlinearity.

 

 

John

 

 See also :

https://www.dropbox.com/s/c5unyq3dqvvg2gk/14A2-B4.pdf?dl=0

 

 This information was sent to me by member al.fe who prefers HDD over SSD for Audio, and is highly qualified in this area.

 

[sandyk]

19 minutes ago, daverich4 said:

Alex, we were talking about digital photography and you responded with a long unrelated post about digital video and audio files. So I will repeat again, the USB Regen did absolutely ZERO to alter the apparent sharpness of your images. 

 I repeat. The images that were saved using the USB Regen then copied back to a folder in the same HDD as the original series of images were noticeably improved. when compared with the unprocessed files.

 It doesn't matter whether the DIGITAL files are those of Photographs, Audio or Video. 

 

Despite what many here may insist , I do NOT imagine these differences with Digital files !!!

[Racerxnet]

19 minutes ago, sandyk said:

 I repeat. The images that were saved using the USB Regen then copied back to a folder in the same HDD as the original series of images were noticeably improved. when compared with the unprocessed files.

 It doesn't matter whether the DIGITAL files are those of Photographs, Audio or Video. 

 

Despite what many here may insist , I do NOT imagine these differences with Digital files !!!

Color space, Gamma, 16-235 or 0-255, White balance??  Driver for the video card? You think a regen changes the video picture???? MPC-HC, MADVR for the renderer? LAV filters? Did you calibrate with Calman, HCFR or other software and a i1Pro for a baseline?

 

Your WAY out there right now. Anyhow, get back on topic please.

 

MAK

[sandyk]

6 hours ago, charlesphoto said:

He can't be serious. Or can he? Thing is, vision is a very different sense than hearing. So in audio it's possible to hear things from certain improvements such as the Regen, but those improvements don't translate into other digital or sensory realms.

 The same rules apply for all DIGITAL processing .

 If anyone wishes to discuss this further with me then please take it to PMs as it is off topic.

[daverich4]

4 hours ago, sandyk said:

 I repeat. The images that were saved using the USB Regen then copied back to a folder in the same HDD as the original series of images were noticeably improved. when compared with the unprocessed files.

 It doesn't matter whether the DIGITAL files are those of Photographs, Audio or Video. 

 

Despite what many here may insist , I do NOT imagine these differences with Digital files !!!

Alex, you clearly are trying to express an opinion on a subject about which you have minimal knowledge. This is not a photo forum however so I will offer, as I did the last time you wandered off into a technical discussion of photography here, to meet you over on The Luminous Landscape or DP Review and try again to explain the basics. 

[sandyk]

17 minutes ago, daverich4 said:

Alex, you clearly are trying to express an opinion on a subject about which you have minimal knowledge. This is not a photo forum however so I will offer, as I did the last time you wandered off into a technical discussion of photography here, to meet you over on The Luminous Landscape or DP Review and try again to explain the basics. 

 

 We aren't talking about Photo basics. We are talking about Digital files which no matter whether Photographic, Audio or Video STILL consist of Binary Data, represented by 1s and 0s .

You are now talking about a subject of which  you have obviously minimal knowledge about too.

 

Again, it's completely OFF TOPIC .

[PeterSt]

12 hours ago, Summit said:

Also is it better to get a bigger CPU that can do the same operations with lower CPU load or will the result be the same in the end?

 

The former.

[PeterSt]

12 hours ago, charlesphoto said:

Sharpening in photography is all predicated first on the sensor and lenses used, (and focusing technique) and then on the software. Bizarre...

 

OT: 

 

Leave out the AA filter, and you're way better off.

Except for one or two photos out of 1000, if you know what you're doing.

[PeterSt]

16 hours ago, Blackmorec said:

One of the most rewarding aspects of hi-fi is that performance isn’t directly correlated with how much money is spent, but it is directly correlated with how much care and attention goes into its installation. 

 

That. But sadly for all the nay-sayers in this thread, it is even far more related to all the seemingly small stuff. Like the source.

 

I have said it before - these days the PC is more important than the DAC. Now who is going to believe that for real eh ?

And don't give me "then the DAC has a few issues" lines.