Changing MB from ASUS to Gigabyte - what can I expect in reinstall etc?

[lfm]

I have not found any spread spectrum on either asus or this Giga mb.
Closest is Intel Speedshift which I have enabled, as I recall.

Lots of different ratios and whatnot.
Active Turbo ratios and allows to set for each cpu if that is multiplier 49,48,47 almost different for each core.

I tested to fix cpu voltage the other day, to 1.25V but that made no difference than raise a temperature 2 degrees whatever it was.

Seems to me so many other considerations than audio are in focus when optimizing performance.

I reported to Intel today, and got a bunch of tests to run and send them results. So will see if they are up to me job and track this down.

[Synthman2000]

Yes spacers do the ground to chassis but I recall having a corner of the MB causing some big problems some time not meeting with spacer/chassis somehow. When I pushed it down with thumb, issue vanished... maybe 9 years ago. (Cannot even recall what issue was now but something fundamental like not even booting) You would think if 1 spacer was earthing then the entire MB would be but it was causing a problem on some circuits.

Not saying this is your issue but a lot of people building a pc can get into unforeseen troubles with CPU mounts and pins, coolers, slots and screwing stuff up. It is pretty easy to mess up as I have found out myself a few times, even with great care. You only need a split second of inattention.

I hope you resolve the problem, good luck hunting.. and if not, well it's -80dB so run your mixing a bit hot maybe. So a bit less gain on the audio side post sound cards outputs and so better SNR so that noise should be buried pretty far down upon playback.

[lfm]

Grounding mb makes perfect sense. I have a vague memory that one mb I swapped 20 years ago that there was on original a earth-thing on one screw of mb, but then it was mostly plastic spacers also. And very careful not to create tensions between different parts of mb by screwing to hard etc, so it still could move a little with heat if needed.

So looked carefully about any advice in manuals to mb, psu and chassis what to look out for.

So a bit surprised there are strictly 8 screws and one metal spacer for mb.

"unforseen troubles", you can say that again and again.

But lucky enough is not showstopper thing, just that not going for acceptable spending this money on new computer. I'm going for perfect that is what had experienced for 20 years of daw computers. Had a Dell Optiplex 2004-2010 with internal RME Hammerfall 9636 and all silent no digital noise or anything in audio. The a new Dell 2010 with internal RME HDSP 9632 until now, still working fine if I wanted to use it. That also dead silent and no digital noise anywhere finding it's way into audio.

So was convinced it was a usb thing in my current case, but not so sure anymore as internal onboard audio produce similar unwanted stuff. But there aren't that many internal interfaces that meet my needs anymore with ADAT as one thing. RME has this HDSP AIO, same as the PCI-one 9632, but twice as expensive as Digiface USB. So felt usb has probably matured and that's why just about all audio interfaces are usb nowadays.

I have an old Avid Fasttrack Solo to evaluate ProTools long ago, that produced the same noise. Presonus Studio 1824c same thing so was returned.

Thought maybe to get an PCI to PCI Express adapter, that cost €50, and test my former 9632 if that is better isolated from digital noise. It's a full size card, not low profile, so will not fit with case all closed but as a test.

Will see what comes from Intel on this, if they stick with you to solve it or not. I hope I provided enough information to at least make them curious and not dismiss the whole thing.

Both mb's had Intel cpu, Intel chipset, Intel usb 3 controller. So either Intel or Microsoft is having an issue. I would not surprised if Microsoft, they seem all focused on security things nowadays and also read that Windows 11 22H2 at some point introduced way high DPC latency which seem not the case anymore. Just tells a story that focus is on other things than audio.

But could be a bios thing also with all the settings that are available with frequencies and voltages and stuff. Just that mb makers are focused on gamers to a large part. That is a huge business.

[camsr]

Is this an issue even if you are only using 2 output channels for stereo playback and nothing else?

Thanks for posting the sample and screenshot. A hunch tells me this issue is software, due to the number of errored samples being 16. But that's just a hunch.

I remember when using Vista that disabling Aero desktop interface had a massive performance boost, in the latency department. That could have been because it triple buffers, or because it was just unoptimized and clunky

[lfm]

Yes, a daw is not needed. An audio player like Foobar2000 has the same result over windows speaker settings.

You gave me a new idea, the appearance settings to turn off stuff you mentioned.
Thanks.
- edit: all was unchecked apart from smooth scrolling so no difference

A new finding was using PowerSettingsExplorer and unhiding Interrupt Steering Settings.
- some parameters for mode like, use any unparked processor with delay
- and set time delay how long a core should be unparked before used for interrupts
- so changed from default to any unparked processor with delay, and increase delay to 300 ms

Made a notable difference. Even that these things make a difference tells a story.
- so many things are happening with cores all the time
- threads are moved all the time according to various settings, round robin, sequential etc
- you can see how the same amount of cores used suddenly are on other cores
- just a small glitch in all this can cause this kind of issue, I assume

[Pictus]

A new finding was using PowerSettingsExplorer and unhiding Interrupt Steering Settings.
- some parameters for mode like, use any unparked processor with delay
- and set time delay how long a core should be unparked before used for interrupts
- so changed from default to any unparked processor with delay, and increase delay to 300 ms

Interesting, what are the names of the options? Can you show a screenshot?
Thanks!

[lfm]

It's a separate part before cpu settings and just called "Interrupt Steering Settings".
That in turn has only 3 parameters
- steering mode, 5-6 options
- time delay for unparked to be used
- cpu load(I think it was)

But it could be a windows 11 thing, my windows 10 laptop did not have many settings as windows 11 do.

To continue experiments...

Right now the best settings in bios were changed to
- C-states off again
- E-cores off
- above Interrupt steering stays as I put it Any unparked Core with delay, tried a few more modes

So just using cores that have the same frequency. E-cores are having 75% only, and probably very little cache so might be half in performance. I think AMI bios call E-cores ATOM cores to tell how weak they are.

What I could see from Process Lasso that shows cores visually you can see that daw application got as much places on E-cores. So distribution maybe have to be with E-core aware software that only place certain non-time critical stuff there, not sure.

So that's why I thought E-cores better gone. You now also see in task manager that cpu is steadily 4.5 GHz, since when E-cores were there is were around 3.8 GHz since that takes average down with lower frequency.

Intel came up with nothing on my reports other than did the cpu tests. Then using other intel chips to contact mb makers. Will see what Gigabyte come up with. I told the full story that other Asus board show same so maybe Intel chipset or on Windows side. If to make mb makers work together with Intel and Microsoft on maybe altering bios alternatives given.

Really looking for comprehesive guides to configure all that is there
- there are loads of voltages that are dynamically altered
- there are frequency boosts high and low changing in a dynamic fashion
- temperatures are measured and reacted on constantly
- cores are changing performance modes constantly all the way down to parking
- cores are awakened and unparked constantly
- threads are constantly moved over to other cores to distribute heat more evenly

You see this clearly as you watch cores running, suddenly the same amount of cores are moved to other cores.
- focus here seem to be heat distribution
- but I think there are flawed routines not having enough time space at some occasions
- this is the assumption anyway
- but it's not the full story and maybe all wrong, since computer is stable as anything.
- but audio is probably the most demanding thing to keep the pace of samples going
- video is one frame every 15-25 ms or so to be ready, for audio every ms counts.

I did all sorts of tests with turning off hyperthreading and runnings 8 cores. CPU doubles but audio is clean but rare occations the ticks are more rare but a bit worse and stronger than with HT on.

In phones I probably have SPL of 90 dB or so on full mix, and listening to a piano track only the faint ticks are just about audible.

So work with mixes progress nicely, and just put some time into this audio now and then since I am stubborn as hell to solve things, or my software business would not have lasted 16 years.

[2DaT]

I did all sorts of tests with turning off hyperthreading and runnings 8 cores. CPU doubles but audio is clean but rare occations the ticks are more rare but a bit worse and stronger than with HT on.

In phones I probably have SPL of 90 dB or so on full mix, and listening to a piano track only the faint ticks are just about audible.

So work with mixes progress nicely, and just put some time into this audio now and then since I am stubborn as hell to solve things, or my software business would not have lasted 16 years.

I really doubt it's CPU/MB related at all. It's either static electricity from your carpet/clothes, or induced noise from your PSU/UPS or mains. Try to move your sound card and headphone cable away from power sources and the computer itself.

Messing with the interrupts will just make things worse. I run my windows stock with high performance on NVIDIA panel and in the system itself and that's it. No dropouts, no "ticks", no noise, no anything. I didn't have to mess with the interrupts, didn't have to disable half of my system, didn't have to mess with frequencies and c-states, didn't have to mess with anything. It just works and it should work on your system without problems too.

If anything, maybe you should try a clean install, because all the things you did could make it worse for audio. But again, I really doubt it's a CPU/MB/Software problem. It could be an old refrigerator next to your room, or a vacuum cleaner that somebody turned on and the power dipped a little bit, or your UPS is acting up.

I would look into analogue side of things. You checked everything reasonable on your PC and it's not helping.

[Pictus]

Messing with the interrupts will just make things worse. I run my windows stock with high performance on NVIDIA panel and in the system itself and that's it. No dropouts, no "ticks", no noise, no anything. I didn't have to mess with the interrupts, didn't have to disable half of my system, didn't have to mess with frequencies and c-states, didn't have to mess with anything. It just works and it should work on your system without problems too.

Here the default install works, but I hate bloatware and
I like to keep the system as fast/responsive as possible.
If it's something I don't use/need, then I don't want it to
stay running in the background.
AMD Ryzen 5600X + tweaked Windows 11 22H2 build 22621.1555
VCV Rack in the background playing some regenerative stuff and browsing(wired LAN) via LibreWolf.

[2DaT]

Here the default install works, but I hate bloatware and
I like to keep the system as fast/responsive as possible.
If it's something I don't use/need, then I don't want it to
stay running in the background.

I have comparable numbers and didn't have to do anything special at all. But I do agree that bloatware can be terrible. Even something simple such as frequency/temperature monitoring program can cause some issues, because they poll low level drivers and some of them may be not that fast or respectful of the audio threads.

[camsr]

I agree that determining if these problems are buffer underruns is most important. You may be able to record the SPDIF signal coming out before it goes analog. Also, if you are using large buffer sizes, and just listening to stuff, I highly doubt any buffer underruns are occuring at all.

[camsr]

OT but, are E-cores and P-cores separate units on the processor die? Disabling E-cores seems wasteful if they are.

[lfm]

I really doubt it's CPU/MB related at all. It's either static electricity from your carpet/clothes, or induced noise from your PSU/UPS or mains. Try to move your sound card and headphone cable away from power sources and the computer itself.

Messing with the interrupts will just make things worse. I run my windows stock with high performance on NVIDIA panel and in the system itself and that's it. No dropouts, no "ticks", no noise, no anything. I didn't have to mess with the interrupts, didn't have to disable half of my system, didn't have to mess with frequencies and c-states, didn't have to mess with anything. It just works and it should work on your system without problems too.

If anything, maybe you should try a clean install, because all the things you did could make it worse for audio. But again, I really doubt it's a CPU/MB/Software problem. It could be an old refrigerator next to your room, or a vacuum cleaner that somebody turned on and the power dipped a little bit, or your UPS is acting up.

I would look into analogue side of things. You checked everything reasonable on your PC and it's not helping.

Thanks.

You're not missing any interrupts, they are just placed on cores being unparked long enough. To make voltages stabilize and things like that I think.

Now with C-states off again, it's probably more on/off thing for cores. But have set now min cores to be 4 being unparked all the time.

If on analog side I would imagine it would be strictly on both channels all the time. It can be as heard and also recorded from phones out to be sometimes on one side. This seems like a glitch in some timing of things, sometimes just affect a couple of samples, sometimes more muting both channels for some samples. Early I talked to RME about it, it's their interface, and he was stumped that both channels were muted.

Looking at this RME asked if I could record ADAT out, and I did to my former daw and the muting was strictly 16 samples long in that case using the Asus MB. So it enters audio stream.
Here a view from audio
http://cloudmanfilms.ownit.nu/digitest-1949tick.jpg

This is no static. Lower track is digital recorded over ADAT, and strict square wave.

And also during some tests ran a mains double converter that always brings 230 VAC 50 Hz, not matter what mains are, usually 238 VAC here. And my mains is not in some industrial area, I listen to various pods from a laptop a couple of hours a day, and no ticks there at all.

And over all that ticks are affected by various settings in bios and windows power plan is another thing to consider why that can be.

Even running onboard audio, even after mb change to other brand, it still happends.

E-cores are some weak cores, more like laptop as I see it. I ran cpubenchmark.net and their program. Running with E-cores it is as average testa sent to them, 31 000 or so. Without E-cores 29 000 cpu rating. So my former daw cpu i7-860 rate 5000. So this is still the boost I want mostly for video editing doing multiple camera editing which barely can be done on former daw.

This cpu, i7-12700F, is just one year since released on market. windows 11 is not that old either. With the amount of parameters to handle, in bios and os, I would not be surprised if one update of something will fix it all.

I am running largest project so far at about 13% utilization. So all these variations of voltages, moving threads between cores and stuff that would be good if going over 50% utilization this is really needed probably. CPU about 40-45 degress Celsius tops, VRM and PCH at tops 50 C. Nothing is about to burst, really.

To take on you tip of distance to PSU I will check some things. Main to cpu goes close to mb, a 8+4 thing, so will check a bit. But still mb is changed, so that both would have same "feature" still far fetched.

The timing of ticks reminds a bit of some sort of interference between frequencies. Like when having harmonic beating kind of thing, but more rare. That some frequencies create troughs and peaks or something. Like two quartz clocks for sample rate interfere just slightly apart. Something like that.

I pushed RME on this sync thing a bit, but they dismiss that interface has problems. Also trying Presonus Studio 1824c with same setup confirms that he is probably right.

I have ran interface with internal clock to rule this out that ADAT sync from rack preamps. And also running other interface like Avid Fasttrack Solo still making these ticks rules out that part too. No external sync at all.

Playing back audio with media player like Foobar2000 also create this. So not a daw particular thing either. Utilization at 1% or so.

It's weird, keep ideas coming please, I will consider them all.

[lfm]

I agree that determining if these problems are buffer underruns is most important. You may be able to record the SPDIF signal coming out before it goes analog. Also, if you are using large buffer sizes, and just listening to stuff, I highly doubt any buffer underruns are occuring at all.

It's good idea and RME suggested recording digital. Here it is.
http://cloudmanfilms.ownit.nu/digitest-1949tick.jpg

Lower track is digital.

I also ran spdif out to other dac and amp in room and it's surely in audio stream.

[lfm]

OT but, are E-cores and P-cores separate units on the processor die? Disabling E-cores seems wasteful if they are.

CPU rating goes down from 31 000 to 29 000 disabling them. To get rid of having threads on cores with half performance seems beneficial.

I used cpubenchmark.net own benchmark they use for presenting results from users out there.