Thanks.
yes, I have a 10 year old Avid Fasttrack Solo which runs fine(usb as well).
- early before computer were setup good it gave ticks too
- now fine
It's been a while since the stronger tick on both channels, mostly these faint ones just on one ear.
I also run mains on a double converter that give perfect 230 VAC 50 Hz to exclude something on mains going through. When on both channels it is possible to be on analog side, I thought.
I also in general listen to various pods a couple of hours a day on another computer and there nothing funny on mains. I also run optical to external dac and listening there same ticks, and as darker track on first pic shows it's sudden zero level on audio and always same length 16 samples.
No network is up, since not online at all unless installing stuff etc.
There was one update on drivers since I reported to RME, but not firmware.
I've done real heavy usb 3 backups doing 2 Gbps along with listening to audio, and nothing is changing. So seems usb is solid in itself.
So search continues...
Intel 12th generation - is it at all possible to get a fixed cpu frequency - no throttling?
-
- KVRAF
- 16888 posts since 8 Mar, 2005 from Utrecht, Holland
It's not so much about people being affected. Your other electronic devices should work better with less RFI through the spectrum spread. It really won't hurt.lfm wrote: Thu Apr 06, 2023 7:17 amAnd who cares about EMI?2DaT wrote: Thu Apr 06, 2023 6:04 am There is no reason to disable spread specturm. If anything, this will reduce the EMI from your computer.
Most of the time you shouldn't mess with the stuff too much.
- we are surrounded by mobile phones and masts
- we surrounded by wifi routers looking for clients
We are the KVR collective. Resistance is futile. You will be assimilated. 
My MusicCalc is served over https!!
My MusicCalc is served over https!!
-
- KVRAF
- Topic Starter
- 7120 posts since 22 Jan, 2005 from Sweden
My assumption is that it might hurtBertKoor wrote: Thu Apr 06, 2023 10:20 am It's not so much about people being affected. Your other electronic devices should work better with less RFI through the spectrum spread. It really won't hurt.
- constantly changing cpu frequency cause jitter of various sort
- maybe on usb too
That is why I wanted to at least try fix cpu in every aspect to see if different.
When I used Asus Suite it said for all perfomance cores a fixed frequency.
- not sure how task manager measure cpu frequency because it jumps around whatever you do.
I had a theory for a bit and large enough jitter on usb bus could be cause of interface get buffer underrun or something, creating gaps like on first linked pic.
But editing pics and just move one side of gap does not quite make it ok, so think something can just mute audio instantly like this and samples lost, kind of.
Shooting in the dark and try anything I can come up with.
-
- KVRian
- 636 posts since 21 Jun, 2013
No it doesn't hurt anything. It's well within acceptable limits and won't cause anything bad. Changing cpu frequency doesn't cause jitters, because USB devices are not clocked by CPU clock.lfm wrote: Thu Apr 06, 2023 10:42 am My assumption is that it might hurt
- constantly changing cpu frequency cause jitter of various sort
- maybe on usb too
Remove that bloatware asap. These apps often cause problems.lfm wrote: Thu Apr 06, 2023 10:42 am When I used Asus Suite it said for all perfomance cores a fixed frequency.
That can happen if your USB cable is bad or if it's very long. Try different one if possible.lfm wrote: Thu Apr 06, 2023 10:42 am I had a theory for a bit and large enough jitter on usb bus could be cause of interface get buffer underrun or something, creating gaps like on first linked pic.
-
- KVRAF
- Topic Starter
- 7120 posts since 22 Jan, 2005 from Sweden
Asus AI Suite is removed since some time, thanks.
I found it unnerving having to exit a lot of stuff that could not be disabled through Startup. Even the more detailed Sysinternals Autostart did not fix it. Stuff for gamers again to throttle or go in and out of turbo by pressing a key or button.
Throttling is solved now thanks to Pictus suggestions.
- maybe the hysteresis for idle throttling had an impact setting to 100%
- Great suggestions and thanks again, Pictus
- one piece if the puzzle in place
It goes steady on 1.99 GHz as TaskManager say.
Which is consistent with 16 performance cores on 2.1 GHz and 4 efficient cores on 1.6 GHz(saw in bios this is what it runs on).
It's been really quiet with ticks now for some hours, but not taking victory in advance. Moody stuff computers....
I found it unnerving having to exit a lot of stuff that could not be disabled through Startup. Even the more detailed Sysinternals Autostart did not fix it. Stuff for gamers again to throttle or go in and out of turbo by pressing a key or button.
Throttling is solved now thanks to Pictus suggestions.
- maybe the hysteresis for idle throttling had an impact setting to 100%
- Great suggestions and thanks again, Pictus
- one piece if the puzzle in place
It goes steady on 1.99 GHz as TaskManager say.
Which is consistent with 16 performance cores on 2.1 GHz and 4 efficient cores on 1.6 GHz(saw in bios this is what it runs on).
It's been really quiet with ticks now for some hours, but not taking victory in advance. Moody stuff computers....
-
- KVRAF
- Topic Starter
- 7120 posts since 22 Jan, 2005 from Sweden
The default cpu speed was 4.5 GHz on my machine when everything in turbo and boost and stuff were still active.
To run constantly over 3 GHz seems a challenge for manufacturers. Even with all these voltage regulators(VRM) going dynamically up and down.
The gain of more cores might be at it's limit right now.
Looking at Process Lasso early after my pc was up and running the active cores change every second from everything to 20 active cores down to 2 cores, up to 12 cores and on it goes. All this dynamic thing happening seems to create audio issues and hard to keep an constant stream with low jitter with those maximizing features as turbo and stuff.
That's why I did go for i7-12700F cpu rather than the ones at 3.6 GHz. It still a leap from my old i7-860 from 2010. Memory at 3.2 GHz running XMP-1, 4 TB SSD and 20 core cpu is a very responsive machine.
Now with all cores running all the time, probably no idle states, utilization increase a bit. A project running 8% before now use 11% or so. And audio seem calm and collected as I want it. And fans don't go bezerk. CPU has not been over 37 Celsius.
To run constantly over 3 GHz seems a challenge for manufacturers. Even with all these voltage regulators(VRM) going dynamically up and down.
The gain of more cores might be at it's limit right now.
Looking at Process Lasso early after my pc was up and running the active cores change every second from everything to 20 active cores down to 2 cores, up to 12 cores and on it goes. All this dynamic thing happening seems to create audio issues and hard to keep an constant stream with low jitter with those maximizing features as turbo and stuff.
That's why I did go for i7-12700F cpu rather than the ones at 3.6 GHz. It still a leap from my old i7-860 from 2010. Memory at 3.2 GHz running XMP-1, 4 TB SSD and 20 core cpu is a very responsive machine.
Now with all cores running all the time, probably no idle states, utilization increase a bit. A project running 8% before now use 11% or so. And audio seem calm and collected as I want it. And fans don't go bezerk. CPU has not been over 37 Celsius.
-
- KVRian
- 1055 posts since 3 Jul, 2006
TLDR:
For a laptop, if your frequencies are crazy:
Limit the OS to 5 cores, in order to reduce current consumption from the CPU power supply.
Set the OS to always use min. 80% of the maximum turbo boost frequency.
This stabilized the frequencies of the cores, on my laptop.
----
I'll share my story here, about trying to optimize a mediocre laptop for Audio, in terms of processor power management.
So, I have an older low end laptop.
It's an Acer Aspire A515-55 laptop, which has a Ice-Lake-U, i5-1035G1 processor.
The base speed is about 1GHz, but can go up to 3.6GHz in turbo boost.
It has 4 physical cores, and due to hyper-threading can show 8 logical cores to Windows.
What I observed is that the core frequencies really vary a lot from 300 MHz to 3.6GHz, but quite chaotically. Switching the switching frequencies of the cores adds latency to the system.
The "realtime" cpu usage graph in popular DAWs fluctuates like hell, and it doesn't take long before there are buffer under-runs, and the sounds starts to break apart.
So then, what I wanted to do is try to make the CPU run at maximum frequencies all the time.
Frequency operation choices are given both by the OS, and the processor itself, responding to several sensors.
The most important sensor is the thermal sensor. Junction temperature inside the processor should not exceed 100C according to Intel's CPU datasheet, or it will wear out the processor very quickly.
What I found is that the laptop is configured in hardware to limit the frequency up to 90C.
This is a good thing and shouldn't be disabled. Or you may end up with a fried CPU.
But there are other hardware sensor limits that may throttle down the operating frequency:
- Dissipated power (if it too much)
- Current consumption (if it is too little per core). This prevents data corruption inside the CPU.
Firstly, a lot of the Windows processor power management parameters are hidden.
What is hidden and what is shown is stored in the registry.
To view all the available power management parameters and choose which to make visible you can use the PowerSettingsExplorer tool.
Secondly, you want to tell Windows to try to keep the frequencies up, even if the load is low.
This should, in principle, limit the changing of operating frequencies in the cores.
You can start from a high performance power profile in windows by making a copy of the hidden
Ultimate Performance profile and start from there (using powercfg).
But even with the ultimate power profile, I observed that the frequencies still fluctuate.
Thirdly, there are programs that can change the caps for the power dissipation (short term and
long term). I used ThrottleStop to try to modify those.I put higher limiting values, but that wasn't enough.
What is interesting about ThrottleStop, is that if your CPU hit a limitation factor, and therefore
reduced its operating frequency. If you click the Limits button, you will see in realtime what
triggered a frequency reduction.
The limiting factors are:
- THERMAL: if you see it most likely means that you are in thermal throttling. The frequency is
reduced so as not to fry the CPU. Do NOT disable this.
- PL1: Long Power Limit
- PL2: Short Power Limit
- EDP other: mainly current limit!!
- BD PROCHOT: I believe also related to thermal throttling. Do NOT disable this!!
In my case, I was not thermal throttling, but I kept seeing the EDP (Electrical Design Point).
My conclusion was that the power supply of my laptop could not deliver enough current for all
the cores to operate continuously at their maximum turbo boost frequency.
I couldn't disable hyperthreading in my laptop, as this doesn't show up in the BIOS.
What I tried to use to reduce the current consumption was to try to turn off one of my cores.
You can use msconfig > boot > advanced options to tell windows how many cores to use at boot time.
I thought it would be nice if I disable hyperthreading by telling windows to limit itself to 4 cores.
But if choose 4 cores, Windows will use 2 physical cores to end up with 4 logical cores.
So then I tried 5 cores at boot. This means two physical cores will account for 4 logical cores
and the third physical core will not be hyperthreaded.
Even with 3 cores, it could not supply them continuously at 3.6GHz.
So I reduced the frequency by using a parameter available in Throttle stop called EPP.
This is related to the Speed Shift feature of the processor (SST).
EPP is 0 to 255. 0 means it tries to keep maximum frequency even at no load, while higher values
will reduce the idle frequency.
I still noticed frequency changing a lot with EPP set to zero, but then everything stabilized when I set this to 4. When I set EPP to 4, the idle frequency was a constant 3GHz, with a stable frequency (90% of the time).
I read that if you use SST to adjust the frequency you can disable Speed Stop (EIST), and that may result in a more stable frequency operation.
I also disabled C1E, which as far as I understand is a power-saving feature.
So, in summary, I found out that the power supply can continuously provide enough current to the CPU to keep 3 out of 4 physical cores at 3GHz (about 80% of the max. frequency) with (mostly) constant frequency.
Testing in my DAW, the realtime CPU usage is much more stable, and I can generally run a lot more plugins without getting Audio glitches. And 5 logical cores is at the moment more than enough for me.
BTW: when operating with EPP zero, at 3.6GHz, the CPU reaches thermal throttling really fast.
With EPP at 4 (3 GHz) the CPU cores stay at roughly 50-55C when I'm using my DAW.
The tools that I used:
- HWiNFO
- HWMonitor
- PowerSettingsExplorer
- ThrottleStop
- MSconfig
- powercfg
Hope this might help others!
For a laptop, if your frequencies are crazy:
Limit the OS to 5 cores, in order to reduce current consumption from the CPU power supply.
Set the OS to always use min. 80% of the maximum turbo boost frequency.
This stabilized the frequencies of the cores, on my laptop.
----
I'll share my story here, about trying to optimize a mediocre laptop for Audio, in terms of processor power management.
So, I have an older low end laptop.
It's an Acer Aspire A515-55 laptop, which has a Ice-Lake-U, i5-1035G1 processor.
The base speed is about 1GHz, but can go up to 3.6GHz in turbo boost.
It has 4 physical cores, and due to hyper-threading can show 8 logical cores to Windows.
What I observed is that the core frequencies really vary a lot from 300 MHz to 3.6GHz, but quite chaotically. Switching the switching frequencies of the cores adds latency to the system.
The "realtime" cpu usage graph in popular DAWs fluctuates like hell, and it doesn't take long before there are buffer under-runs, and the sounds starts to break apart.
So then, what I wanted to do is try to make the CPU run at maximum frequencies all the time.
Frequency operation choices are given both by the OS, and the processor itself, responding to several sensors.
The most important sensor is the thermal sensor. Junction temperature inside the processor should not exceed 100C according to Intel's CPU datasheet, or it will wear out the processor very quickly.
What I found is that the laptop is configured in hardware to limit the frequency up to 90C.
This is a good thing and shouldn't be disabled. Or you may end up with a fried CPU.
But there are other hardware sensor limits that may throttle down the operating frequency:
- Dissipated power (if it too much)
- Current consumption (if it is too little per core). This prevents data corruption inside the CPU.
Firstly, a lot of the Windows processor power management parameters are hidden.
What is hidden and what is shown is stored in the registry.
To view all the available power management parameters and choose which to make visible you can use the PowerSettingsExplorer tool.
Secondly, you want to tell Windows to try to keep the frequencies up, even if the load is low.
This should, in principle, limit the changing of operating frequencies in the cores.
You can start from a high performance power profile in windows by making a copy of the hidden
Ultimate Performance profile and start from there (using powercfg).
But even with the ultimate power profile, I observed that the frequencies still fluctuate.
Thirdly, there are programs that can change the caps for the power dissipation (short term and
long term). I used ThrottleStop to try to modify those.I put higher limiting values, but that wasn't enough.
What is interesting about ThrottleStop, is that if your CPU hit a limitation factor, and therefore
reduced its operating frequency. If you click the Limits button, you will see in realtime what
triggered a frequency reduction.
The limiting factors are:
- THERMAL: if you see it most likely means that you are in thermal throttling. The frequency is
reduced so as not to fry the CPU. Do NOT disable this.
- PL1: Long Power Limit
- PL2: Short Power Limit
- EDP other: mainly current limit!!
- BD PROCHOT: I believe also related to thermal throttling. Do NOT disable this!!
In my case, I was not thermal throttling, but I kept seeing the EDP (Electrical Design Point).
My conclusion was that the power supply of my laptop could not deliver enough current for all
the cores to operate continuously at their maximum turbo boost frequency.
I couldn't disable hyperthreading in my laptop, as this doesn't show up in the BIOS.
What I tried to use to reduce the current consumption was to try to turn off one of my cores.
You can use msconfig > boot > advanced options to tell windows how many cores to use at boot time.
I thought it would be nice if I disable hyperthreading by telling windows to limit itself to 4 cores.
But if choose 4 cores, Windows will use 2 physical cores to end up with 4 logical cores.
So then I tried 5 cores at boot. This means two physical cores will account for 4 logical cores
and the third physical core will not be hyperthreaded.
Even with 3 cores, it could not supply them continuously at 3.6GHz.
So I reduced the frequency by using a parameter available in Throttle stop called EPP.
This is related to the Speed Shift feature of the processor (SST).
EPP is 0 to 255. 0 means it tries to keep maximum frequency even at no load, while higher values
will reduce the idle frequency.
I still noticed frequency changing a lot with EPP set to zero, but then everything stabilized when I set this to 4. When I set EPP to 4, the idle frequency was a constant 3GHz, with a stable frequency (90% of the time).
I read that if you use SST to adjust the frequency you can disable Speed Stop (EIST), and that may result in a more stable frequency operation.
I also disabled C1E, which as far as I understand is a power-saving feature.
So, in summary, I found out that the power supply can continuously provide enough current to the CPU to keep 3 out of 4 physical cores at 3GHz (about 80% of the max. frequency) with (mostly) constant frequency.
Testing in my DAW, the realtime CPU usage is much more stable, and I can generally run a lot more plugins without getting Audio glitches. And 5 logical cores is at the moment more than enough for me.
BTW: when operating with EPP zero, at 3.6GHz, the CPU reaches thermal throttling really fast.
With EPP at 4 (3 GHz) the CPU cores stay at roughly 50-55C when I'm using my DAW.
The tools that I used:
- HWiNFO
- HWMonitor
- PowerSettingsExplorer
- ThrottleStop
- MSconfig
- powercfg
Hope this might help others!
== VDX == One Man can make a difference!
My music is on https://soundcloud.com/vdxi | Info | More Info
My music is on https://soundcloud.com/vdxi | Info | More Info
-
- KVRAF
- Topic Starter
- 7120 posts since 22 Jan, 2005 from Sweden
Thanks.
Reason to get to fixed frequency was that I had a theory that it might induce jitter somehow when audio is transferred to audio interface over usb. Succeding this did not fix the little faint ticks that appear sometimes, and just in one ear most of the time. Still a mystery.
I did test yesterday and set the core parking on again, changing value from 100% used to default 4%. All audio suddenly sounded completelty trashed, like massive overs. And on Process Lasso I could see how cores were started and changed around like crazy every second that Lasso polled. Not sure why thread allocation would jump around like that.
- before I found that hidden setting for core parking it did not sound that bad
- so might be a setting in bios I used that control more and os less
- but changing the idle demote/promote probably also affect if cores step by step is parked or just shut down
Overall theory is that windows is doing certain maintenance or similar different days in different ways that affect audio somehow. Just minutes after booting and starting audio application you have like a 1s stall of audio when windows throws things to do things on internet that is not up, not even network is up. Windows is menace in how bad scheduled jobs behave, not checking if internet is up first before throwing calls to network. I had such issues with applications that had autoupdate stuff that did such too, on windows 7, and finding those and turning that off fixed things. These things were enough to have Cubase Pro restart audio engine and if recording that was aborted too.
I ordered a Presonus Studio 1824c to come this week and will test that if different. Presonus seems keen to work on low latency. Seems incredibly good value as well.
Reason to get to fixed frequency was that I had a theory that it might induce jitter somehow when audio is transferred to audio interface over usb. Succeding this did not fix the little faint ticks that appear sometimes, and just in one ear most of the time. Still a mystery.
I did test yesterday and set the core parking on again, changing value from 100% used to default 4%. All audio suddenly sounded completelty trashed, like massive overs. And on Process Lasso I could see how cores were started and changed around like crazy every second that Lasso polled. Not sure why thread allocation would jump around like that.
- before I found that hidden setting for core parking it did not sound that bad
- so might be a setting in bios I used that control more and os less
- but changing the idle demote/promote probably also affect if cores step by step is parked or just shut down
Overall theory is that windows is doing certain maintenance or similar different days in different ways that affect audio somehow. Just minutes after booting and starting audio application you have like a 1s stall of audio when windows throws things to do things on internet that is not up, not even network is up. Windows is menace in how bad scheduled jobs behave, not checking if internet is up first before throwing calls to network. I had such issues with applications that had autoupdate stuff that did such too, on windows 7, and finding those and turning that off fixed things. These things were enough to have Cubase Pro restart audio engine and if recording that was aborted too.
I ordered a Presonus Studio 1824c to come this week and will test that if different. Presonus seems keen to work on low latency. Seems incredibly good value as well.