[SOLVED] [split] Missing HPET clocksource option

[hkphantomgtr]

FYI, "hpet" is gone in Clocksource since FW 1.2.4.

[agent_kith]

FYI, "hpet" is gone in Clocksource since FW 1.2.4.

Do you mean you have HPET enabled in the BIOS, but can't find that setting in the kernel config any more?

I havn't touched that part of the code, but will check and confirm later today (or in the week)

[hkphantomgtr]

FYI, "hpet" is gone in Clocksource since FW 1.2.4.

Do you mean you have HPET enabled in the BIOS, but can't find that setting in the kernel config any more?

I havn't touched that part of the code, but will check and confirm later today (or in the week)

Yes, now I have my NUC.  In the CMOS there is an option to turn on/off "High Precision Event Timer".  Your help will be highly appreciated, again. 

[agent_kith]

Yes, now I have my NUC.  In the CMOS there is an option to turn on/off "High Precision Event Timer".  Your help will be highly appreciated, again. 

Can you run the following command, and tell me what it says? Also generate a diagnostics file and send it to the specified email so I can check the logs

Code:

cat /sys/devices/system/clocksource/clocksource0/available_clocksource

[hkphantomgtr]

Yes, now I have my NUC.  In the CMOS there is an option to turn on/off "High Precision Event Timer".  Your help will be highly appreciated, again. 

Can you run the following command, and tell me what it says? Also generate a diagnostics file and send it to the specified email so I can check the logs
 

Code:

cat /sys/devices/system/clocksource/clocksource0/available_clocksource

tsc acpi_pm

And I've emailed the diagnostic file to the specific email a/c.  Please take your time. 

[agent_kith]

tsc acpi_pm

So it looks like you only have the above two as your clock source. I believe TSC is pretty good as well, so that's worth a try

And I've emailed the diagnostic file to the specific email a/c.  Please take your time. 

See this in the system logs:

Code:

hpet: HPET dysfunctional in PC10. Force disabled.

Look in your ACPI or power saving and see if you can disable PC10 state, play around with CPU states and see if you can disable just that, failing that you might need to disable power saving entirely Doing so should get HPET back after a reboot. Let me know how it goes.

[hkphantomgtr]

tsc acpi_pm

So it looks like you only have the above two as your clock source. I believe TSC is pretty good as well, so that's worth a try

And I've emailed the diagnostic file to the specific email a/c.  Please take your time. 

See this in the system logs:
 

Code:

hpet: HPET dysfunctional in PC10. Force disabled.

Look in your ACPI or power saving and see if you can disable PC10 state, play around with CPU states and see if you can disable just that, failing that you might need to disable power saving entirely Doing so should get HPET back after a reboot. Let me know how it goes.

Thanks, AK!  You're right.  After disabling the power control setting (disabling "Dynamic Power Technology" , or mainly "OS ACPI C2 Report"), my HPET is back, at the expense of lower temperature, and the low latency value.  I really like the HPET sound as the sound is more vivid, with better focus, much better transition, more "on-beat", and denser texture.   In short, more holographic.

[hkphantomgtr]

tsc acpi_pm

So it looks like you only have the above two as your clock source. I believe TSC is pretty good as well, so that's worth a try

And I've emailed the diagnostic file to the specific email a/c.  Please take your time. 

See this in the system logs:
 

Code:

hpet: HPET dysfunctional in PC10. Force disabled.

Look in your ACPI or power saving and see if you can disable PC10 state, play around with CPU states and see if you can disable just that, failing that you might need to disable power saving entirely Doing so should get HPET back after a reboot. Let me know how it goes.

Thanks, AK!  You're right.  After disabling the power control setting (disabling "Dynamic Power Technology" , or mainly "OS ACPI C2 Report"), my HPET is back, at the expense of lower temperature, and the low latency value.  I really like the HPET sound as the sound is more vivid, with better focus, much better transition, more "on-beat", and denser texture.   In short, more holographic.

To further my findings, with the same CMOS setting, after changing to "TSC" clocksource, the latency becomes as good as before.  huh.... so, with my new NUC, seems the CMOS setting doesn't affect the latency, but the choice of "clocksource".

And last night, in a second thought, it's quite strange to me that, even in the default CMOS setting, including HPET is enabled, the Snakeoil OS can't use HPET, or the system/hardware can't use HPET with the CMOS default power management setting...... so where is the problem?

[agent_kith]

To further my findings, with the same CMOS setting, after changing to "TSC" clocksource, the latency becomes as good as before.  huh.... so, with my new NUC, seems the CMOS setting doesn't affect the latency, but the choice of "clocksource".

On some motherboards, I honestly can't discern any notable difference between HPET and TSC. IIRC need to disable hyper-threading.

I believe one is the older technology, and less "accurate" under certain conditions, and the other is meant to be a replacement. If memory serves, HPET is meant to be a replacement for TSC (could be the other way round though ).

And last night, in a second thought, it's quite strange to me that, even in the default CMOS setting, including HPET is enabled, the Snakeoil OS can't use HPET, or the system/hardware can't use HPET with the CMOS default power management setting...... so where is the problem?

HPET is disabled in the kernel. There's apparently some problems with HPET when in C10 mode (power saving mode), so the kernel disables the HPET timer (and falls back to TSC IIRC). So even though HPET is enabled in the system BIOS, the OS will refuse to use it.

If you're using custom kernels, power saving is disabled by default so HPET is always working. Using stock linux kernels, full suite of power saving is enabled hence you may get that missing HPET.

[hkphantomgtr]

To further my findings, with the same CMOS setting, after changing to "TSC" clocksource, the latency becomes as good as before.  huh.... so, with my new NUC, seems the CMOS setting doesn't affect the latency, but the choice of "clocksource".

On some motherboards, I honestly can't discern any notable difference between HPET and TSC. IIRC need to disable hyper-threading.

I believe one is the older technology, and less "accurate" under certain conditions, and the other is meant to be a replacement. If memory serves, HPET is meant to be a replacement for TSC (could be the other way round though ).

And last night, in a second thought, it's quite strange to me that, even in the default CMOS setting, including HPET is enabled, the Snakeoil OS can't use HPET, or the system/hardware can't use HPET with the CMOS default power management setting...... so where is the problem?

HPET is disabled in the kernel. There's apparently some problems with HPET when in C10 mode (power saving mode), so the kernel disables the HPET timer (and falls back to TSC IIRC). So even though HPET is enabled in the system BIOS, the OS will refuse to use it.

If you're using custom kernels, power saving is disabled by default so HPET is always working. Using stock linux kernels, full suite of power saving is enabled hence you may get that missing HPET.

O... I c.  Thanks for your explanation.  Now I understand more.
BTW, I've already disabled hyper-threading.  Have a nice day!