Intel® NUCs
Support for Intel® NUC products
11772 Discussions

NUC11PAHi7: The fan stops during video transcoding / CPU overheats

Geeooff
New Contributor I
1,913 Views

Hey guys.

I'm owner of a very brand new Intel NUC 11 Performance Kit.

I'm using it for many small development projects and for entertainment.
It's so using Ubuntu Server 20.10 with QEMU-KVM and Docker.

I tried to install also Plex Media Server, to see the capabilities of the Intel Xe iGPU to, maybe, replace my Nvidia Shield (only the server-side).

I have quite a few difficulties to install the proper OpenCL and VA drivers/API for this iGPU, but this is not the main topic.

I notice that during hardware video transcoding the fan just stops and won't restart!
After a few minutes, the CPU reach 95 degrees Celsius and will thermal throttle. The entire case becomes hot and I can barely touch it.

It's very concerning... I tried to reboot (remotely) the NUC to restart the fan and because a temperature alert displayed the NUC won't boot. I'm using it remotely, and even If I set up the "no POST warnings" in the BIOS it will be stuck there until I reboot the NUC again (so... I suppose another BIOS bug/feature).

Of course I checked the fan profile: I disabled the fanless capability, put lower minimal temperature than the "Cool" profile, but this seems to works on Windows for an entire day, but not for Linux.

Any ideas ?

 
0 Kudos
10 Replies
Alberto_Sykes
Employee
1,878 Views

Geeooff, Thank you for posting in the Intel® Communities Support.


You mentioned "works on Windows for an entire day, but not for Linux", just to let you know, Intel® has tested and validated the Intel® NUC 11 Performance kit - NUC11PAHi7 using Windows* as the Operating System, all the drivers that we have available in our web site will work only with Windows* as you can confirm in the links below, so, some problem might arise when using a different OS, in this case, Linux:

https://www.intel.com/content/www/us/en/support/articles/000005628/intel-nuc.html

https://downloadcenter.intel.com/product/205073/Intel-NUC-11-Performance-kit-NUC11PAHi7


In reference to this scenario, in the following link, you will find some suggestions or troubleshooting steps recommended to try when the Intel® NUC is getting overheated, some of them you already tried:

https://www.intel.com/content/www/us/en/support/articles/000033327/intel-nuc.html


If the problem persists after that, then please attach the SSU report so we can verify further details about the components in your platform, please check all the options in the report including the one that says "3rd party software logs":

https://downloadcenter.intel.com/download/26735/


Any questions, please let me know.


Regards,

Albert R.


Intel Customer Support Technician


Geeooff
New Contributor I
1,874 Views

Thank you for your answer.

I already applied the usual solutions:

  1. the BIOS was the latest one, applied on day one.
  2. the fan profile was set to a conservative (means "noisy") profile as safety measure

I can understand the Linux Support is in early state for 11th gen, that's why I use Ubuntu 20.10 because it should contain a kernel that knows this architecture, and especially the Xe iGPU.

Nevertheless, I'm using an Intel non-free package for Video Acceleration, so no third party:
Package binaries: Ubuntu – Details of package intel-media-va-driver-non-free in groovy
Source code: intel/media-driver (github.com)

I'm just surprise that software (even loaded at kernel level) can affect the CPU fan profile. The fan profile should be managed at BIOS / CPU microcode, not at software level, whatever the OS. Right ? (I'm not blaming anyone, I'm no driver developer).

Anyway, I will try to gather more infos to transfer this to your engineers, and will wait for new BIOS releases.

 
Geeooff
New Contributor I
1,860 Views

@Alberto_Sykes 

I made one more try with Plex.
I forced the transcoding of an UHD video to 720p while I was watching the temperatures using sensors, htop, and intel_gpu_top.

In the first moments, Plex was reporting hardware transcoding, and intel_gpu_top reported Video processing activity as well.
Then after a few minutes it seems the hardware transcoding dropped, replaced by software transcoding with no reason (same video).

The cores temperature raised to 90-100 degrees. The fan seems to rotate, but it stayed at minimum speed all the time. I think the difference with my previous try is I disabled the "fanless" option in the BIOS this time (duty cycle at 30% minimum).
So "something" forced the fan to go to its minimum speed?

You will find attached some console shots and the SSU reports.

 
 
Alberto_Sykes
Employee
1,847 Views

Geeooff, Thank you very much for providing that information and the reports.


In order for us to answer your questions and to try to fix this problem, we will do further research on this matter, as soon as I get any updates I will post all the details on this thread.


Regards,

Albert R.


Intel Customer Support Technician


Geeooff
New Contributor I
1,828 Views

Hello @Alberto_Sykes 

Please let me add another report to the previous one.

I booted up the NUC 11 on Ubuntu Server 20.04 LTS that I just installed on the second drive.
I wanted to see if with Intel-provided packages (only for 20.04) it will work better.

To have the best hardware support I could have, I installed:

  1. the Ubuntu HWE kernel (5.8 based)
  2. the Ubuntu OEM 20.04b kernel (5.10 based, the one I used for the tests)

I then Installed a few packages:

  1. Intel GPGPU packages to support Video Acceleration and OpenCL, from Intel repository
  2. beignet-opencl-icd, required for Plex to do HDR to SDR tone mapping
  3. Plex Media Server, from Plex repository
  4. intel-gpu-tools to watch the iGPU usage with intel_gpu_top
  5. several drm-related packages as dependencies of previous packages
  6. lm-sensors to watch the temperatures

Plex reported Hardware Transcoding support.

I tried to play a simple 1080p/H.264 video and try to transcode it to 720p/H.264 (through Plex).

After a few seconds the play dropped and Plex crashed.
I had barely the time to see activity in intel_gpu_top.

No problem with the fan this time, but dmesg looks ugly...

You will find attached the SSU reports for this new 20.04 LTS setup.

I hope it will help too.

 
Ronny_G_Intel
Community Manager
1,701 Views

Hi Geeooff,

I have been looking into this report but unfortunately our support for Linux* is very limited on this platform since we have only tested and validated Windows 10*, see this URL for Operating System Compatibility reference information https://www.intel.com/content/www/us/en/support/articles/000005628/intel-nuc.html

Also, keep in mind that we dont provide specific drivers for Linux, the drivers should be provided by the Linux distribution so I would recommend that you reach to the Ubuntu* community for support on this issue.  

 

Additional notes: 

- I couldn't find any reference compatibility information for 11th Generation Processors on the Ubuntu website.

- I checked the SSU attached and I am not 100% sure the system is properly recognizing Intel Iris Xe Graphics.

 

Regards,

Ronny G

Geeooff
New Contributor I
1,697 Views

Thank you again for your answers.

I admit my post complaining about two distinct things could confuse people:

  1. about fan control under linux
  2. about hardware accelerated media decoding/encoding under linux

The first point is still valid, I still think the fan control must be decoupled from the OS and even with some privileged access (from Windows or Linux drivers) the BIOS should ultimately be the last reliable tier to maintain fan control over dangerous temperatures. If I set up a custom but stricter fan profile than the "Cool" one, I'm expecting more aggressive fan profile, not fan shutdown under load.

The second point is concerning but could be part of a second post. Developers, users (and myself) of Plex Media Server are really interested to make a NUC the perfect consumer-grade server for this great piece of software. I will try to see to which project website I should refer to for those issues, but it's not clear right now. I would be pleased to find the right entry point (Intel owns many GitHub projects, Ubuntu PPAs, and other kernel-related projects too, which one should be used for support?).
 

 
Ronny_G_Intel
Community Manager
1,524 Views

Hi Geeooff,

I am not very knowledgeable about how the Operating System manages and combines the cooling options configured in BIOS with its own cooling parameters. My understanding is that Linux takes full responsibility of power management and cooling by taking that away from the BIOS while Windows based on Operating Systems follows the BIOS settings.

Since this issue is only happening when using Linux, I would recommend checking with the Linux Support Community.

 

Regards,

Ronny G

 

n_scott_pearson
Super User Retired Employee
1,510 Views

In desktop designs (like the NUCs), fan speed control is in the hands of the NUC's EC and implemented completely independent of what is happening in the runtime environment (other than the affects on temperature of what is happening within the runtime environment). The ACPI specification states that, if the fan is being actively (and independently) controlled (as it is in all NUCs), you cannot make use of any ACPI- and fan-based passive cooling techniques. Consequently, there are no fan speed control methods implemented within the the BIOS' ACPI tables and nothing is ever done by the O/S in that respect.

In this case, it is how the runtime environment is using the processor - and its affect on processor temperatures - that is responsible for the differences between the temperatures seen and the fan speed control implemented. Now, there are definitely differences in how Linux and Windows use the processor - whether configured for balanced vs. performance operation, for example - and this could result in differences in the temperatures seen.

Hope this helps,

...S

SCrom2
Beginner
1,435 Views

I'm having this issue in combination with my plex media server too. Running linux 18.04.

Once a transcode remotely starts, the temp flies into the sky until it powers the nuc down and I need to wait for it to cool down.

 

Do I really need to go back to Windows to stop this issue from happening?

Reply