1 Rookie

 • 

22 Posts

58834

July 8th, 2021 09:00

isHighlyActiveTopic

Latitude 5420/7420/7520 CPU Throttling Issue on Linux

Hi there!

I have Latitude 7520 laptop with i7-1185g7 and this model has a huge problem with CPU throttling.

Besides this model users report that 5420/7420 models affected too.

This laptops have a two issues:

  1. Throttling to 400mhz after running power-intensive workloads for a short amount of time -- which caused by incorrect work of legacy interface. This issue can be resolved by blacklisting "intel_rapl_msr" module or install intel/thermald v2.4.6
  2. The same issue but cpu stuck on 1800mhz -- it is another issue and currently I haven't any idea how resolve it. Looks like something from firmware locks CPU power.

Message from Intel developer(pls see here) :

The problem is that TMEM sensor reaches its limits of 42C in 4 seconds, so the system is throttled from max power. Even at the start the temperature is 39C. So not much margin. Not
sure what can be done here

So I am trying to find and solve this problem. Maybe someone from Dell can help. A lot of users buy Latitude laptops for work with Linux and as I can see Dell sells linux version of this laptops.

What I see currently:

  • In Windows it works correctly(3.3-4.8ghz on max load)
  • Latest mainline kernel(`5.13.0-1-mainline`) doesn't fix it
  • Latest bios doesn't fix it(1.7.1 for my laptop)
  • It depends on laptop vendor. I have MSI laptop with the same i7-1185g7 and it works properly, one difference is: Dell show "base_frequency" as 1800mhz, MSI as 3000mhz. Maybe it's related
  • Load/Unload modules, disabling intel_pstate, tune tdp limits, changing thermal mode(quiet,cool,balanced,performance) doesn't affect this problem

Related links:

I will be very appreciated if someone help me to resolve this issue. Thanks!

1 Rookie

 • 

7 Posts

December 8th, 2021 15:00

I tried to open a ticket for this with Dell. What a waste of time.

As far as I understood things:

  • There is no Linux team support at Dell (at least for this model)
  • All I got was a link to the Ubuntu forum homepage
  • There is no agreement between Dell and Canonical/Ubuntu to provide support
  • Noone at Dell can help me further (with a Linux only system)

So even though their product page claims that my "Latitude 7420 supports" Ubuntu, there is no support from Dell itself. https://www.dell.com/support/home/en-nz/drivers/supportedos/latitude-14-7420-2-in-1-laptop?ref=oscompatibility

Does anyone of you have their Windows partition still intact? Maybe try pinging the Dell support with this. They need to 'check some things' and apparently they can only proceed with Windows. @SebastianH I think I read on the thermald issue tracker that you had opened a ticket? How did that work?

If you just bought a Latitude 5/7xxx for work with Linux: just return it if you can.

Who knows when or if this will ever get fixed.

1 Rookie

 • 

33 Posts

December 8th, 2021 17:00

If I had to "assign blame" or "ask the right people to fix the problem" I'd probably start with the folks who maintain thermald, or perhaps Canonical (if they have custom patches for thermald).  I know that Ubuntu is the "bees knees" but for this particular issue the solution may be to switch to a RedHat derived distro.  I've been using Rocky 8 on my 5420 and yes it has been glitchy every once in a while, but overall it is a solid performer.  (I find it strange that Dell used to offer RedHat Enterprise Linux across a lot of their Latitude and Precision line and now they don't. RedHat would offer hardware support, as far as creating patches for supported software to work around hardware defects.)

I know it's personal preference but I think it's a stretch to recommend Linux users "just return it if you can".  As others have pointed out, this is not just a Dell issue- there are Lenovo laptops with the same chipset that have the same issue.  

35 Posts

December 9th, 2021 00:00

I installed Windows on a USB stick and with that the temperature rises up to 100°C and after some time goes down to 70°-80°C and throttles down the CPU clock. But not as heavily as with Linux. On Linux it's stuck to 1800MHz and 50°C after seconds.

As this is a business notebook for me and I have "Pro Support" I contacted Dell and described the problem. The first offered solution was to change the electronics which I denies because it makes no sense. Then I was told that "The Linux on the Internet does not support the notebook" and I replied that I would like to get the "official supported Linux" from Dell. Last answer was that some technical person wants to contact me which I am waiting for.

On a daily basis I currently can work with 1800MHz but I selected this notebook to compile and run CPU heavy things. I know that there are XPS out there but I like the modular building of the Latitudes and my last Latitude was very solid over the last 3 years and performed well.

No I am hoping that I will get contact to someone who really can change something.

35 Posts

December 13th, 2021 04:00

Unfortunately the latest reply was that the notebook is not supported because it has been delivered with Windows. I replied in a long mail that the manual and homepage states that it should be supported nevertheless. Wish me luck, I will share all results here and on Github.

December 13th, 2021 04:00


Very interesting Sebastian, specially the fact that someone from Dell is going to contact you to know more about the problem.

I'm exactly in your same situation: I can more or less work using the workaround with the 1800 limitation, but sometimes I need to perform intensive CPU tasks and this is just not acceptable.

My laptop is supposed to be certified by Dell to work with Ubuntu:

https://ubuntu.com/certified/202010-28335
But it turns out that actually it's not.

This it's not a cheap computer, so I'm afraid that if someone from Dell cannot confirm that somebody is taking care of this we would have to return the laptop and look for a more compatible one.

Please if someone from Dell contact you about this share it with us

Thanks !

1 Rookie

 • 

7 Posts

December 13th, 2021 14:00

I said the same thing - with no success. See if you can escalate the request higher? There does not seem to be a proper Linux team, so just "higher", I guess.

(I was not able to, but I assumed it was because I was using the WhatsApp channel)

Cheers

35 Posts

December 17th, 2021 06:00

I just had a zoom meeting with a very nice guy from the Dell support (thanks again for this meeting!) and I explained the situation and demonstrated the problem sharing my s-tui output while stress testing.

As I bought my notebook with Windows (there is no option for Ubuntu here in Germany) he pointed out that this is officially no case for the pro support, which I can fully understand.

Nevertheless, he promised to dig deeper and forward the findings to colleagues. Hopefully someone at Dell will pick up this case and take a technical look at it.

At this point I fear the the only option is to wait for a BIOS update fixing the problem or some "hack" that will release the power of the "tiger" (lake) on our Latitudes.

35 Posts

January 17th, 2022 11:00

Updated BIOS and thermald to latest version: Nothing has changed yet.

January 18th, 2022 00:00

Thanks so much for the update Sebastian

I'm afraid I'm going to give up on this and I will ask for a replacement of my laptop now that it's still on warranty. 

January 18th, 2022 08:00

Update: Interestingly enough, immediately after posting that I'm giving up with this laptop I was contacted by Dell Support, what I really appreciate, but I think that the problem is really well explained in this issue, so I would suggest to read the description of this issue instead of contacting user by user.

Said that, they told me that they will redirect the issue to the product team so I hope they push for a solution

1 Rookie

 • 

21 Posts

January 18th, 2022 13:00

I'm having the same problem, in a Latitude 5420 with an i5-1135G7. Bought it for work, as a software developer, with 32GB of RAM, and it throttles badly everyday, but to 1400Mhz. Tried all the possible combinations on the BIOS and no dice.

After a BIOS update some time ago, it began throttling to 400Mhz instead of 1400, which was already unacceptable. I had to reinstall my distro (Ubuntu 20.04 LTS) from scratch.

I'm also monitoring this topic, and I'm opening a complaint about it. I reckon I'll have to return it as well if this is not solved. It's a really expensive laptop, and it's currently being outclassed by an old Xeon running with DDR3 (a secondary machine I have), because the Xeon does not throttle.

1 Rookie

 • 

33 Posts

January 18th, 2022 14:00

For those who are calling Dell support and being told "This shipped with Windows so we won't support you" I would push back on that and demand escalation- especially if you have ProSupport.  Ubuntu 20.04 is a supported configuration on these laptops.  Period.  End of statement.  They need to support it!  If I buy a Dell laptop and then I buy a 16GB memory module from Dell, and then the laptop fails a month after I install the module... Dell will not refuse support just because "the laptop wasn't shipped with that 16GB module already installed".

The bad thing about this is that Dell has enough "fleet sales" of these laptops that they can afford to not care about individual users when it comes to actually solving a problem that requires more than just replacing parts.

1 Rookie

 • 

33 Posts

January 18th, 2022 16:00

I think I found the problem here.

To note, I have the same configuration as @JoseMachado : a 5420 with the i5-1135G7, 32GB of RAM, and 1T NVME.

Using Rocky Linux 8.5.  I grabbed the source for the latest release of thermald: 2.4.8 and used it to build an RPM package.  Installed the RPM package.  Then I ran pixz on a 120GB qemu image, which maxed out all of the CPU cores.  Something very interesting happened: the processor hit a full 4GHz and remained there for a bit.  I've never been able to get it to hit 4GHz until now.

I did this test on two separate occasions.  On the first occasion, the ambient temperature in the room was 80 degrees F.  (What can I say, I like my rooms toasty ).  In this case, the CPU hit 4GHz, remained there for about 3 minutes, then dropped to 1.5GHz.  It remained at 1.5 for about 40 seconds, then it jumped back up to 4GHz.  It sat there for about 10 seconds, then went down to 1.8GHz and fluctuated between 1.5 and 2.7GHz.

On the second occasion the room was 70 degrees F.  The progression was somewhat different: it hit 4GHz, remained there for 3 minutes, then dropped to 1.5.  Remained at 1.5 for 40 seconds.  Then jumped back up to 4GHz for 80 seconds... then dropped to 1.5.  It repeated this cycle for 15 minutes until I stopped the test.

A couple of conclusions can be made here.  First, the cooling capability of the 5420 is not very strong at higher ambient temperatures.  Second, this is a thermald problem, not a hardware problem.  I notice the current version of thermald in Ubuntu 20.04 repository is 1.9.1 with some Ubuntu customizations.  The stock thermald that comes with RHEL8 clones is 2.4.1, and it had the same or a similar problem as the Ubuntu version.  The thermald version 2.4.8 does not appear to have this problem.

Dell may be able to fix this by including their own thermald package version 2.4.8 that would replace the stock Ubuntu thermald package.  Developers or power users may be able to get around this problem by using a RHEL8 derivative instead of Ubuntu; I can make the RPM (or RPM SPEC file so you can build your own RPM) available to anyone who wants it.

Shown below is the ThermalMonitor program that can tie in with thermald to fine-tune the setpoints and monitor what's going on.  The blue peaks in this graph were the CPU hitting 4GHz.

Screenshot from 2022-01-18 19-12-25.png

35 Posts

January 19th, 2022 00:00

If you are falling back to 400MHz that's most probably a problem by a not running or not properly configured thermald.

For me it does not work when thermald is started with systemd during startup, I always have to start thermald after complete bootup to make it work. Also you have to use the following thermald configuration: /usr/sbin/thermald --systemd --dbus-enable --adaptive

Important is the "--adaptive" the two others are only needed when invoked via systemd. You can test is by stopping the thermald and then starting by hand: /usr/sbin/thermald --adaptive --no-daemon

Can you try this?

35 Posts

January 19th, 2022 00:00

But even the thermal capabilities are not good, the CPU should hit ~95°C before throttling down. This is also the way it behaves with Windows. So I still insist a software/BIOS problem.

No Events found!

Top