HP Elitebook 865 G10 w/ AMD Ryzen 9 PRO 7940HS: cannot sustain a 35W load

[darkbasic]

Hi,

The Ryzen 9 PRO 7940HS is being sold with a 35-54W configurable TDP. I didn't expect it to be able to sustain the full 54W, but the baseline 35W is bare the minimum to make the upgrade worth over the cheaper AMD Ryzen 7 PRO 7840U (which is 15-30W). I bought the bulkier 16" over the 14" mostly because of thermals: I wanted to be sure that it had enough room to breath without overheating. I'm a full stack software developer and I use my laptop mostly docked doing heavy software compilations, but occasionally I also make sustained use of its iGPU (games, compute...). It's the latter which is more problematic: you rarely notice a 30% mid compilation slowdown (unless you're compiling something super heavy like Chromium which translates into hours of additional waiting), but it's a night and day difference in games where 30% makes the difference between playable and unplayable. Also the heat buildup is much more steep with CPU+GPU combined loads to the point that sometimes the thermal protection shuts the laptop down before it has the chance to throttle down to 25W.
The ambient temperature is 20 °C (~68 °F) and if I start a game it runs in the 35-40W range for the first 10 minutes and then it starts throttling down to 24-25W in the next 5 minutes. Temperatures in the 35-40W range are around 100 °C (212 °F), while sustained temperatures at 25W never exceed 83 °C (181 °F).
First and foremost it should never ever shutdown because it reaches the thermal threshold. Also it should be capable of sustaining at least 35W: it's a big bulky 16 inches chassis from a reputable manufacturer sold with a 35-54W APU.
I've already upgraded to the latest 01.04.10 bios revision via LVFS (fwupd) and I'm running Arch Linux with the latest kernel 6.8.5, but I've also backported the AMD P-State Preferred Core patchset from kernel 6.9 in the hope that it gets a little bit colder by preferring the most efficient cores whenever possible.
Anyway this has little to do with Linux: it's either a thermal design issue or a production defect of my specific sample.
I hope it's the latter because I've used Dell laptops in my business for decades and never had such issues but I've always considered HP to be as reputable as Dell and I would be be very surprised if that ends up being a design flaw.
Please let me know your experience so that I can compare it to mine.

[Jupitero]

Hello.

Your laptop shutting down due to CPU thermal issues in a normal environment should not happen. I presume the fans are working hard and the airflow vents both in and out of the laptop are unobstructed and cleaned of dust/lint/debris? Has this problem been there from the very beginning?

Linux + AMD power management support has received a lot of updates and I expect this to continue. The Linux kernel AMD support just hasn't been on the same level as with Windows and may lack some crucial power management stuff on the chipset level as well.

Because HP does not support anything beyond Windows, they may ask you to try recreating the issue with Windows. So, if you have the means to install Windows on a separate drive and you can recreate the symptoms, then you will have easier time with the HP support as well. That's something I would try just to exclude any software problems.

[darkbasic]

The airflow vents are unobstructed, the fans are spinning like crazy and I had this problem since the very beginning.
I didn't try to reproduce the issue with Windows yet, but I will if I have to. My understanding is that it's the firmware which decides when to throttle, so whether it is capable to sustain 35 W or not shouldn't be influenced by the operating system.

The shutting down issue on the other end could be influenced by how aggressively the frequency governor asks the CPU to boost, but in the end it shouldn't happen anyway (it should simply ramp down earlier in response).

HP how much sustained power did you engineer this laptop to handle? Is it a bad thermal paste or a design flaw? I've benchmarked it against any other Phoenix laptop available on the market and on average it struggles to keep pace even with the AMD Ryzen 7 PRO 7840U due to thermals.

[darkbasic]

As I suspected It turns out HP makes good laptops, but apparently TERRIBLE docks.

The issue is caused by the HP Thunderbolt G4 dock: as soon as I unplug the dock and attach the bundled USB-C power adapter instead it can easily sustain 41 Watts of load.

This is not the only issue I have with this dock, so hopefully it's a faulty one: https://h30434.www3.hp.com/t5/Business-Notebooks/HP-Elitebook-865-G10-loses-connection-with-the-HP/m...

[darkbasic]

The more I dig into this the more it gets weird: whether it sustains 40W while gaming or not doesn't depend on the power adapter but on the usage of external monitors via the dock. In fact I can charge the laptop using the dock and it can sustain 40W during gaming if the external monitors are not attached. I'm unsure about the reason (frame rate is uncapped so it should use max power either way), I will try to experiment further.

[darkbasic]

AMD engineers said the Embedded Controllers control the throttling behavior so the fact that the issue only occurs with the Thunderbolt Dock G4 is most likely an issue with the dock's Embedded Controller: https://bugzilla.kernel.org/show_bug.cgi?id=218305#c65

Another user has a similar issue where his Ryzen 7 7840HS  gets stuck at 544MHz frequency after resuming after unplugging the power cord during sleep.

AMD's suggestion is to have someone's at HP have a look at the issue because Embedded Controllers are basically a black box to anyone but the system manufacturer. I'm sure there is someone at HP who would care about this, because the Thunderbolt Dock G4 supports a wide variety of systems including the HP Dev Zero which comes preinstalled with Linux, but I don't know how to reach them. Can you please point me to the right person/support channel?

[Jupitero]

---

@darkbasic wrote:

Can you please point me to the right person/support channel?

(I don't work for HP nor do I have any more resources or contacts there than you have)

Well, your first step would be to open a case with HP. As the HP support person likely* does not have a solution, ask him/her to escalate the problem. HP has their non-public product bulletins with known problems and solutions.

Your situation is a bit murky since Linux is not supported with your laptop, but then again HP does support the Thunderbolt G4 dock with "Linux Ubuntu 20.04 LTS 5.14 Kernel", as mentioned in the TB4 Dock Quickspecs document, so that's something to test - unless it won't work at all with your laptop hardware - could be tricky.

If you could cross test the dock with another laptop model and can reproduce the same problem, your case could have better chances of going through to the actual dock firmware team at HP.

If the support case is closed without a solution, you can always try their official social media feeds to "make some noise".

Finally, you could also try to contact the HP dock firmware team manager (or individual dock engineers as very last resort) directly via email since your support case may not have reached them. You'll have to do some sleuthing at Linkedin to find the manager and formulate the subject+message in a way that the manager reads it and won't just move it to trash...