The fans in my AMD Framework 13 spin up to a high speed, and then drop back down after 1 second.
This happens randomly when I am using it, when I am in any application. The fan noise does not seem to be regulated by lower fan speed to prevent such from occurring. It is sort of annoying when you are concentrating on a task and you hear a random bust of air coming from the system.
Also, does anyone’s AMD FW13 drain battery fairly quick while using Firefox? (Fedora version, not flatpak). Even when I have one tab open (this page that I am writing this post in) the system gets hot where it kinda starts to get uncomfortable. My power mode is always in power saver:
My CPU is currently at 101.84 F
38800 / 1000 = 38.8 C (101.84 F)
gxnni@necrozama:/sys/class/thermal/thermal_zone0$ cat temp
neofetch screenshot (since idk what else to share)
Same on Windows, fan is off most of the time and when it turns on its rather noise, so high rpm on the lowest setting and it modulates too quickly
I’m glad that I’m not the only one, and that it might only be a Linux issue. If we’re both using AMD FW13, maybe it’s something to do with the firmware. Either way, it’s jolting when you’re super concentrated on something. Hopefully something about it is addressed.
I recently spoke with support about this, and it is behaving as designed:
The fan speed is controlled by external temperature sensors outside the main APU. So it will take a longer time for the fan to spin up. The fan start temperatures are actually 40/46C. But the sensors are located further away from the main APU. You can see the configuration here: https://github.com/FrameworkComputer/EmbeddedController/blob/lotus-zephyr/zephyr/program/lotus/azalea/overlay.dtsi#L23
Perhaps future ec firmware versions will tweak the fan curve?
They should consider it. In my eyes, if the sensor is too far away from the main source of heat on the board, seems like a massive design flaw. If they need to adjust some values to make it not spin up as fast, that would be great. No other system that I have used crank up the fan to this amount, just for it to halt after a second. (don’t even get me started on the supreme thermal control on macbooks, because I do not want to compare apple’s to oranges) (pun intended) That would greatly increase the usability and customer experience.
The fan spinning up very high and then stop again after a second doesn’t sound like something that comes from a temperature control, especially if the sensor is a bit further away, where I would imagine the temperature to change a lot slower (because of thermal mass). Unless they have a fan-curve like “40C is off, 41C is full blast”, where it suddenly goes over 40 and then is under 40 again a second later.
So the fan spinning up full blast for a second sound like a different bug. Why does it even spin up that fast? If the temperature slowly raises on more or less idle-load, the fan should never need to ramp up that fast. It’s like the sensor goes “OH SHIT, IT’S REALLY HOT … oh wait, false alarm”.
Thank you for your input. I believe that the threshold should be changed if this is the case. There should be tiers to how much fan speed is associated to temperature of the system. A gradual increase would be way better than a full blast (… for a second).
That is the question that I am trying to get an answer for! That’s the reason that I created this thread.
I don’t know if there are multiple sensors, for example in the link @46hawk posted, there is also
cpu-f75303, which I don’t know where it is. But if that’s in the CPU itself (from the name of it), with
temp_fan_off = <46> and
temp_fan_max = <54>, which is a very small range. And on the CPU die itself, temps probably spike a lot more, even from some short background task on an otherwise idle CPU (which then goes down quickly again because of the thermal mass of the cooler). So maybe that fan-curve is too aggressive?
Either way; it’s something that I am not comfortable enough to mess with just yet. If users are experiencing this regardless of host OS, this is something the manufacturer (Framework) should take a look at.
I also have this little beast here and must say i’m happy with the noise levels. i also use firefox and it seems not to act like yours with the fan jumps. i also did a little browser benchmarking and it it started to give more cooling just in the upper 80 degrees (celsius).
but i must say - one of the first things i did after installing ubuntu and the stuff to make fingerprint sensor working was installation of auto_cpufreq - because on my ryzen 5000 desktop (windows) i also have this issue with permanent fan speed jumps. this is because of the aggressive boost overclocking and a sff enclosure.
regarding the question about multiple sensors or not - i thought of there are some more internal sensors for measuring hotspots in the cpu/apu but i cannot provide sources for this piece information (sorry about that).
Thank you for your response. I am also working on some out of the box Linux customizations. To assist myself in the future (in case I distro hop) I am tracking all of my configurations in a handy Libre Office file. I am daily driving Fedora for the first time on my AMD FW13, and it has been an interesting experience. Nice of you to mention auto_cpufreq. This is something I had recently learned about while doing some background research about power management on Linux. To my surprise, power management might as well be non-existent on Linux since it does barely anything (at least with the default tools). I first heard of the auto_cpufreq package from the Chris Titus Tech YT channel, where he covered the benefits of using it. From what he described, it looks fairly promising and I think I may check it out:
However, I hate to be the bearer of bad news, I don’t think this will address the rapid spikes in fan noise. Though it may help place a band-aid over the issue at least. I cannot say for sure, before I do more testing of my own.
Thank you (again) for your comment either way!
you are very welcome!
i like the idea of documenting the steps needed to create a reproducable workable linux installation on your baby!
i also throught about that but in a more technical way (what i did not go so far). in my job our sysadmins use a tool called ansible to make such states reproducable by just running a “playbook” what is basically a configuration unit in a certain format. but i did not start with it yet.
when you are speaking of fedora - there is another post in this forum about OCI images - this is another step further by just creating dowloadable ready to use fedora images for the framework laptop with certain hardware configurations - but i think they are in the beginning of that topic. maybe you can provide your docs there
link to the topic: Custom Fedora OCI images for Framework laptops
have a nice evening!
Very cool! At my job, imaging software is only used for Windows systems.
I’ll need to check it out! A lot of this is so new to me (especially the concept of Linux images), so it will probably take a while before I can make meaningful contributions. Of course this is something that is possible, considering everything in Linux is a file. This is def something I should look in to more. If I can contribute to open source initiatives that make my life, and everyone else’s easier, I am all for it!
Tracking this. As you have new details please update it here. Thanks