I’m running latest Fedora on my Framework 16 and have an interesting phenomenom: When I have the battery installed, my CPU seems to run 30-40° hotter than without the battery installed (while the PSU is attached in both cases). That causes the laptop to go into emergency throttle frequently, where the CPU goes down to 500something MHz for 2 or so minutes. In direct comparison, idling without the battery installed (but with midplate and keyboard/touchpad installed) the CPU is somewhere around 60°C. As soon as the battery is present, the CPU Temperature almost immediately jumps 30-40° and runs at much higher temperatures.
My first thought is that your battery removal is shifting the processor mode from one thing to another; that is unusual, since it is normally mains power that would have that effect. Nevertheless, I would dig around on that theme first. Consider looking at dmesg --follow and then do some battery removal/installations etc to see what your system is doing.
I notice also that your CPU seems to be under 50-60% load, which is very high. Have you run it that level just to illustrate the thermal phenomenon? That is still a lot of load to be running over the long term.
Finally, consider trying another distro; Ubuntu on a bootable memory stick should be enough to try an alternative.
At that moment I was recording that video (OBS encodes on CPU for me) and also had a game (going medieval) running in the background. I need some load on the CPU to demonstrate that behavior reliably. Did the experiment again with dmesg --follow running (also with journalctl -f) and neither is logging anything related to the battery attachment or removal. you can also see that the power preference as well as the governor did not switch during the battery removal or attachment. What is interesting is that the CPU seems to cap at 50W as soon as the battery is removed. That feels adequate to the 54W TDP that AMD lists for this CPU. As soon as the battery is installed, note that almost 90W are ostensibly flowing through the CPU, which would easily explain the higher temperature
What is interesting is that despite the CPU consuming 40W more power, the battery is actually not doing anything charge or dischargewise, it’s at 0.00W Power input/output at the same time
Yes, but for some reason your CPU frequency is dropping from 4.5 GHz to 2.5 GHz when you remove the battery. So something looks like it is making capability judgements based on power systems available.
I’d try another OS next - should take half an hour or so to get the same measurements on another distro.
You have a point there, though I didn’t notice the frequency drop when the battery is removed. So there is definitely something going on with the CPU overextending itself as soon as the battery is present, even if the battery is not utilized (I’m running on the 240W PSU, so the laptop should have ample power to work with even without battery). What’s also interesting is that I’m still on the “balanced” profile, not the performance or powersave profile. And even in Powersave, the CPU only goes down to 3.7GHz with the Battery installed, and to 1GHz when the battery is removed. So strange xD
I would recommend to get onto the Fedora Forum and ask them as well (if you haven’t already done so). I ado agree with a previews poster that you should grab a life Distro using a USB stick and see if the results are different.
Yes. This can be an EC firmware thing. On the Framework 12, it has a mode where it lets the CPU use 20W continuously instead of just 15W if the charger is high enough wattage. But the minimum charger wattage required to shift into that mode is higher in circumstances where the battery is not installed because there is no fallback for when the system demands too much power then, so the solution is to increase the minimum wattage required for 20W continuous operation in no-battery mode.
A similar thing COULD be going on here, but I don’t know why it would get to the point of making the CPU emergency throttle like that. Maybe the cooling system isn’t up to the task? Dust, probably?
INSANE DISCOVERY: FRAMEWORK 16 SUPPORTS HOT-PLUGGING THE BATTERY!!!11111111111!!1
I’m not joking, no one I have known (IRL on in this forum) had ever tried that.
The CPU shouldn’t reach 101C for more than 5 seconds during normal use, you may want to check the cooling. Check out this topic Uneven CPU thermals!
I actually had uneven thermals before, which disappeared once I pivoted away from liquid metal. Since then, individual core heating was not really prevalent anymore. I am intrigued by that thread however, when doing my research I read that this cpu does not provide per-core heat measurements, but the screenshots there suggest otherwise?? I basically ran each core individually to determine heat buildup…
I airsprayed all cooling outlets thoroughly, those are clean. I also examined the heatpipe with an IR Thermometer to check if there maybe is a disruption of heatflow, but there is no meaningful gradient between either end, which to me seems like the heatpipe is doing its job. I DO notice that even with the fans pinned at 100%, there is not really any airflow noticable by hand. I mean you can feel it, but judging by the noise those fans make when running full juice I’d honestly expect more.
This IS a batch1 laptop if that helps, that sadly also means it’s long out of warranty
I rewatched your video and noticed that your CPU power was at 85W sustained, at this level no amount of repasting the thermal paste/TPM is going to cool it down. idk how did you manage to do that as I’m pretty sure the sustained CPU wattage is about 45-54W. Did you use any kind of “tinkering” software like ryzenadj or Smokeless_UMAF?
70C at 50W is very good cooling. So the problem isn’t cooling or low power without battery, but the high (85W) with battery and AC connected. The CPU may boost to 75-ish watts peak but should reduce to 54W after a few seconds so the CPU won’t reach 100C. Find a program that can show the PL1 and PL2 values?
I haven’t watched the video yet, but this is my thoughts as well. We know that the power budgets between battery attached vs standalone mode are drastically different, so running hotter with battery is to be expected IMO.
under Ubuntu, I was not able to reproduce this behavior at all. Plugging the battery caused the CPU Speed to shortly drop to 2.5GHz, then back up to 4 something, while the temperature reported stayed the same. Mission Center unfortunately did not have a wattmeter under ubuntu so I can’t say anything about the actual power consumption.
I found a tool I installed earlier, (Coolercontrol) which I wanted to use in the past to get it to cool earlier (because I thought that the cooling was just too little to keep it from throttling). I uninstalled it and it appears that for now, at least the power-save governor is keeping the CPU cool (which it was not able before) so perhaps there really are multiple tools somehow fighting over this (I did not find any other tool that sounded like it was playing with the governors, though)
as you can see on the Screenshot, on “balanced” I reach over 110W
How about s-tui? You can even adjust the power profile and governor there. You might need to sudo to show real-time power usage. Also, amdgpu_top can show individual core temperatures
Powersave is more like automatic mode, and the power is controlled by Power preference
