I have a Batch 5 framework 13 Ryzen laptop and I’ve been noticing some weird temps. It was fine for the first month or so I have been using it usually no more than 75C briefly with web browsing and office work, but now it is pegged at 85+ often hitting 100C all on sensor 4 on the CPU. I don’t want to have to replace the thermal past but will do it if I have to (I have some NT-H2 if that is appropriate, or can use thermal phase change pads).
My only concern is if this would void the framework warranty. If it would then I have no issue sending it back as I can easily use one of my other laptops for work.
You are not alone. I have a batch 9 Framework 13 AMD Ryzen 7 7840u, and I’ve also been having issues with the thermal performance. This does not seem to be limited to a single sensor. But it’s very easy to push the overall temperature to 100c, even with a single core workload. The laptop was doing this right from the get-go. I contacted support and explained the problem to them. After some back and forth, I was instructed to try repasting the CPU/Heatsink with some NT-H1 that I had lying around from my last desktop build. This did not seem to make a difference. Support then sent me a tube of their stock thermal paste and instructed me to repaste it again. Still no difference. I noticed that the amount of paste between the heatsink and CPU would be pretty much none. Like the mounting pressure was enough to squeeze it all out. So I started to experiment, and I’ve found that if I leave the screws for the heatsink just a little loose, then the thermal performance is significantly improved. By a little loose, I mean that I turn the screws back in one turn at a time on each screw and that I stop one turn after I feel resistance on the screws.
Now, my laptop performs as it should, and it can sustain an all-core load and maintain a temperature of 90c after the fans kick in. Before, it would hit 100c with only one core, and I wouldn’t see the temperature drop until I stopped testing.
Great that this works for you, but I find it a bit curious. From a physics perspective, the more you squeeze out, the thinner the paste layer gets, the better your thermal conductivity. The optimum would in principle be a Metal-Metal contact without thermal interface material - but due to air trapped between the sheets, you need some thermal filler, i.e. the paste.
Paste is alright but the ptn is closer to liquid metal than to paste especially in the kind of power envelope of a laptop. I also didn’t believe it until I tried and then moved on to lm but that really wasn’t worth it.
I’m borrowing a space on your topic since I was about to open a very similar one myself.
My company recently bought a few new Laptop 13 with AMD Ryzen 7 7840U mainboards. Only one of them has been checked out to a colleague, with Windows 11 freshly installed on it. He has been tasked to report on his experience during a few weeks, so we can have feedback on these laptops suitability for work.
The feedback here is notably that the CPU seem to have its last CPU temperature sensor through the roof on idle. Using CPUID’s Hardware Monitor, the most common idle behaviour is as follow :
TZ00 : between 50 and 53°C
TZ01 : between 50 and 53°C
TZ02 : between 50 and 53°C
TZ03 : between 79 and 83°C
Strangely enough, starting a videoconference (Google Meet over Google Chrome in this case) does not trigger any change in this profile.
It seems to me that either one of the cores is doing all the (light) work, or one of the core is not dissipating heat properly.
I have seen similar topics on the community forum, including yours. Some people are suggesting repasting the CPU, hoping for better thermal efficiency. For the sake of trying and having better knowledge of the Framework hardware, we might try that, but I must say that just like you, most topics I read seem to have a very mild improvement after repasting.
Last activity here dates back to february. Are you still experiencing some kind of unbalanced heat dissipation as of today ?
I don’t think the 4 on-die sensors reported by the CPU represent different cores really. Rather, they are sensors at different locations on the die. My interpretation was that Temp4 (TZ03) is closer to the cores, but I don’t really have a source for that. So that one being higher than the others seemed normal to me.
FWIW, I installed PTM7950 when I started to regularly push my cpu with compute-intensive tasks and temps went down. Temp4 is still way higher than the others, but reaches a steady-state of about 80°C after the fans have spun up and temperature stabilized under stress-testing. That seems acceptable to me if it is some equivalent of Tjunction.
Did this change the relative temperatures between the 4 sensors reported by the acpitz-acpi-0 interface, or only overall temperatures?
80°C on stress testing seems perfectly fine indeed, but 80°C on idle looks concerning to me. As we have 3 other AMD mainboard in stock, we will probably take one out and do some testing to see if all are displaying the same behavior (and if yes, we might try to repaste it)
May I ask how you were able to display acpitz-acpi-0using the sensors tool ? We just installed a fresh Ubuntu 24.04 on a second Laptop 13 with AMD 7840U in order to investigate temperatures, and it seems sensors-detect cannot see any CPU-related sensors.
That being said, the current idle temperature we can feel seems very very low. No fan rotation at all apparently. I don’t think this one is running at 80°C, so there might be a problem with the first one we distributed
Never had any issues with NT-H1 thermal paste on any of my desktop builds.