I am also seeing temperatures in the 95-100C range when playing Helldivers 2. A bit weird that I’m hitting stress test level temperatures when gaming, but it would appear that 100 degrees is still a safe temp for this cpu?
According to the technical specifications from the manufacturer’s website https://www.amd.com/fr/products/apu/amd-ryzen-7-7840hs
100 degrees is the maximum temperature, 100 degrees for
any laptop is not very good.
I’d report this to FW Support, maybe there is a fix to help with that.
1 Like
Yep! Already did a week ago, and it’s getting escalated. Thanks 
I’m running a cinebench right now too, since that should help tell me if my performance to thermal ratio is poor
I don’t have any equipment or a good frame of reference to say what the sound level is.
Is this on the dGPU or iGPU? I’ve yet to play Helldivers, but whenever I’d game on the iGPU, I’d hit the same temps as a full CPU load would do. If it’s on the dGPU and that’s the limiter to the FPS, CPU is routinely 15-25c cooler 
Genuinely did not occur to me to do that after reading this post:
Sorry for the late response, not too sure how accurate my thermal camera is, but after ~17 minutes of a CPU-only stress test (100c on CPU the whole time), hottest spot on the keyboard is around ~37c, whereas on the bottom the hottest spot was at ~40c
Thanks for the info, maybe it should be put on one of those laptop cooling pads when gaming or doing other stressful work.
I’m playing on the dGPU, on Linux (Nobara 39), and when I’m on Performance power profile, I’ve seen it spike to 103. When in power-saver mode though, it stays generally below 95.
I’ve reached out to FW and AMD and at least AMD seems to suspect something may be wrong, but we’ll see shrug
1 Like
As an update: Framework replaced the device, and now the thermals are within the expected temps!
3 Likes
really curious if anyone has removed the stock option, and conformal coated and used conductonaut on both apu and gpu…
Out of curiosity, did your old unit have a large per-core gap in temperature? I had recently read some other posts about such gaps and tested my own system on Windows for it. (Not aware of how I might test per-core temps on Linux) Turns out between the hottest and coldest cores, the gap might start out around 20c, then typically narrows down to ~15c, with 1-2 cores consistently being around 85c while another is stuck at 99/100c. I have finally contacted support about my unit as that definitely looks unusual, and went ahead and ran CB24 to see how mine had compared to others, was definitely on the low end
1 Like
Just FYI since I didn’t see someone mention this yet, but Tctl doesn’t appear to be a real temperature value. It is apparently a proxy for the amount of cooling required by the processor (Source: Kernel driver k10temp — The Linux Kernel documentation). I regularly see it jump from 50 °C to 100 °C when something like gcc starts. Even with the CPU at 100%, it sometimes drops to 70 or 80 °C before jumping back up.
I think the other sensors are accurate, but I can’t seem to find which sensor is actually providing CPU temp
1 Like
I had that exact thing happen to me. Core 6 would shoot up to 100C almost immediately and cores 0 and 1 would stay about 14C cooler under full load. I got a replacement mainboard and now the difference is down to 9/10C and no core is hitting 100C anymore even under full load. My Cinebench scores rose and the CPU is boosting 100-200MHz higher.
So you can be prepared they asked me to run a test, it sounds like you are already using CB24, and send screenshots with thermals. I used HWInfo to show average under load temps and HWMonitor to double verify. Running fresh tests and sending screenshots each time they asked me to: Ensure best performance was enabled, remove all expansion cards, make sure the BIOS is up to date, reset the BIOS to optimized defaults, reset the mainboard by pressing the chassis intrusion switch for 2 seconds 10 times, perform a memory module shuffle, run on 100% battery, run while unplugged and draining the battery, check the task manager to make sure nothing is eating CPU processing in the background, take photographs of the laptop so they can see if there is damage, and also video record a run of cinebench with the thermal monitoring software on top and in full view. Additionally if you aren’t using any of their supported operating systems (Windows 11, Ubuntu, and Fedora) they will probably ask you to do the testing in those.
1 Like
That’s good to know, in that case, I think that would be acpi zone 3, as that seems to always be within a degree of tctl and matching what I saw on windows. Initially thought that was what tctl was too as that was matching all the other numbers I saw for temperature using other utils.
That’s great to hear! Is it quieter in general now? Yep, I went through the support testing process over Monday/Tuesday and yesterday got informed a replacement mainboard would be coming. Was also Core 0/1 for me that were the coolest, wonder if this might have been a cooler contact gone wrong.
I would say the fan volume is about the same. And core 1 is also still the cool one on the replacement board. I don’t know if it is a cooler contact issue or just how AMD designed the chips but it is remaining consistent there.
I don’t seem to see an ACPI zone 3 sensor… Which OS are you using? For me, on Ubuntu 24.04, I don’t really see much sensor data… I get some for my NVME, WiFi card, RAM, and the GPUs, but pretty much nothing for the CPU.
Ubuntu’s package manager(and possibly kernel) is a bit outdated, even btop can’t display the current battery power. Try using Fedora then?
Was using openSUSE Tumbleweed, but looking back at the Ubuntu 24.04 session screenshots, looks like the acpi zones were not there. Like Charlie_6 mentioned, the Fedora releases should also have new enough packages/sensor modules to expose that, tho I’m surprised that 24.04 was missing that.
Ubuntu must have picked up the commit that broke it but not the one this fixed it.
1 Like
Not really worth switching distros just to get a temp reading. I’m on kernel 6.8.0-36-generic, so if kernel support exists, I’m sure something will pop up eventually to make use of it. Currently, though, if I run cat /sys/class/hwmon/hwmon*/name I don’t see an ACPI device, only the k10temp and other hardware, like my SSDs, the WiFi card, etc.
I’m thinking of building btop myself so I can view my GPU there as well, but I don’t feel a need to constantly monitor power. I could read /sys/class/power/_supply/BAT1/uevent for current and voltage just fine, if needed.
Having said that, I did see some interesting stuff when charging where current would oscillate like a capacitor charging and discharging. This was at relatively low SOC, and I’ve been meaning to follow-up eventually but just haven’t found the time. I’m wondering if it was a self-calibration procedure since it didn’t persist very long
Be careful with any software that reads values for the dGPU on a regular basis. It will keep the dGPU awake just to read the values.
1 Like