FW16 AMD 7840HS, no dGPU. Just APU.
Mine settles at 83.8 C in performance mode, with amdgpu_top showing 45W.
I had an older mainboard, with PTM (I replaced the LM), that would sit at 100 C, but it was replaced with a more recent one that sits at 83.8 C. The PTM made the temp range between cores improve. I.e. with LM the range was 20 C +, with PTM is was 10C range or less. But it still got Tctl 100 C with the PTM.
So, there might be something else at play to cause 100 C.
- PTM
- Heatsink solder bad. (There is some solder that holds some bits of the heatsink together, it might be bad)
- CPU bad. Maybe its just a bad CPU that gets too hot, and even with a great heatsink, it does not improve.
With the replaced mainboard, I suspect the (2) and (3), but I have no way to test.
As the PTM made the core temp range less (10 C or less), I think the PTM was working. But maybe the heat sink itself was faulty also.
So, I am just saying that PTM might fix the problem, but there are other causes.
In summary, it is possible to get a mainboard that does not hit 100 C.
Running with:
stress-ng --cpu 0
After a few minutes it is steady at:
sensors output:
cros_ec-isa-0000
Adapter: ISA adapter
fan1: 3343 RPM
fan2: 3101 RPM
ambient_f75303@4d: +48.9°C
charger_f75303@4d: +50.9°C
apu_f75303@4d: +58.9°C
cpu@4c: +83.8°C
gpu_amb_f75303@4d: -0.1°C
gpu_vr_f75303@4d: -0.1°C
gpu_vram_f75303@4d: -0.1°C
gpu_amdr23m@40: -0.1°C
Technically, if I really try, I can get Tctl up to 100 C if I really try. But in normal operation, it stays below 90 C.
This thread shows where the heatsink solder can sometimes be bad:
The whole uneven-cpu-thermals thread is a really good read, if you have the time.