Laptop 13 Pro (Core Ultra X7 358H): CPU at 100 °C at only 33 W, fan stays at 0 RPM while the die heats up

Hi all,

My Laptop 13 Pro (Core Ultra X7 358H, Arc B390, 32 GB, a few days old) runs at 100 °C under almost any sustained load, and I’d like to know if others see the same or if my unit has a cooling problem before I contact support.

Setup: Fedora 44 (kernel 7.2.15), BIOS 3.04 (updated via fwupd), performance platform profile, lid open on a stand, Framework 100 W charger.

I ran the same 35-minute test sequence (stress-ng, 7-Zip, vkmark) on BIOS 3.02 and 3.04, logging package temp, power, clocks and EC fan RPM every second:

BIOS 3.02 BIOS 3.04
1 thread (~13 W package): max temp 97–100 °C 100 °C
Fan during that 1-thread test 0 RPM (whole minute) 0–2,300 RPM
All-core: time to reach 100 °C 8–10 s 2 s
All-core sustained power ~32–33 W 33 W
Seconds throttling in last 5 min 201–299 of 300 299 of 300
Fan at sustained load ~4,800–4,900 RPM ~4,900 RPM

What stands out to me:

  1. 100 °C at only 33 W, with the fan already at ~4,900 RPM. Phoronix reported around 85 °C under sustained full load on a 13 Pro. On mine, the cooler seems to barely handle 33 W.
  2. The fan starts too late. With one thread at ~13 W, the die went to 97–100 °C while the fan stayed at 0 RPM. The EC seems to follow the board sensor cpu_f75303 (31–36 °C at that moment) rather than the die temperature. 3.04 starts the fan a bit earlier, but still not before 100 °C.
  3. It hurts real use. In games the CPU and GPU share the 33 W budget. Uncapped, the CPU sits at 100 °C and the Arc B390 drops to its 1,000 MHz minimum (pl1 throttle). Capping the CPU at 2.4 GHz brings the package to ~88–90 °C and lets the GPU run at 2,200–2,500 MHz, almost doubling FPS (Diablo IV: ~24 → ~40–46).

Single-core clocks and benchmark scores are within spec (7-Zip matches Notebookcheck), and BIOS 3.04 fixed the GPU PL4 cap (SoftwareFirmwareIssueTracker #263). So performance is “fine” on paper, but only by living at Tjmax.

Questions:

  • Do other 13 Pro (Intel) owners see 100 °C at ~33 W sustained, or is ~85–90 °C typical?
  • Has anyone else seen the fan stay at 0 RPM while the die is near 100 °C under light load?
  • Is this worth a support ticket for heatsink contact or thermal paste, or is it expected with the current EC fan curve?

I have full logs and plots and can share them. Thanks!

Contact support. Probably not how it was designed though support is going to be able to give whatever plethora of details that has been requested.

Presumably some or most of the OP was written by AI. Consider disclosing this at a bare minimum and for a more “human” response, whittle down the data and drama that is AI generated. All it knows is how to barf information.

Understandibly, it is easier to have some of the work automated but the resulting AI response from an AI response is probably not what anyone wants to bother with.

This is not meant as a personal attack at the OP. Just trying to help keep the human engagement going on these forums since that is much more valuable to everyone.

  1. Heat conduction is poor, high die temperature but low board sensor temperature
  2. bad cooling system, the fan is only 4900 RPM, it should be 6000-7000 RPM at max load, try using ectool fanduty 100
  3. Low thermal capacity, 100C is reached with only 2 seconds.

Thank you for the replies.

Regarding the AI thing, I used AI to write part of the post because I had run so many benchmarks and tests that I wanted to post the results in a readable way. I find it very useful, but it may be a matter of preference.

I carried out more tests with the fan maxed out using the framework_tool . At 100% it reaches the 6500 RPM numbers. However, all my tests were run using the performance power profile in Fedora, and this may be the cause of reaching 100 degrees in the CPU package. For whatever reason a sustained 33W skyrockets the temps regardless of whatever the fans can do.

Switched to balanced and do not have temp issues anymore, but maybe at the cost of performance as the package runs with less wattage.

I will contact support but I suspect that this is how the Laptop was designed to perform, anyway if I find out more I will post it here.

I have this EXACT problem, along with a cpu core 3 issue where there is a parity error in the core causing reboots

Here’s a summary of some debugging I did, i think my cpu is in worse state during these peak events:

I think I found why the fan ignores the die temp. Run sudo framework_tool --thermalget and look at the peci-temp row (the CPU die sensor):

sensor warn high halt fan_off fan_max name
4 - 120 127 103 105 peci-temp

The fan ramp for the die sensor starts at 103°C and maxes at 105°C, but the CPU throttles at 100°C, so the die temp can never drive the fan. Only the board sensors do, and they lag behind. I changed it with:

sudo framework_tool --thermalset 4 -1 -1 -1 60 90

That ramps the fan from 60°C to 100% at 90°C. It stuck through a reboot for me, but I reapply it at boot with a small systemd service just in case.

It still can’t keep up with load spikes: under all-core load the die goes from ~50°C to 100°C in 1–2 seconds, and the fan takes a few seconds to spin up. So brief 100°C peaks are expected. Sustained load should hold clocks better now, though.

Mine is worse than just heat. One core (CPU 3) keeps logging mce: [Hardware Error]: CPU 3: Machine Check: 0 Bank 0: 8000004000010005 (a corrected internal parity error). That’s 11 times since day one, including at idle 16 seconds after boot. About 10 seconds of all-core load hard-resets the machine, and the firmware crash records (BERT) show a fatal uncorrectable machine check. You’ve run yours at 100°C for 35 minutes without crashing, so I’m treating mine as a defective CPU and going through support.

I tried some tests, putting fan to 100% manually then doing a 15s load test:

yeah… i think my cpu is bad. I am able to set the max power limit to 30w (i also tried 35w), and the ONLY way it doesn’t crash is if i set a limit on my cpu core #3 to 3.5ghz, otherwise on a burst workload, the machine instant shuts down. before fans can ramp.

I tried 4ghz on the core, and still at least a parity error happens, so it seems well under the amount its supposed to handle at boost.

with that core limited, all others are fine to go near 100c for a split second while waiting on fans.

Hoping FW will replace but at least my machine won’t crash by limiting this core, but im concerned long term if this core gets even worse and i have to limit it even further.

The exact changes I did for anyone else with problems:

What's installed now
- CPU 3 capped at 3.5 GHz, with the other 15 cores at full speed.
- Power limited to 35 W, both long- and short-term, in both places the chip enforces limits. Stock was 35 W long-term and 60 W short-term.
  - Both are set by /usr/local/bin/cpu-safety-limits. cpu-safety-limits.service runs it early at boot, and a hook reruns it after resume from sleep. I checked that power profile switches don't reset them.
- Fan curve: ramps from 50°C on the CPU die and runs at 100% from 80°C. It's applied at every boot by fan-peci-thresholds.service.

They approved a mainboard replacement for me. Will run full load tests with different wattage limits when i get the new one :slight_smile:

This is a very informative follow-up, Zach1, and useful for anyone experiencing something similar. It’s good that they approved the replacement. I hope you get it sorted out soon.