[RESPONDED] Excessive CPU thermal(?) throttling on 12th-gen

So now the issue has been identified, and the resolution isn’t something that can be fixed with a (alpha state, potentially bricky) BIOS update, when is the recall happening?

A precedent has been set. Can I assume if I go through the trouble of spending a month going back and forth with support that I will have my mainboard replaced too? @nrp

I found a solution on my i5-1240p that gets me stable framerates without having to set the power profile to “power save”.

Just two steps:

  1. Set the Fans to max while gaming, because the normal fan curve is not agressive enough: sudo ectool fanduty 100.
  2. Set the maximum frequency to 2900 MHz: sudo cpupower frequency-set --max 2900MHz.

Both steps were necessary for me even with PTM7950. If I don’t set fans to max then the temp isn’t controlled in time and it’ll throttle to 400MHz, and if I don’t set max frequency under 3000MHz then it’ll overwhelm the fans even at max RPM and then throttle to 400MHz again.

To undo these when you’re done your game, run sudo ectool autofanctrl and sudo cpupower frequency-set --max 4400MHz.

Would be really nice if these laptops could settle into a sustainable clock on their own though.

Tested on cyberpunk at 1440p @ low settings with a Radeon RX 7600XT eGPU over thunderbolt 4, got very stable 35-40 FPS. Without these changes I’d get 40-50 FPS but with frequent drops to 5-10 FPS when it throttled, which was unplayable for me.

Would be really nice if Framework produced a working computer, or at least didn’t abandon us with expensive, out of warranty e-waste.

1 Like

I’m seeing the same problem after a few minutes on a Google Meet with the camera enabled. I don’t use any 3d games or anything that should really be pushing the cpu hard enough to throttle. I used to be able to use Meet with video enabled just fine though who knows what changes Google’s made.

I have noticed that watching the temperature sensors the temperature is normally pegged at 90-100C when using Meet but occasionally it will jump up to 180C for a single sample. I assume this is a bug, either in firmware or in linux that’s causing it to read the sensor incorrectly. I wonder if that’s what’s triggering the throttling if it’s a firmware bug.