Thank you, that’s what I thought. I’ve seen it happening a few times here under different condition: high CPU usage, high GPU usage and even high NVMe usage (2TB Samsung 980 Pro)
But it usually goes away after 10-20 minutes of waiting on my Laptop (DIY i7-1280p), so it doesn’t stay fixed on the 400 MHz.
Probably worth logging your temps and package power under load if you can. See if it’s regularly throttling at 100C. If so, I would replace the heatsink/fan (it’s straightforward), or at least blow it out with compressed air and replace the thermal paste. There are several other smaller thermal pads in addition to the CPU/iGPU paste, so I don’t know how successful cleaning vs replacement will be in general. Those other pads were also somewhat degraded in my case.
For me it throttled also while the CPU was below 100°C (80-85) but with high load on the NVMe.
I can run 4 VMs with Linux (2x CachyOS, Fedora, Zorin; 8 GB RAM, 4 Cores each) at the same time, let them all make updates, without any issue.
But if I fire up the VM with Windows 11 (AzureAD joined and Intune enrolled; 16GB RAM, 8 Cores) chances are really high it throttles down to 400 MHz a few seconds after I login (because than MDM kicks in and there are running a few checks and also virus protection is scanning the disk). This is at around 40-50 % CPU load (on the host ofc.) and 80-85°C CPU temp.
Guess I need to order PTM for the CPU and something to get the heat away of the disk. While I’m on it I think I should also look for a solution to keep the RAM cooler.
Thank you for the detailed insights you provide, they confirm my thoughts that it’s a general heat issue with the laptop (which triggers the bug).
Ok. I get what you’re saying. Ill share it with the Internal teams and see if we can come up with something.
It is not clear to me if you consider the thermal paste to be degraded in your case. But from your diagnosis (up thread) I wonder if the liquid in the heat pipes has found a leak. It would only need a pin hole and probably wouldn’t be noticeable. Not sure if you have multiple pipes, but if one failed that could well be the problem in your particular case.
Hmm. Could be. I threw the old cooling assembly in the trash. But I might go dig it out and experiment.
Disabling BDProchot is a bad idea.
Yes, the modern CPUs have sophisticated thermal sensors. But features like BDProchot is a last resort emergency feature. And as demonstrated by Jonathan, there was a problem. So if in the future you have a real issue and the BDProchot is disabled, you won’t know.
Last resort emergency exists for a reason. Because if it doesn’t work it could lead to worse, damage of the components, or even a fire. Safety features have to be conservative, and have to have multiple backups.
Jonathan, the reason it doesn’t unassert itself without turning off power(latching) is because that’s usually how safety features work. It wants you to turn it off for a reset.
In the Pentium 4/Athlon days, AMD got a lot of flack because without a heatsink the CPU would burn itself up, whereas the Intel chips did not. Safety features are tried and true battle tested.
I don’t think you were unpleasant.
It’s still a false alarm. In Jonathan’s case. The CPU was working comfortably at 100C because the CPU’s power limit is continuous variable even down to as low as 17W without the temperature exceeding 100C. In Mahe’s case, the CPU went down to 400MHz but the temperature was only 80-85C, which is far from an emergency. You were right on the last resort emergency thing, but the criteria of “an emergency” in the FL13 Intel 12th board is wrong.
One caveat: I don’t think it was “comfortably working” at 100C. But latching into an unrecoverable state with no documentation is not reasonable in my vew.
Part of the issue is the thermals on this laptop are not great at the best of times. Fans blow like crazy, the base gets very hot. So the pathological “17W at 100C” doesn’t feel that different from ordinary operation. There’s not much in the behavior to indicate to the user that something is wrong. So when it latches, cooling is not the obvious culprit. And it seems some causes of latching are not direct CPU/iGPU cooling issues regardless.
Better documentation from framework of this failure state would make a big difference.
The problem may be Framework’s implementation of hardware but it’s still an emergency protection feature and other measure must be done to resolve the issue over outright disabling it.
Again it’s a bad idea.
I have the same issues. Including that the moment my processor starts doing anything it almost immediately reaches 100c. so ordering a fan assembly right now.
The new assembly came in, and it doesn’t throttle anymore. ran a 10 minute multi threaded cinebench test and the processor didn’t go below 2ghz. And in Linux s-tui temps don’t go above 80 running core 0-7 at 3000 and the others at 2400
Edit, this (image) probably didn’t help
Yikes! Yeah, you should probably clean your fan more often.
You can’t anyway, or at least when I tried it immediately gets reset, see here
