CPU heat emergency shutdown

[Timur_Born]

Hello Asus team,

I think you need to take another look at your current Sense Skew vs. SIO CPU temp implementation, at least at default/Auto values.

Currently these settings not only lead to the CH6 keeping my 1800X from "soft" throttling (Tctl including offset 95C), but also from "hard" throttling (real CPU temp at 95C).

Even worse, it keeps the CH6 from doing an emergency shutdown due to the SIO CPU temp never reaching 110C. Instead the CPU will start throwing errors (WHEA) and finally crash into Code 8, the latter of which is *not* a full shutdown.

This means that an overheated CPU (over 100C while SIO assumes <75ish C) will still be fed around 1V Vcore even when cooling failed completely, measured via DMM at the socket during Code 8. This seems rather dangerous to me?!

Turning off Sense Skew fixes these issues, but at the cost of idle temps reporting too high for SIO CPU and thus fan-headers.

[Timur_Born]

According to Elmor the overheating emergency shutdown is initiated by Tctl, not by SIO CPU. This makes no difference concerning the problem, though, because the current Sense Skew defaults still prevent this from happening. A Code 8 happens shortly before emergency shutdown would/should happen and then keeps about 1.0 V Vcore applied to the CPU.

This is what temps look like on my 1800X when Sense Skew is used with default settings. This was screenshot short before a Code 8 happened (reproducibly):



And this is what temps look like in the very same situation when Sense Skew is disabled. These temps are real and emergency shutdown happened shortly after taking the screenshot:



As you can see from the CPU (Socket) temps the reported CPU temps in the first screenshot are very problematic. When the fan-curve maximum is set to 75C then the CPU isn't even fully cooled at this point, despite running well towards 110C for real.

[Timur_Born]

Here is a shaky mobile camera video that demonstrates the Code 8 shutdown. This was at "Optimized Defaults" BIOS settings, no overclock or anything. Consider that I was able to capture this on video, because I knew it would happen soon.

The Code 8 crash happens at 0:48 right after I stopped ITB AVX load. This should have been an overheating shutdown instead, but it was not.