Hi
With the latest linux BSP 2.8b5 you increased the temperature for the critical shut-down to the values from kernel 4.1. In the patch there you wrote that the old values were well verified.
When I understand right, you and Freescale/NXP do not see any risk for the CPU to overheat and get damaged with this value 10 grad over the max value.
Do you see any risks to apply this change with mainline-kernel 4.19.y?
Regard Stefan
Hi
This is a bit of a trick question.
The question is what the purpose of the critical temperature trip point is.
Should the product’s thermal design should ensure that Tj max is never reached and thus the shut down due to over temperature is a last resort thing or does the thermal design rely on the system to shut down whenever it gets to hot?
Of course operating near or above Tj max. for a prolonged time will negatively impact reliability. On the other hand shutting down the system below Tj max. and then waiting for user interaction will depending on the application negatively impact the user experience.
So, for years we had the higher trip points and we and customers did validate their designs with those trip points. Lowering them by 15 degrees Celsius without at least further tests from our side seemed unreasonable.That was the reason for this commit.
We since made additional tests in our temperature cabinet and came to the conclusion that lowering the critical trip point to Tj max. makes a good compromise reflect by this follow up commit.
I see the same risks and advantages in whatever kernel you choose for setting a higher vs. a lower trip point. So you have to judge for your application what values are best.
I probably would consider having temp_passive more than 5 degrees below temp_critical so that the throttling has more time to prevent reaching temp_critical.
Max