darp10: Add custom fan points

[XV-02]

If I'm reading this correctly, we're moving the fans between 27% and 77% duty cycles across the thermal range?

If that is correct, why are we capping at 77%?

[crawfxrd]

It's harder to make our fan tables now by only looking at the Clevo tables since:

• They use a ladder (10 levels) instead of a linear scale
• They use target RPM instead of PWM duty
• They no longer document noise generated

Since we don't support target RPM, I used the average between 2 of the 6 profiles that Clevo provides as the duty.

• Fan1 max: 7700 RPM
• Performance, level 9: 7700 RPM
• Quiet, level 9: 5400 RPM

I don't know the actual duty required for the 7700 RPM target, so I considered it 100%.

[XV-02]

For the fan curves, where are we actually sampling the temps? are we using the "Package" temps or the "CPU" temp as would be reported with sensors?

[crawfxrd]

We use PECI to read the package temperature every 250+ ms.

• src/board/system76/common/peci.c
• src/board/system76/common/main.c

From the PECI user guide:

• The GetTemp() command is used to retrieve the maximum die temperature from a target PECI address.
• The processor temperature is averaged over 256 ms.

[XV-02]

Okay, looks like the Max fan speeds are ~9000rpm +/- ~10% on the lab units. The duty cycle is working, and package/for temps are successfully driving fan speeds at the desired duty.

Do you want to adjust the curves in light of different values for 100% duty than you expected? For reference, in a 25⁰C environment, building kernels and running stress loads aren't pulling temps much above 80⁰C, or fan speeds much above ~6200 rpm. If not, I'm comfortable to approve this.

[crawfxrd]

No, I'll leave it as-is. The primary concern with moving away from the default points on all boards is having the fans cool below 70⁰C.

I want to look into #453 more which will change how all this is defined anyways.