Kernel Task stuck at 15% when finger rests on touchpad

[adityahota]

I see with the latest release that kernel task is taking up about 15% of CPU usage when my finger is resting on the trackpad. This was fixed earlier in a release but it looks like something has regressed.

Also, do you know if there is a way to disable the touchscreen as with the previous AML replacement?

[xxxzc]

Releases except 201208(use VoodooI2C 2.5.2) are using the same VoodooI2C 2.4.4 and SSDT-I2CX(SSDT-TPDX renamed).

You can copy https://raw.githubusercontent.com/xxxzc/xps15-9570-macos/ced5c266dc29c6c3de0efc9e87458f3e10677760/ACPI/SSDT-TPDX.dsl to SSDT-I2CX and uncomment TPLT = Zero to disable touchscreen.

[adityahota]

~Thanks--I was able to disable the touchscreen. However, it still looks like I am getting between 15 to 20% kernel task usage even at idle. Is anyone else getting that issue?~

~I tried changing the layout-id for AppleALC back to 72, but that didn't seem to affect anything.~

Looks like the usage went down a few minutes after restarting. I will keep an eye out!

[MaKS-NUC]

With the latest release 210110 kernel task is taking up about 14% of CPU usage when the finger is resting on the trackpad and about 12% when is resting on the touchscreen.

[adityahota]

Doing the steps to change SSDT-I2CX.aml (described by XXXZC) fixes the issue for me. I get 4 to 5 percent with my finger on trackpad and 2 percent without. Unfortunately, that disables touchscreen.

let me know if you need a compiled version.

[xxxzc]

@MaKS-NUC @adityahota you can try to add the SSDT in https://github.com/xxxzc/xps15-9570-macos/issues/74#issuecomment-751622506 and remove https://github.com/xxxzc/xps15-9570-macos/blob/master/OC/config.plist#L202-L208 to enable interrupt mode on touchscreen.

[MaKS-NUC]

I tried described changes but the result is the same ~14% kernel task. I will continue with touchscreen disabled ~ 5% kernel task.

[MaKS-NUC]

One more thing:

I managed to get better performance without sacrifice battery life by modified CPUFrienDataProvider.kext, tested with Geekbench5.

With OLD CPUFriendDataProvider.kext the result is: 

With NEW CPUFriendDataProvider.kext the result is 10% better Since-Core Performance:

And using VoltageShift (https://github.com/sicreative/VoltageShift) with these parameters: 

The result is more than 10% in Multi-Core Performance:

 I attached the modified kext for testing.

P.S. The CPU cooling system is repasted with Arctic MX-4 Thermal Compound.

CPUFriendDataProvider.kext.zip