Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

No, Linux is getting lower performance than Windows because distributions aren't shipping the tooling needed to inform the firmware that the OS can manage platform thermals, and so the firmware is defaulting to a safe state. Of course, the reason they're failing to do so is because Intel never publicly documented any of this.


I don't understand what's so difficult about this.

Why is Intel not providing the documentation? And what is it so difficult to reverse engineer?

Is this tool correctly reading the DPTF data and correcting for it?


> I don't understand what's so difficult about this.

It requires parsing multiple undocumented binary file formats, making sense of how the information in each is interlinked, and writing code that keeps track of system state, compares it against the firmware policy, and then imposes the appropriate constraints.

> Why is Intel not providing the documentation?

I have no idea.

> And what is it so difficult to reverse engineer?

https://github.com/mjg59/thermal_daemon/compare/028bde5bf0f1... was my initial implementation of this (remember to click "Load diff" for anything that's too large for github to display by default). There's a fair amount of complexity here!

> Is this tool correctly reading the DPTF data and correcting for it?

Thermald reads the DPTF data, throttled doesn't. I can't promise that thermald reads it correctly given the lack of a spec, but to the best of my knowledge it does.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: