Contribute
Register

[Guide] Dell XPS 13 9360 on MacOS Sierra 10.12.x - LTS (Long-Term Support) Guide

Status
Not open for further replies.
bozma88 said:
Yes, but what is controlling PM, then? HWP?
Why is HWP enabled? This is a fresh install where I never used any HWP enabling method.
I did use the Clover HWP flag in a prior test install. Does it set a permanent flag in a CPU register?
I'm quite intrigued.
Click to expand...

Hardware controlled PM is enabled with HWP (Speed Shift). I didn't know it would affect the IGPU though (thought it was just CPU)... but I haven't really read about it yet.
 
RehabMan said:
Hardware controlled PM is enabled with HWP (Speed Shift). I didn't know it would affect the IGPU though (thought it was just CPU)... but I haven't really read about it yet.
Click to expand...

Yes, but where did I enable Speedshift?
What's the sense of Clover flag or HWPEnabler (and why everyone is talking about them) if Speedshift is enabled by default?

BTW, in my opinion, in Kaby Lake Speedshift starts to become preferable, because of the ultra-short frequency rise time compared to OS-driven P-states. Kaby Lake halves the already short frequency rise time of SKL Speedshift.
 
bozma88 said:
Yes, but where did I enable Speedshift?
What's the sense of Clover flag or HWPEnabler (and why everyone is talking about them) if Speedshift is enabled by default?
Click to expand...

There is a Clover option for HWP... and a kext... I haven't looked into it in detail.
If you didn't use it (I didn't check), then the effects you're seeing are just PM in the kernel (XCPM).

BTW, in my opinion, in Kaby Lake Speedshift starts to become preferable, because of the ultra-short frequency rise time compared to OS-driven P-states. Kaby Lake halves the already short frequency rise time of SKL Speedshift.
Click to expand...

And I suppose it is possible Apple will start using it at some point...
 
RehabMan said:
Not likely. Maybe AppleIntelInfo.kext outputs Speed Shift status?
Click to expand...
If you tell me where/how to look, I will look and report.
What I am seeing is a very snappy system and correct reporting in IPG.
The system is able to go TDP-up for 10 seconds (25W) and then settles at 15W full load.

If you want to dive deeper I can help.


In my AppleIntelInfo dump I see this:

Code: 
IA32_TSC_DEADLINE................(0x6E0) : 0x4357780B5C
(HWP Supported but not (yet) enabled)


__
~ I'm reading Pike's post with attention, will report back.
 
Last edited:
I completely disabled SpeedStep and TurboBoost in BIOS.
The system seems even more aggressive in frequencies rise-fall.
(?!)

Screen Shot 2017-01-28 at 16.44.38.png

This is the diff of AppleIntelInfo dumps with SpeedStep disabled | enabled in BIOS: https://www.diffchecker.com/Xh6Qb062
 
bozma88 said:
If you tell me where/how to look, I will look and report.
What I am seeing is a very snappy system and correct reporting in IPG.
The system is able to go TDP-up for 10 seconds (25W) and then settles at 15W full load.

If you want to dive deeper I can help.


In my AppleIntelInfo dump I see this:

Code: 
IA32_TSC_DEADLINE................(0x6E0) : 0x4357780B5C
(HWP Supported but not (yet) enabled)


__
~ I'm reading Pike's post with attention, will report back.
Click to expand...

I have no time to look into it now (eventually, I'll look into it on my Skylake NUC).
You can probably find information in Intel CPU data sheets that will tell you which MSR is tweaked (not to mention Clover and HWPEnabler source code).
From there you should be able to determine if/where AppleIntelInfo.kext is reporting this information.
 
RehabMan said:
I have no time to look into it now (eventually, I'll look into it on my Skylake NUC).
You can probably find information in Intel CPU data sheets that will tell you which MSR is tweaked (not to mention Clover and HWPEnabler source code).
From there you should be able to determine if/where AppleIntelInfo.kext is reporting this information.
Click to expand...

Done.
The MSR responsible for toggling HWP is IA32_PM_ENABLE, and on my system is not enabled (0), as AppleIntelInfo dump reports.
So, what's driving the power management I am seeing, even with SpeedStep and TurboBoost disabled in BIOS and no custom SSDTs?
 
bozma88 said:
So, what's driving the power management I am seeing, even with SpeedStep and TurboBoost disabled in BIOS and no custom SSDTs?
Click to expand...

XCPM in the kernel, and perhaps ACPI_SMC_PlatformPlugin is involved.
 
Status
Not open for further replies.
Copyright © 2010 - 2024 tonymacx86 LLC
Back
Top