Contribute
Register

Skylake HWP Enable

Status
Not open for further replies.
1. It is hard to say which (HWP or XCPM control) is better, if base on performance, not know. If base on power consumption then absolutely HWP. Unlike previously, PM ssdt is very important to CPU. Now the main purpose of PM ssdt is to ensure x86platformplugin is loaded.

2. Previous version of Clover will stop to generate CpuPm ssdt if HWP is enable, therefore we use ssdtprgen.sh. Recently version will generate a null CpuPm ssdt (without any PStates) if HWP is enable, that make sure x86platformplugin is loaded.

3, I do not think it will harm your CPU, if you are running Sierra, basically no need to input HWPvalue. Since Sierra 10.12.2, I no longer input the HWP value, HWP seem to complete optimised by Apple. May be HWP will be standard on next version of MacOS.

Thank you for your kind explanation, now it makes more sense to me. So right now, I enabled the HWP with clover configurator without putting any values. I do not check the -xcpm boot flag. I still have ssdt.aml in my EFI - clover - patched folder. I think this is all right? Is there anything else that I need to do? (I am using a i7-6700k)

Thank you shuhung
 
Thank you for your kind explanation, now it makes more sense to me. So right now, I enabled the HWP with clover configurator without putting any values. I do not check the -xcpm boot flag. I still have ssdt.aml in my EFI - clover - patched folder. I think this is all right? Is there anything else that I need to do? (I am using a i7-6700k)

Thank you shuhung

Uncheck -xcpm from Clover is useless, XCPM mode still on, it equal to XCPM plus HWP, unless you willing to give up XCPM control, otherwise you may not find out the difference.

The only way to disable XCPM is use FackeCPUID like 0x0306a0 and add NullCPUPowerManagement.kext. It is very funny (totally instane), I always avoid to use this kext which complete disable CPU speedstep function (except Skylake Kabylake) .
 

Attachments

  • HWP.zip
    2 MB · Views: 316
Uncheck -xcpm from Clover is useless, XCPM mode still on, it equal to XCPM plus HWP, unless you willing to give up XCPM control, otherwise you may not find out the difference.

The only way to disable XCPM is use FackeCPUID like 0x0306a0 and add NullCPUPowerManagement.kext. It is very funny (totally instane), I always avoid to use this kext which complete disable CPU speedstep function (except Skylake Kabylake) .

This is very interesting, great work. Normally you steer clear away from both the -xcpm and nullCPUPowerManagementkext for Skylake. Interesting to know the Null setting.
 
Uncheck -xcpm from Clover is useless, XCPM mode still on, it equal to XCPM plus HWP, unless you willing to give up XCPM control, otherwise you may not find out the difference.

The only way to disable XCPM is use FackeCPUID like 0x0306a0 and add NullCPUPowerManagement.kext. It is very funny (totally instane), I always avoid to use this kext which complete disable CPU speedstep function (except Skylake Kabylake) .

Thank you for your reply, I patched the FakeCPUID, and disable xcpm(it wasnt enabled before anyways). and I use clover to enable the power management to get x86platformplugin. Everything seems to be working but how do I know if the HWP is working? I ran Geekbench and got a little higher score on multicore but lower on single core. Is that normal?

Thanks :)
 
it seems like the front green numbers are how aggressive speed step is, in the intel power gadget going from 20 to 40 makes the graph a lot more spiky and shorter together. faster speeds steps = increase power efficiency, it shouldn't result in lower performance but I haven't done any testing.

it raises and lowers the clock speed faster it seems. I only use macOS for productivity stuff so 40 seems fine.
 
Last edited:
Thank you for your reply, I patched the FakeCPUID, and disable xcpm(it wasnt enabled before anyways). and I use clover to enable the power management to get x86platformplugin. Everything seems to be working but how do I know if the HWP is working? I ran Geekbench and got a little higher score on multicore but lower on single core. Is that normal?

Thanks :)


It does not gain more performance but get better power consumption, less frequency fluctuation on each load steps, higher CPU utilization on each speed step.

You can try difference HWP values and compare to each loadings (by observation the activity monitor and Intel Power Gadget), with lower frequency and higher CPU utilization on each load step should be better.

May slightly reduce performance during light or medium loads but have significantly power saving.
 
I'm still following this thread without changed to HWP on my Skylake system.
I still haven't understood what "ssdt" is?

The range is too wide, it is hard to explain what “ssdt” is, you should google it. In this post “ssdt” is used for power control.

If you do not have a customer ssdt.aml (CpuPm), let Clover do it for you, just check generate PStates, CStates.

You should ensure that native power management is function correct first, although you feel that your system is running fine.

1. Did X86PlatformPlug loaded

2. Did AppleLPC loaded
 
You should ensure that native power management is function correct first, although you feel that your system is running fine.

1. Did X86PlatformPlug loaded

2. Did AppleLPC loaded

How do you check to see of they are loaded?
 

Attachments

  • AppleLPC.png
    AppleLPC.png
    10.4 KB · Views: 704
  • x86platformplugin.png
    x86platformplugin.png
    23.3 KB · Views: 663
I followed these instructions and had some success. I have a cool and stable overclock the 4.4GHz. The only downer is that my "About this mac" speed has dropped from a claimed 4.01 GHz to 1.99 GHz. Benchmarks and the Intel Power Gadget show it slammed at 4.4GHz, 27.5C, 5.7W with 20 and 2C.

Thanks for giving me a working OC. I just wish for vanity's sake that the computer would own it ;)
 
Status
Not open for further replies.
Back
Top