i7 3770k Ivybridge Power Management

[Hackmodford]

I've been trying to get power management working correctly.

 

If I use the imac13,2 profile I get 2 pstates (1.6 and 3.9)

 

If I use -xcpm I will get around 7 pstates but I do get errors in the console.

 

My question is this. Does a desktop machine need a range of pstates? Or are the 2 just fine?

 

Also I found this post that describes patching the appleintelcpupowermanagement.kext. Is this necessary?

http://myhack.sojugarden.com/2012/08/my-new-hackintosh/

sudo perl -pi -e ‘s|\xE2\x00\x00\x00\x0F\x30|\xE2\x00\x00\x00\x90\x90|g’ /Extra/Extensions/AppleIntelCPUPowerManagement.kext/Contents/MacOS/AppleIntelCPUPowerManagement

[Allan]

You need generate correct power managent.

 

Try create SSDT.

curl -o ssdtPRGen.sh https://raw.github.com/Piker-Alpha/RevoBoot/clang/i386/libsaio/acpi/Tools/ssdtPRGen.sh
chmod +x ssdtPRGen.sh
./ssdtPRGen.sh

and

sudo ./ssdtPRGen.sh 77 3900

instead of applying the patch AICPM. use this in config.plist

<key>KernelAndKextPatches</key>
<key>AsusAICPUPM</key>
<true/>

is the same effect as applying the patch AICPM.

 

download the AppleIntelCPUPowerManagementInfo.kext and add it to your System/Library/Extensions/ folder. Rebuild cache with the app of your choice and reboot.

 

At start open your terminal app and type this command :

cat /var/log/system.log | grep "AICPUPMI:"

If you can see a lot of P-States, you did a good job.

 

test

 

Credits: Whit3Spirit

[Hackmodford]

AppleIntelCPUPowerManagementInfo.kext caused a kernel panic for me

But I did change the AsusAICPUPM setting in clover.

I have so far seen 1.6, 3.5, 3.7, 3.9 in HW Monitor

 

Strange thing is MSRDumper only reports 16 and 39

[Allan]

Congrats, this shows that your States is working.

 

look about your CPU: i7 3770K

 

you can try applying the patch in AICPM, but then removing AsusAICPUPM, do not forget to backup the original kext.

 

but if this method worked for you, the patch is not necessary.

[Hackmodford]

@Allan

 

So even though MSRDumper is reporting only two states. The fact that I can see a few more via HWMonitor means it's working?

Most of the time it's either 1.6 or 3.9 and VERY rarely do I see stuff like 3.5,3.7,3.8

 

My original question still stands, do the pstates really make that much of a difference for a desktop machine?

[Allan]

run Geekbench (64 bits), and see HWMonitor.

 

pstates make much difference on the desktop machine.

 

besides better performance, will prevent fraying Hardware.

[Hackmodford]

Well if I did that HWMonitor would just show 3.9 pretty much constantly.

 

The other "stress test" I use is firing up the Dolphin Gamecube emulator

[Allan]

Please take a screenshot of HWMonitor.

[Pimentel]

Hello

 

Take a look in my guide: http://www.insanelymac.com/forum/topic/295587-power-management-for-sandy-and-ivy-bridger/

 

if you're using ML 10.8.5 or MV... try add this boot flag(chameleon)/argument(Clover): -xcpm to get more P-States

[bronxteck]

this still works http://www.insanelymac.com/forum/topic/295200-testers-needed-cpu-power-management-for-sb-and-ib-xeons-or-i7-39xx-on-x79-or-c60x-chipset/?p=1990582 i guess no one on that thread cared to test for themselves.

[Hackmodford]

Is there anything special I have to do to use voodopstate.kext? Do I have to remove any kexts? That seems kinda interesting.

 

The reason I'm not using -xcpm is because of errors I get in the console. (though it appears to be working 100% the errors are not right according to Pike)

[bronxteck]

no nothing that i know of has to be removed... you control it via the app. 

[Hackmodford]

I wonder if I could use it and keep the macpro3,1 profile to fix the audio popping issue?

 

I'd much rather be in control of the speedstepping than try to match it to an existing Mac...

[Hackmodford]

voodoopstates.kext doesn't work for me

 

Whatever... I'm using the macpro3,1 profile and enabled xcpm. Pike said it doesn't work for mac pro but it seems to be working fine even though I do have errors in the console.

[bronxteck]

try adding p and c states option via chameleon. if not try the _CST option with the kext.

[Gringo Vermelho]

My original question still stands, do the pstates really make that much of a difference for a desktop machine?

At the bottom, HWMonitor shows how many watts your CPU is drawing. You should see the wattage going up and down as the operating frequency goes up and down. But it's not just about saving on the electricity bill, it's also about heat.

 

Having your CPU clock actively regulated according to what it is actually doing, as opposed to constantly running at full tilt, means less energy is wasted, less heat is generated, which translates to longer component life. Next to yourself, heat is your PCs worst enemy lol

 

So yes, definitely great for desktops.

 

Laptops normally have more "steps" than desktops, which helps to improve battery life. Also laptops are much more difficult to cool so you'll want tighter regulation there.

[Hackmodford]

try adding p and c states option via chameleon. if not try the _CST option with the kext.

 

I'm using a generated SSDT.

So in clover I have "dropoemssdt" on and turned off generate c/p states.

 

What is the _CST option?

--------------------

Basically it seems with the iMac option it's either 1.6 or 3.9

[bronxteck]

those options are for chameleon as that's what i am using. i have yet to get clover configured properly. the wiki is a bit outdated it is not keeping up with all the major improvements made daily.

[Hackmodford]

Well I'm back to using -xcpm this is what the console looks like when searching for xcpm

2/14/14 12:04:18.000 PM kernel[0]: XCPM: registered
2/14/14 12:04:19.000 PM kernel[0]: IOPPF: XCPM mode
2/14/14 12:04:19.000 PM kernel[0]: XCPM: P-state table mismatch (error:0x12)
2/14/14 12:04:19.000 PM kernel[0]: X86PlatformShim::sendPStates - pmCPUControl (XCPMIO_SETPSTATETABLE) returned 0x12

Everything seems to be working fine. I'm starting to think that the last two are just a kernel bug...

 

Update: I just searched for x86platformShim and found these also.

2/14/14 12:04:19.000 PM kernel[0]: X86PlatformShim::start - Failed to send PStates
2/14/14 12:04:19.000 PM kernel[0]: X86PlatformShim::start - Failed to send stepper

But still, I can't find anything wrong...

[Gringo Vermelho]

That was my impression as well, everything seems to work fine even with those messages.

 

The only reason that I don't use -xcpm is...that I don't like error messages in my logs.. ridiculous I know..

[Hackmodford]

Dude it's our famous person OCD issues 

 

I might even go back to using the macpro profile to fix audio popping and using xcpm, even though he said it doesn't work. It works fine for me 

 

So using the macpro3,1 profile (to get rid of audio pop)

and using my custom ssdt specifically made for xcpm.

and using the xcpm flag

2/14/14 3:38:16.000 PM kernel[0]: XCPM: registered
2/14/14 3:38:16.000 PM kernel[0]: XCPM: registered

(interesting no "IOPPF: XCPM mode" message but no errors either)

 

Here's an example of the Hardware Monitor graph.

 

 

[gils83]

 

Xeon E3-1245 v2 ivybridge at 3,4

 

 

 

 

[Hackmodford]

Okay I found a problem. If I wake from sleep while using xcpm the computer will spontaneously restart soon after.

 

Here's HWMonitor using the normal imac13,2 PM.

[gils83]

 

for me, I used no patch, my bios is set by default and I do not use DSDT  

 

GA-Z68-AP-D3

[Hackmodford]

I jumped over to the windows side to see how it handled the speed stepping. I observed that Windows also just goes between 1.6 and 3.9 as a general rule.

 

So I'm going to say the correct answer for me was this

Use imac13,2 smbios (because of my graphics card)

No -xcpm

Generated ssdt from piker's script.

DropOEMSSDT = True

GenerateCstates = False

GeneratePStates = False