Important: 10.7.4 and Sandy Bridge CPU PM Temporary Fix

[artgub]

...
Page 409 of the ACPI Spec.

Package {
?CoreFrequency                 // Integer (DWORD) 
Power                        // Integer (DWORD) 
Latency                      // Integer (DWORD) 
BusMasterLatency            // Integer (DWORD) 
Control                      // Integer (DWORD)
Status                       // Integer (DWORD)
}

Oh thanks! One little question more.. Where should I get those Power values for my 2600k? Couldn't find it in Intel data sheets.
Thanks.

Good question, here is what I did and that doesn't mean it's the correct way but it did seem logical and logic is my god. I compared the power values for the existing entries 0x1800,0x1700,0x1600,0x1500 and 0x1400 and here is what they were in decimal 247000,224000,202000,182000, and 163000. So I subtracted 247 from 224 and got 23. Then 224 from 202 and got 22 etc.. The end result from subtracting each one results in the following difference values. 23000,22000,20000,19000 (you get the gist). So to get the value for 0x1300 I subtracted 17000 from 163000 and got 146000 and did a similar for the other two. Using this I came up with the following values for 0x1300, 0x1200, and 0x1100. 146000, 130000, and 116000. Which in hex is 23A50, 1FBD0, and 1C520 respectively. So that is what I used and it works, is it right beats me, but it works.
MSRDumper PStatesReached: 16 20 24 29 33 34 35 36

Well. Nice observation.
But when I'm solving my progression this way I'm getting 83000 power value for 0x00001000 state when it should be 80000 (I guess) (I have this value in my dump).
Thoughts?
Thank you.

...
Package (0x06)
{
    0x000008FC, 
    0x00031510, 
    0x0000000A, 
    0x0000000A, 
    0x00001700, 
    0x00001700
}, 
Package (0x06)
{
    0x00000898, 
    0x0002C6F0, 
    0x0000000A, 
    0x0000000A, 
    0x00001600, 
    0x00001600
}, 
Package (0x06)
{
    0x00000834, 
    0x00027CB8, 
    0x0000000A, 
    0x0000000A, 
    0x00001500, 
    0x00001500
},

// oops

Package (0x06)
{
    0x00000640, 
    0x00013880, 
    0x0000000A, 
    0x0000000A, 
    0x00001000, 
    0x00001000
}
..

 

[intamara]

OK so I fixed cores 1-3in my SSDT, and I am able to boot with stepping. Well sort of.

10.7.3

May 11 13:07:08 Mac-Pro kernel[0]: MSRDumper CoreMulti(16)
May 11 13:07:08 Mac-Pro kernel[0]: MSRDumper PStatesReached: 16 23 24 25 26 40

10.7.4

May 12 14:53:47 Mac-Pro kernel[0]: MSRDumper CoreMulti(16) 
May 12 14:53:47 Mac-Pro kernel[0]: MSRDumper PStatesReached: 16 17 18 19 20 40

Not really the stepping that I was hoping for. Almost pointless. Anybody have any ideas?

Try to set as MacBook Pro 8,3

mine: not overclock
10.7.4
MacPro 3,1
May 13 08:50:22 localhost kernel[0]: MSRDumper CoreMulti(16)
May 13 08:50:22 localhost kernel[0]: MSRDumper PStatesReached: 16 17 18 19 20 34 35 36

MacBook Pro 8,3
May 13 08:57:36 localhost kernel[0]: MSRDumper CoreMulti(16)
May 13 08:57:36 localhost kernel[0]: MSRDumper PStatesReached: 16 20 24 29 33 34 35 36

iMac 12,2
I have read someone's post @ insanelymac said it has no speed step

May 13 09:14:23 localhost kernel[0]: MSRDumper CoreMulti(16)
May 13 09:14:23 localhost kernel[0]: MSRDumper PStatesReached: 16 34 35 36

 

[replicant]

Ok, so I made the edit to the plist and rebooted; Opened up HWMonitor.app and I see that I'm still at a 16x multiplier.
I have a Gigabyte GA-Z77M-D3H with Intel i5-2500K.

I'm sure I'm missing something; is there a way to verify that things are working as they should? Running the demo of Geekbench (32bit mode only) showed a minimal increase in the score (about a thousand points) after I made that change to the plist.

Thanks much.

 

[sirio688]

Hi, installed 10.7.4, restarted, Kernel Panic
-x -v -s , Kernel Panic
Please Help Me !

 

[Sav]

Hi, I just upgraded.
My motherboard is a Gigabyte Z68X-UD3H-B3, with the F10 bios and the custom DSDT I downloaded from the DSDT archive.
I lost audio, trim support and, of course, the cpu performances were cut by 50%.
With multibeast I managed to reinstall HDA kext and the trim support.
For the cpu issue I followed the instructions on the first post:

1) I edited the org.chameleon.Boot.plist

2) set the bios with: Real-Time Ratio Changes in OS=Disabled - CPU EIST Function = Enabled

To be sure, I also disabled the C1 state in the bios settings.

Now the geekbench score is ALMOST back to what it was before (I lost 200 points).

 

[LosChristianos]

Hi, installed 10.7.4, restarted, Kernel Panic
-x -v -s , Kernel Panic
Please Help Me !

Looks like you need to re-install NullCPUPowerManagement.kext from MultiBeast.

 

[LosChristianos]

I guess we need to start an SSDT database and write a tutorial on how to edit it for each particular processor.

I'm going to have a go, i7 2600K stock speed, no overclocking.

Well, I had a go at this; and much to my surprise, it worked.

MSRDumper CoreMulti(36) 
MSRDumper PStatesReached: 16 23 35 36 37 38

Wow! I'm shocked - my little locked processor is happy to turbo boost to 38x for single threaded load and happily runs at 35x under full load. It seems that under 10.7.4 turbo boost is being triggerred more aggressively as Intel specified. Perhaps even better then windoze. Neat-o!

** I used the SSDT.AML attached to a previous message here -- no mods; I tried it on a whim.

Well I tried the same SSDT and it works

P-States reached: 16 23 25 26 35 36 37

No doubt it could be better but it's better than 16 and 35 only.

One thing I did notice running MSRDumper opening Adobe apps is the P-State would stay at 16, quit and re-open the app it would jump to 35.

Now it jumps to 23 or 26, apps opening super fast.

GeekBench score at 11834

Cinebench test results same as in 10.7.3

Will definitely try and use an SSDT that is tailored to my setup at some point.

So in short, if your run an i7 2600 be it locked or unlocked it's worth trying Greggen's kindly supplied SSDT found on page 9.

I'm running an i7 2600K (stock speed no OC) on a Z68X-UD3H-B3 F10.

Thanks

 

[Wolfster]

I can also confirm that Greggen's SSDT worked for me(i7 2600k) when upgrading to 10.7.4. I just put it my extra folder.
My geek bench scores were at an all time low of 6000 ish and now they are back up at 12700.
I haven't checked which states are being used, as Im not sure where to check so if someone could let me know that would be appreciated. However, checking HW monitor, it seems to only stay at x16 and x20 most of time and then to x38 when opening a new application though so Im not sure it working perfectly.

 

[pehera]

I can also confirm that Greggen's SSDT worked for me(i7 2600k) when upgrading to 10.7.4. I just put it my extra folder.
My geek bench scores were at an all time low of 6000 ish and now they are back up at 12700.
I haven't checked which states are being used, as Im not sure where to check so if someone could let me know that would be appreciated. However, checking HW monitor, it seems to only stay at x16 and x20 most of time and then to x38 when opening a new application though so Im not sure it working perfectly.

I have the same MB and CPU but if i use that SSDT my lovest multiplyer is x19 and highest x42.
Are you using Drop SSDT in org.chameleon.Boot.plist ?

 

[Sav]

Just tried Greggen's SSDT, by leaving the previous bios settings I posted before, with my original org.chameleon.Boot.plist.
My I5 2500 went from 16 up to 36 pstates and the geekbench test reached an higher result.
Thanks guys