Jump to content

10.12 PB Haswell-E Clover kernel patching for XCPM

matryska

By matryska,
in OSx86 10.12 (Sierra)

 Share
12 posts in this topic

Recommended Posts

matryska
matryska
Posted
Posted

Thanks to Pike, I was able to get Clover kernel patching on the fly working for xcpm. I expect this should work for the developer betas too.

E5-1650 V3 @ 4Ghz

X99E-ITX/ac

 

10.12 public beta (16A293a)

no FakeCPUID

​Currently testing with iMac 14,2 SMBIOS but I intend to revert to Mac Pro 6,1 with freqVectorsEdit.sh

 

Haswell E > Haswell

for other processors change as per https://pikeralpha.wordpress.com/2016/07/26/xcpm-for-unsupported-processor/

 

_xcpm_bootstrap 10.12 © Pike R. Alpha, 0xC4 > 0xC1

F: 83C3C483FB22
R: 83C3C183FB22

_xcpm_cpuid_set_info 10.12 © Pike R. Alpha, 0xE9 > 0xE6

F: 0FB6C483C0E983F847
R: 0FB6C483C0E683F847

_xcpm_idle MSR (0xE2) locked 10.12 © Pike R. Alpha, 0x0F30 > 0x9090

F: 20B9E20000000F30
R: 20B9E20000009090

xcpm reboot 10.12 © Pike R. Alpha, 0x55 > 0xC3 (ret) # Additionally (if booting results in a immediate reboot)

F: 554889E5415741564155415453504189D64189F74889FB45
R: C39089E5415741564155415453504189D64189F74889FB45

sysctl machdep.xcpm reveals

machdep.xcpm.mode: 1
machdep.xcpm.vectors_loaded_count: 1

Turbo boost does not reach flatline. This is noticeable with GeekBench. I have yet to find a solution; this is no different to 10.11.

Flatline turbo boost can be achieved with NullCPUManagement + okrasit's Msr199.kext

 

Update:

Wow. It works. okrasit has patched the kernel to fix the flatline issue as well. 

 

xcpm performance fix 1 © okrasit

F: 89D8C1E008B99901
R: B800FF0000B99901

xcpm performance fix 2 © okrasit

F: 4489E8C1E008B99901
R: 90B800FF0000B99901

xcpm performance fix 3 © okrasit

F: 4489F8C1E008B99901
R: 90B800FF0000B99901
  • Like 2
Link to comment
Share on other sites
nmano
nmano
Posted
Posted

I try this for DP6

F: 89D8C1E008B99901

R: B800FF0000B99901

 

F-90 B8 00 FF 00 00 B9 99 01

R-90 B8 00 FF 00 00 B9 99 01

 

F-00 00 4C 8B 75 08 B9 99 01

R-90B800FF0000B99901

 

This is like okrasit patched but XCPM not worked.

Geekbench score better.

Link to comment
Share on other sites
matryska
matryska
Posted
  • Author
Posted

F-90 B8 00 FF 00 00 B9 99 01

R-90 B8 00 FF 00 00 B9 99 01

 

This code does not appear to be patching anything. With the other 2 entries instead of okrasit's code, I retain XCPM / frequency vectors and Geekbench score is the same with turbo flatline.

 

Are you looking to get XCPM working? If so we will need the results of the following:

kextstat -l | grep -y applelpc
kextstat -l | grep -y x86plat
kextstat -l | grep -y acpiplat
Link to comment
Share on other sites
ergot
ergot
Posted
Posted

I will try the patches, so that mean we do not need NullCPUManagement + okrasit's Msr199.kext? by the way where u can find Msr199 kext?

 

kextstat -l | grep -y applelpc is not giving me any output, the other 2 give me an output...how to verify that applelpc is loaded?

 

 

kextstat -l | grep -y x86plat:

90 1 0xffffff7f81cb7000 0x17000 0x17000 com.apple.driver.X86PlatformPlugin (1.0.0) 737E5937-A52E-32C0-A6C2-4C026B96848C <89 81 62 39 11 7 6 5 4 3 1>

126 1 0xffffff7f81cce000 0x7000 0x7000 com.apple.driver.X86PlatformShim (1.0.0) 60DA68BB-E888-37A1-BCA6-9930A920A266 <90 89 81 7 4 3>

 

kextstat -l | grep -y acpiplat:

13 2 0xffffff7f83350000 0x60000 0x60000 com.apple.driver.AppleACPIPlatform (5.0) D748FA50-380C-3F0D-BAD9-EB7A15848EA8 <12 11 7 6 5 4 3 1>

 

 

 

Thanks

Link to comment
Share on other sites
matryska
matryska
Posted
  • Author
Posted

I will try the patches, so that mean we do not need NullCPUManagement + okrasit's Msr199.kext? by the way where u can find Msr199 kext?

 

patching AppleLPC in clover now AppleLPC is loaded...

Yes. We no longer need NullCPU/Msr199.kext. This kext was posted by okrasit on this forum, perhaps have a look via Google but you shouldn't need it.

 

Now that AppleLPC is loading for you, is it working? You should be getting flatlines at peak multiplier.

Link to comment
Share on other sites
matryska
matryska
Posted
  • Author
Posted

So what do I need to do to run this?

 

My i7 5820k with another person with same mobo/cpu and near same OC's SSDT gave me perfect flat line in El Cap, same clover/etc I get less mult and never peak OC mult. 

 

You need to apply those patches via Clover Configurator and it should give you peak multi with flatlines

 

I am confused exactly how to apply this patch. I have a E5-2630v3 and I'm running Yosemite. Any way I can get this to work for me so I can make my processor always be turboin'?

 

I've only seen patches for 10.11 and 10.12. I don't have any test systems with Yosemite so I'm no help I'm afraid. Would recommend upgrading to a newer OS if possible.

  • Like 1
Link to comment
Share on other sites
cmanns
cmanns
Posted
Posted

You need to apply those patches via Clover Configurator and it should give you peak multi with flatlines

 

 

I've only seen patches for 10.11 and 10.12. I don't have any test systems with Yosemite so I'm no help I'm afraid. Would recommend upgrading to a newer OS if possible.

oh okay sweet! 

 

My cpu is doing x18 and x30 mostly with maybe x40 on geek bench. Should be doing x12 -> x44 using someone else's SSDT that has same cpu/mainboard/overclock and it works flawless on El Cap. Will def try this out and report back

Link to comment
Share on other sites
cmanns
cmanns
Posted
Posted (edited)

I put xcpm performance fix 3  & all the other fixes above (Some my clover config already had) for 10.12.x from whoevers config am using. 

 

machdep.xcpm.vectors_loaded_count: 1 = 0 for me

I only hit X419 mult, normally sits x18 to x30 even during geekbench. 

In El Capitan mult's go x12 to x44~ and it will flatline x44 give or take during geekbench. 

Scores aren't that much different oddly

 

 

Williams-Mac-Pro:~ cmanns$ kextstat | grep AppleIntelCPU
Williams-Mac-Pro:~ cmanns$ kextstat | grep AppleIntel
   78    0 0xffffff7f82620000 0x2000     0x2000     com.apple.driver.AppleIntelSlowAdaptiveClocking (4.0.0) 6FE984DD-A1FE-309E-83CF-B346989A6F17 <77 4 3>
Williams-Mac-Pro:~ cmanns$ kextstat -l | grep -y applelpc
Williams-Mac-Pro:~ cmanns$ kextstat -l | grep -y x86plat
   75    1 0xffffff7f81b81000 0x17000    0x17000    com.apple.driver.X86PlatformPlugin (1.0.0) F80CF40D-4ADE-3F20-AB94-B5EAB9B898E6 <74 73 43 42 11 7 6 5 4 3 1>
  114    1 0xffffff7f81b98000 0x7000     0x7000     com.apple.driver.X86PlatformShim (1.0.0) 4AAFC86D-5E06-31C5-A6C5-4F00C3A04F34 <75 74 73 7 4 3>
Williams-Mac-Pro:~ cmanns$ kextstat -l | grep -y acpiplat
Edited by cmanns
Link to comment
Share on other sites
  • 10 months later...
damnsoul
damnsoul
Posted
Posted

so sad ,the patch not work for me !

I also have the same cpu, I don't know why not work for me. the Pike's tutorial is hard to me, I hope someone can help me solve this problem.

 

my computer list:

Gigabyte x99 phoenix sli

E5-1650 v3

MSI GTX 1070

16GB DDR4 2133 Recc x2

Samsung SM951 512GB

 

the kextstat message:

kextstat -l | grep -y applelpc
  102    0 0xffffff7f82a06000 0x3000     0x3000     com.apple.driver.AppleLPC (3.1) F51595F0-F9B1-3B85-A1C3-F984DAD4107E <88 12 5 4 3>
kextstat -l | grep -y x86plat
   89    1 0xffffff7f82082000 0x17000    0x17000    com.apple.driver.X86PlatformPlugin (1.0.0) 7A4289C8-1162-389D-8FBF-99730287DB72 <88 87 51 19 11 7 6 5 4 3 1>
  125    1 0xffffff7f82099000 0x7000     0x7000     com.apple.driver.X86PlatformShim (1.0.0) 8E873270-42BC-309E-ACE0-EAE2333623C6 <89 88 87 7 4 3>
kextstat -l | grep -y acpiplat
   13    1 0xffffff7f82f03000 0x60000    0x60000    com.apple.driver.AppleACPIPlatform (5.0) 65E05472-6AE7-3308-8CC8-FA6CB0DB2AEE <12 11 7 6 5 4 3 1>

kextstat&AppleIntelInfo0806.txt

Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...