K3Rl Posted February 23, 2014 Share Posted February 23, 2014 Folks, I got my system up and running on mavericks quite well with clover boot loader. But i must say I'm a bit confused by the power management: When I take a look into the Intel Power Gadget, my system seems to speedstep nicely between 0.8 and 3.3 ghz on the CPU, and also GPU stepping seems to work fine. But when I run MSRDumper, It seems to me that only the lowest + 2 turbo boost states should actually work: 23.02.14 09:03:46,000 kernel[0]: MSRDumper PStatesReached: 8 31 32 33 At the moment my system runs a patched Kernel bc I was experiencing kernel panics (I know my GA board should not lock the register, but Iit seems to run fine now), Clover is on vanilla config.plist (only with DumpOEM=true and SSDT for HDMI audio) – so GeneratePStates and GenerateCStates are both on. And of course native AppleIntelCPUPowerManagement.kext. I tried a generated SSDT for Power Management but that ended in the same result. So can I trust MSRDumper or the Intel Utility? Or does the CPU even step via front side bus when it detects the missing ability to change multiplicators? Would seem strange to me though. Ideas very much appreciated! Link to comment Share on other sites More sharing options...
Dontview Posted February 23, 2014 Share Posted February 23, 2014 Hey, I have exactly the same situation. My CPU is a haswell i5 4570. I use clover and SSDT for Power Management. Didn’t find any solution so far. 1 Link to comment Share on other sites More sharing options...
Liankao Posted March 1, 2014 Share Posted March 1, 2014 Me,too.I have tried using the ssdt.aml generated by ssdtPRGen.sh,But HWmonitor showed only two states,no matter how I configure clover. My CPU is i5-4200m. Link to comment Share on other sites More sharing options...
lunacory Posted March 12, 2014 Share Posted March 12, 2014 If you open the log file of Intel Power Gadget, you can see steps was only few ones. The graphic is not accurate. My cpu is E3-1230 V3 (3.3G-3.7G), with Gigabyte H87N-WIFI, non-SSDT, the speed steps are 8\33\34\35\36\37 (blue means normal、red means boost) I check My RMBP13's cpu, I5-3210M (2.5G-3.1G), the speed steps are 12/22/25/26/27/28/29/30/31 (blue means normal、red means boost) 2 normal speeddsteps VS 3 normal speedsteps This result is acceptable Link to comment Share on other sites More sharing options...
Dontview Posted March 12, 2014 Share Posted March 12, 2014 Well, but your two CPUs aren't Haswell. Speedstepping with Haswell CPUs is a different case, since they using the osx kernel instead of acpm. Link to comment Share on other sites More sharing options...
Laberlohe Posted March 12, 2014 Share Posted March 12, 2014 Erm, the Xeon E3-1230 V3 is a Haswell cpu (like all E3 v3 // E3 v2 was IvyBridge). Generally speaking it is a i7-4770 without the graphics part... Link to comment Share on other sites More sharing options...
Dontview Posted March 12, 2014 Share Posted March 12, 2014 Erm, the Xeon E3-1230 V3 is a Haswell cpu (like all E3 v3 // E3 v2 was IvyBridge). Generally speaking it is a i7-4770 without the graphics part... You are right - sorry my bad. Link to comment Share on other sites More sharing options...
wastez Posted April 21, 2014 Share Posted April 21, 2014 Thats a simple thing. Mavericks is able to manage the pm over the kernel (xnu CPU Power Management) if the x86platform kext is loaded. The old pm is working over the aicpupm.kext. In the meanwhile there are not so much p states. But p-states are not the hole pm, this is what you see when you are using the intel power gadget because it displays p and c states. If you want to use the old pm you can do two things: Edit your ssdt or use a kernel where xcpm is not integrated. Link to comment Share on other sites More sharing options...
Slice Posted May 21, 2014 Share Posted May 21, 2014 Guys, why use ssdtPRGen.sh if you use Clover? It generates the same tables. The script needed for Chameleon users. 1 Link to comment Share on other sites More sharing options...
Pike R. Alpha Posted May 21, 2014 Share Posted May 21, 2014 Folks, I got my system up and running on mavericks quite well with clover boot loader. But i must say I'm a bit confused by the power management: When I take a look into the Intel Power Gadget, my system seems to speedstep nicely between 0.8 and 3.3 ghz on the CPU, and also GPU stepping seems to work fine. But when I run MSRDumper, It seems to me that only the lowest + 2 turbo boost states should actually work: 23.02.14 09:03:46,000 kernel[0]: MSRDumper PStatesReached: 8 31 32 33 At the moment my system runs a patched Kernel bc I was experiencing kernel panics (I know my GA board should not lock the register, but Iit seems to run fine now), Clover is on vanilla config.plist (only with DumpOEM=true and SSDT for HDMI audio) – so GeneratePStates and GenerateCStates are both on. And of course native AppleIntelCPUPowerManagement.kext. I tried a generated SSDT for Power Management but that ended in the same result. So can I trust MSRDumper or the Intel Utility? Or does the CPU even step via front side bus when it detects the missing ability to change multiplicators? Would seem strange to me though. Ideas very much appreciated! Haswell processors do not use AppleIntelCPUPowerManagement.kext but XCPM – power management via mach_kernel – and thus you are doing it wrong. Make sure to use a SMBIOS setup that supports FrequencyVectors in IOPlatformPluginFamily.kext/*/Resources/Mac-[board-id].plist and accept that you will see less P-States with MSRDumper.kext. Using the Intel Power Gadget or my AppleIntelCPUPowerManagementInfo.kext will however show every single frequency change. Link to comment Share on other sites More sharing options...
Simonej Posted May 21, 2014 Share Posted May 21, 2014 Sorry Slice, I had Haswell i5 with Clover, Gen P,C states + Drop OEM and it work well (better than SSDTGen i think), but now I have X79 16020v2 i need to use SSDT, with Clover i have states error in my Console. I'm happy to help if Clover has problem with my setup. Thank you. Link to comment Share on other sites More sharing options...
K3Rl Posted May 22, 2014 Author Share Posted May 22, 2014 Haswell processors do not use AppleIntelCPUPowerManagement.kext but XCPM – power management via mach_kernel – and thus you are doing it wrong. Make sure to use a SMBIOS setup that supports FrequencyVectors in IOPlatformPluginFamily.kext/*/Resources/Mac-[board-id].plist and accept that you will see less P-States with MSRDumper.kext. Using the Intel Power Gadget or my AppleIntelCPUPowerManagementInfo.kext will however show every single frequency change. Thanks for the answer, Pike! So the speed changes that I had been seeing were all correct - I removed the SSDT and everything works the same - which indeed proves this to be correct. Thanks again and keep up the great work! Link to comment Share on other sites More sharing options...
Recommended Posts