856 replies to this topic

[Mohamed Khairy]

now with latest revision  multi per and frequency is multiplied  by 3
value x8  readed as x24 also frequency from 800 to 2400
!!

[Greggen]

Been testing some more. This FakeSMC and Slice's FakeSMC, with no plugins, gives the exact same values and behavior (clocks down to non-turbo modes when using more then one core). The both give all the states as well. When adding IntelThermal.kext to this FakeSMC nothing changes. But when I add IntelCPUMonitor.kext to Slice's it removes 3 p-states (31, 32 and 33) but other then that it behaves correctly. Using the turbo-values as it should and so on. Dunno if there is a problem with my DSDT (more or less totally unedited), SSDT or something else. But as it works with one and not the other it seems like it should be fix-able without the need of DSDT and/or SSDT edits.

I would rather use this branch of FakeSMC as it seems more active and SuperIO gives me more values then the other branch. Of course I'm willing to help with any testing for this!

[Mohamed Khairy]

with latest rev

 multiper.png   31.94K   11 downloads

[eep357]

Greggen, on 02 May 2012 - 07:30 PM, said:

Perfect! Now it updates =)

But the speed stepping still doesn't seem 100% correct. I have all the values (16, 31, 32, 33, 42, 43, 44, 45) but when the CPU uses all cores it goes down to 31 instead of 42. Im fully aware that the multipliers over 42 might not work in OSX (but I want them on other OSes and they do show with MSRDumper) but with the other FakeSMC (Slice's branch) it still behaves correctly. When using just one core with your FakeSMC it behaves correctly (sits at 45/44) but from there it goes to 33 with 2 cores, 32 with 3 and 31 with all 4. State 42 and 43 are there though, they pop up occasionally, but not when they should.

Going to try out what happens if I set the highest value to 42 in BIOS. I'll be back with that info.

EDIT: Ok, it' still the same so it had nothing to do with the values being over 42. Now they are set to 42, 41, 40 and 39. With one core it goes to 41/42 and then its the same behavior as above. Steps down to 33, 32 and stops at 31.

I'm no expert as I have different hardware(guessing u have sandy-bridge from post but hardware not in sig ) but aren't the highest turbo/multi values only applicable while system uses 1 core and when using all cores overall clocks cannot go to max speed? My impression was this was how turbo boost worked, super speed to one core only to help with quick single thread processes as multicore would use too much power/heat at sustained turbo clocked rates?

[scorcho]

Mohamed Khairy, on 04 May 2012 - 06:16 AM, said:

with latest rev

multiper.png

I am seeing the same on my Sandy Bridge system. Motherboard is a Gigabyte P67A UD4-B3, chip is a 2500K. Multiplier is stuck at 0.0x and the CPU package fluctates betweeen 409.60GHz and 1.10THz. These readings changed after the latest Sandy Bridge multiplier test code.

[scorcho]

Kozlek - thank you for your quick work. I noticed you rolled back the changes to the pre May 4 code and swapped the IntelThermal.kext. My multiplier and CPU Package reading again match what I have in Windows. I only have two steppings (16x and 43x), translating to 1.6ghz and 4.3ghz respectively. I am using the latest Chameleon to 'add' P/C states, and have not attempted to manually add them myself.

If you require any log dumps on my end, please let me know within the next hour or two while I am in front of my system. I'm more than happy to help.

 Screen Shot 2012-05-04 at 9.18.36 AM.png   143.31K   105 downloads

Here is how your handiwork currently looks on my system. I only have one system fan, which isn't attached to the motherboard, so no readings. Everything else appears to match what I see in Win7, including the SMART temperature data off an external drive connected to a Marvel SATAIII controller. Mighty impressive.

[Mohamed Khairy]

thanks now fixed  but with out good sync as msrdumper

[kozlek]

scorcho, on 04 May 2012 - 01:12 PM, said:

Kozlek - thank you for your quick work. I noticed you rolled back the changes to the pre May 4 code and swapped the IntelThermal.kext. My multiplier and CPU Package reading again match what I have in Windows. I only have two steppings (16x and 43x), translating to 1.6ghz and 4.3ghz respectively. I am using the latest Chameleon to 'add' P/C states, and have not attempted to manually add them myself.

If you require any log dumps on my end, please let me know within the next hour or two while I am in front of my system. I'm more than happy to help.

Screen Shot 2012-05-04 at 9.18.36 AM.png

Here is how your handiwork currently looks on my system. I only have one system fan, which isn't attached to the motherboard, so no readings. Everything else appears to match what I see in Win7, including the SMART temperature data off an external drive connected to a Marvel SATAIII controller. Mighty impressive.

Looks like only GPU0 tachometer shows as GPU 5... That's interesting. If you still in front of your system please upload SMC keys dump. You could download atMonitor and go to  Utilities->System Info->Smc sensors and use copy&paste

Update: don't bother I just fixed it

[eberts]

Hi kozlek,

first of all - thank you, I really appreciate your great work!
However, I have a little problem: My Apple GT120 shows only 60 rpm and under load it decreases rpms (although fan spin goes up). I know you wrote about it somewhere already but is it possible to get it fixed?

Thx a lot
eberts

[Greggen]

eep357, on 04 May 2012 - 06:53 AM, said:

I'm no expert as I have different hardware(guessing u have sandy-bridge from post but hardware not in sig ) but aren't the highest turbo/multi values only applicable while system uses 1 core and when using all cores overall clocks cannot go to max speed? My impression was this was how turbo boost worked, super speed to one core only to help with quick single thread processes as multicore would use too much power/heat at sustained turbo clocked rates?

Yea, I have a 2500k. And you are correct about the turbo-stuff. But it still doesn't work as it should. With one core only the "super speed" works but after that it clocks down to 33 which is the highest non-turbo setting. And for 3 cores its 32 and 4, 31. What it should do is clock down from whatever setting you have for one core to the next step (in my case 45, 44, 43, 42), not jump from turbo to non turbo just because you use 2 cores or more.

Anyhow, it seems like I kinda sorted it out. I made a correct SSDT and the turbo-steps work. And it acts the same for whatever branch I use. BUT I have totally lost all p-states between 16 and the lowest turbo-setting. What I have now is 16, 42, 43, 44, 45 and its working as it should. I guess that the different FakeSMCs handle the original, incorrect, SSDTs differently or something like that.

What I am supposed to do now, to get a few states between 16 and 42 I don't have a single clue about, but I guess the problem doesn't have anything to do with FakeSMC... or does it? XD

EDIT: Oh, btw, how come that MSRDumper is capable of showing 42+ while HWMonitor isn't? Is the monitor somehow getting the value from AICPM (which isn't working with values over 42 if I'm not mistaken?) while MSRDumper gets the values from hardware?

[TH3L4UGH1NGM4N]

Not sure why but RadeonMonitor is not loading during boot.  Running on a 6870 now previously was using a 4850 where the temp used to show up in both istats and hwsensor but doesn't show up anymore not sure why.

Here's a pic of the console where it shos the radeonmonitor not being loaded.



And here's what hwmonitor currently shows




Another thing I noticed is that when I install the intelthermal kext my gpu no longer displays video so I have to vnc in and delete the kext.  Running a Q8200 temps shows up in iStats but not in hwmonitor.

[Greggen]

Ok, looks like I sorted it out. I was using MacMini5,1 in my SMBios (this was from Chameleon Wizard). THIS IS INCORRECT! It seems like it's very important to type Mini with lower-case M

Macmini5,1

Gave me a bunch of new states and it's behaving identical in every FakeSMC. I guess they handle incorrect values differently, but with a corrected SSDT and correct model-id its all good.

[kozlek]

eberts, on 05 May 2012 - 07:57 AM, said:

Hi kozlek,

first of all - thank you, I really appreciate your great work!
However, I have a little problem: My Apple GT120 shows only 60 rpm and under load it decreases rpms (although fan spin goes up). I know you wrote about it somewhere already but is it possible to get it fixed?

Thx a lot
eberts

I need more information. Please show me your kernel log.

TH3L4UGH1NGM4N, on 05 May 2012 - 09:03 PM, said:

Not sure why but RadeonMonitor is not loading during boot.  Running on a 6870 now previously was using a 4850 where the temp used to show up in both istats and hwsensor but doesn't show up anymore not sure why.

Here's a pic of the console where it shos the radeonmonitor not being loaded.



And here's what hwmonitor currently shows




Another thing I noticed is that when I install the intelthermal kext my gpu no longer displays video so I have to vnc in and delete the kext.  Running a Q8200 temps shows up in iStats but not in hwmonitor.

I think you did installed something wrong. Try fresh installation: delete all previous fakesmc.kext and plugins, download latest revision from GitHub and install fakesmc and plugins with installer.

Greggen, on 06 May 2012 - 10:28 AM, said:

Ok, looks like I sorted it out. I was using MacMini5,1 in my SMBios (this was from Chameleon Wizard). THIS IS INCORRECT! It seems like it's very important to type Mini with lower-case M

Macmini5,1

Gave me a bunch of new states and it's behaving identical in every FakeSMC. I guess they handle incorrect values differently, but with a corrected SSDT and correct model-id its all good.

Mac model is important. I use custom mac model injected with chameleon and information provided with CustomProfile.kext. Download links in my sig.

[eberts]

Hi kozlek,

here are the relevant lines from kernel log:


May  5 19:53:29 localhost kernel[0]: SuperIODevice: found Winbond W83627DHGP on port=0x2e address=0x290
May  5 19:53:29 localhost kernel[0]: FakeSMC: Opensource SMC device emulator. Copyright 2009 netkas. All rights reserved.
May  5 19:53:29 localhost kernel[0]: FakeSMC: Hardware monitoring plugins. Copyright 2012 kozlek, usr-sse2, slice, navi. All rights reserved.
May  5 19:53:29 localhost kernel[0]: FakeSMCDevice: 100 preconfigured key(s) added
May  5 19:53:29 localhost kernel[0]: SMC: successfully initialized
May  5 19:53:29 localhost kernel[0]: IntelThermal: CPU family 0x6, model 0x17, stepping 0x6, cores 2, threads 2, TJmax 100
May  5 19:53:29 localhost kernel[0]: NVClockX: G96 [GeForce 9500 GT] device-id=0x640 arch=0x400000
May  5 19:53:29 localhost kernel[0]: W836xMonitor: started
May  5 19:53:29 localhost kernel[0]: com.apple.AppleFSCompressionTypeZlib kmod start
May  5 19:53:29 localhost kernel[0]: com.apple.AppleFSCompressionTypeZlib load succeeded
May  5 19:53:29 localhost kernel[0]: NVClockX: registering temperature sensors
May  5 19:53:29 localhost kernel[0]: NVClockX: registering tachometer sensors
May  5 19:53:29 localhost kernel[0]: NVClockX: registering frequency sensors

Just loading SuperIOSensors.kext, NVClockX.kext and IntelThermal.kext since GeForceX.kext crashes (don't have GraphicsEnabler=yes - so probably no readable gpu rom)

Thank you
eberts

[kozlek]

eberts, on 07 May 2012 - 08:12 AM, said:

Hi kozlek,

here are the relevant lines from kernel log:


May  5 19:53:29 localhost kernel[0]: SuperIODevice: found Winbond W83627DHGP on port=0x2e address=0x290
May  5 19:53:29 localhost kernel[0]: FakeSMC: Opensource SMC device emulator. Copyright 2009 netkas. All rights reserved.
May  5 19:53:29 localhost kernel[0]: FakeSMC: Hardware monitoring plugins. Copyright 2012 kozlek, usr-sse2, slice, navi. All rights reserved.
May  5 19:53:29 localhost kernel[0]: FakeSMCDevice: 100 preconfigured key(s) added
May  5 19:53:29 localhost kernel[0]: SMC: successfully initialized
May  5 19:53:29 localhost kernel[0]: IntelThermal: CPU family 0x6, model 0x17, stepping 0x6, cores 2, threads 2, TJmax 100
May  5 19:53:29 localhost kernel[0]: NVClockX: G96 [GeForce 9500 GT] device-id=0x640 arch=0x400000
May  5 19:53:29 localhost kernel[0]: W836xMonitor: started
May  5 19:53:29 localhost kernel[0]: com.apple.AppleFSCompressionTypeZlib kmod start
May  5 19:53:29 localhost kernel[0]: com.apple.AppleFSCompressionTypeZlib load succeeded
May  5 19:53:29 localhost kernel[0]: NVClockX: registering temperature sensors
May  5 19:53:29 localhost kernel[0]: NVClockX: registering tachometer sensors
May  5 19:53:29 localhost kernel[0]: NVClockX: registering frequency sensors

Just loading SuperIOSensors.kext, NVClockX.kext and IntelThermal.kext since GeForceX.kext crashes (don't have GraphicsEnabler=yes - so probably no readable gpu rom)

Thank you
eberts

Sorry I don't support NVclock kext anymore. If you could make a photo of panic screen while booting osx with -v bootarg it would  help to solve the problem with geforcex driver

[mindlessmissy]

Does GeForceX stop updating GPU Shaders, Core and Memory values after they reach maximum values or is it just me ?

[eberts]

kozlek, on 07 May 2012 - 08:38 AM, said:

Sorry I don't support NVclock kext anymore. If you could make a photo of panic screen while booting osx with -v bootarg it would  help to solve the problem with geforcex driver

Sorry, just tried it again - GeForceX works perfectly. Apparently, I loaded it together with NVClockX before so it crashed.
HWMonitor now shows 2040 rpm and it's rising under load.
Thx again

[kozlek]

mindlessmissy, on 07 May 2012 - 01:34 PM, said:

Does GeForceX stop updating GPU Shaders, Core and Memory values after they reach maximum values or is it just me ?

That's wrong AGPM profile. Not all available standard mac models compatible with PC hardware. Sometimes you must use different CPU and GPU profiles. For example iMac11,2 compatible only with first generation Core CPUs and AMD GPUs. If you're using nVidia GPU you'll probably get wrong GPU power management.

[mindlessmissy]

kozlek, on 07 May 2012 - 04:31 PM, said:

That's wrong AGPM profile.

You are CORRECT.

I edited my AGPM kext, now the values are updating as per usual ..

Thanks for the hint !

[AppleGeek]

Thank you kozlek for hard work.You are great.

I want to ask you a little question.I have radeon HD 5770 and I use RadeonX plugin for showing GPU-Core tem but this plugin only show core tem.

I want also to show gpu core speed, memory speed and fan load.Is this possible?

Other kext's that I use :