Contribute
Register

How should you load GPUSensors?

Status
Not open for further replies.
Joined
Dec 23, 2012
Messages
157
Motherboard
Gigabyte Z170 HD3P
CPU
i7-6700K
Graphics
Pascal Titan X
Mac
  1. iMac
  2. MacBook Air
Classic Mac
  1. iMac
Mobile Phone
  1. iOS
I just installed a passively cooled GT 730 (4GB version by Zotac), and I'd like to read its temperature, specially because it's passively cooled. It seems to behave correctly (after more than 30 minutes of endless stressing OpenGL work, touching the passive heat sink by hand feels like a MacBook Air at full load, so I guess it's in safe temp, but anyway I'd like to be able to measure it).

I'm at the latest 10.11.6 version (I mean, including the latest security update). I'm using NVIDIA web drivers because the GT 730, despite being a Kepler (GK208), does require the web drivers, at least on 10.11.6.

HWMonitor was installed by MultiBeast when I installed El Capitan last August. The numbers it had provided these months are fairly complete (temp for each core, SSD temps, etc...), except there were no GPU numbers. I didn't pay much attention to that because I was using the HD530 and I thought it would mean that the HD530 has no temp sensor.

But now, with the GT 730 I see there're no GPU temp numbers either. The first I did was to update to the latest HWMonitor (6.23.1417). But it didn't fix it. The GPU temp doesn't appear either.

Trying to find the cause, I found that GPUSensors isn't being loaded, but the rest of sensors do:

If I do sudo kextstat | grep hwsensors , I get this:
22 0 0xffffff7f80c4f000 0x5000 0x5000 org.hwsensors.driver.CPUSensors (1707) 92A9DE1E-049C-3CC9-9BF3-6F064AECABC8 <14 7 5 4 3>
31 0 0xffffff7f80c56000 0x8000 0x8000 org.hwsensors.driver.ACPISensors (1707) D7761E01-6AA4-337E-8D93-2B3E142914E0 <14 11 7 5 4 3>
44 0 0xffffff7f80c26000 0xe000 0xe000 org.hwsensors.driver.LPCSensors (1707) DD7AEF15-8549-39F8-A052-DF4867AABECF <14 12 11 7 5 4 3>

If I look for error messages at /var/log/system.log, I don't find any error saying something like GPUSensors wasn't able to be loaded:
sudo cat /var/log/system.log | grep ensors

Jan 3 12:14:00 localhost kernel[0]: CPUSensors: CPU family 0x6, model 0x5e, stepping 0x3, cores 4, threads 8, TJmax 100
Jan 3 12:14:00 localhost kernel[0]: CPUSensors: setting platform keys to [j43 ]
Jan 3 12:14:00 localhost kernel[0]: CPUSensors: base CPU multiplier is 40
Jan 3 12:14:00 localhost kernel[0]: CPUSensors: started
Jan 3 12:14:00 localhost kernel[0]: ACPISensors (TZ00): 1 sensor added
Jan 3 12:14:00 localhost kernel[0]: ACPISensors (TZ00): started
Jan 3 12:14:00 localhost kernel[0]: ACPISensors (TZ01): 1 sensor added
Jan 3 12:14:00 localhost kernel[0]: ACPISensors (TZ01): started
Jan 3 12:14:00 localhost kernel[0]: IT87xxSensors: started
Jan 3 12:14:00 localhost kernel[0]: PTIDSensors: started

So, the question is: Why are CPUSensors, ACPISensors, IT87xxSensors, and PTIDSensors being loaded, but GPUSensors won't?

At /Library/Extensions/ , I have:
FakeSMC.kext
FakeSMC_ACPISensors.kext
FakeSMC_CPUSensors.kext
FakeSMC_GPUSensors.kext
FakeSMC_LPCSensors.kext

But, however, there's no plugins folder inside FakeSMC.kext:
bash-3.2$ ls /Library/Extensions/FakeSMC.kext/
Contents

bash-3.2$ ls /Library/Extensions/FakeSMC.kext/Contents/
Info.plist MacOS

bash-3.2$ ls /Library/Extensions/FakeSMC.kext/Contents/MacOS
FakeSMC

So, why are CPUSensors, ACPISensors, and LPCSensors being loaded if there's no plugins folder? And why GPUSensors won't?

Then I remembered I boot with Clover, which has its home in the EFI partition. In fact, some searches seem to indicate that FakeSMC sensors kexts should be loaded from the EFI partition. I still don't have the knowledge for knowing if this is true or not, but /Volumes/EFI/EFI/CLOVER/kexts/10.11/ is empty (zero files), so I conclude that CPUSensors, ACPISensors, and LPCSensors are being loaded from /Library/Extensions/ in some mysterious way, as there's no plugins folder.

So, how could I proceed next? Any hints at getting GPUSensors loaded?

Thanks!
 
Last edited:
I have a similar issue with Sierra. Updated my vid card last week to a 280x and it wont show up on the HWMonitor list of reported values. Tried /EFI/CLOVER/KEXTS/OTHER folder and also with the .pkg install for HWMonitor that put a 721kb FakeSMC.kext into S/L/E.

Neither works, but the Heaven and Valley Benchmarks report the GPU temp just fine, as does Windows 10.

Odd...

Edit: Kept the most recent Kexts but replaced the hwsensors app with an older version 6.15.1370. Got my GPU temp back. I also found that if I load the app manually rather than load at startup, I get a better result.


Ersterhernd
 
Last edited:
I have the .kexts in EFI/CLOVER/KEXTS/OTHER and HWMonitor in Applications folder. Did not have to do anything else.

Make sure "Inject Kexts" parameters are set to YES in your config.plist.

Works fine without doing anything else.
 
Install and using GPUSensors is usually pretty straighforward, maybe GK208 just isn't supported, so the kext doesn't load?
But shouldn't I get some error message if that's the case? Is there any log file that could catch errors when loading GPUSensors, apart from /var/log/system.log ?

BTW, should I rename FakeSMC_GPUSensors.kext to GPUSensors.kext? (the other kexts load even if they have the FakeSMC_ prefix)
 
Some news: Loading GPUSensors with kextload worked. So I followed izo's advice, and copied all the FakeSMC kexts to EFI/CLOVER/KEXTS/OTHER. Now sudo kextstat | grep hwsensors reports CPUSensors, ACPISensors, LPCSensors, and GPUSensors, but HWMonitor still doesn't show GPU temp, no matter if I use the latest version, or 6.15.1370 as suggested by
ersterhernd. To make things more mysterious, the GPUSensors kext unloads itself after a few minutes, so sudo kextstat | grep hwsensors no longer reports it after a few minutes...
 
Does it report any diagnostic messages in Console.app now?
Nope. I also tried to get more verbose hints using kextutil instead of kextload, but no errors regarding unsupported GPU chip, card, nothing. It's strange, but GPUSensors doesn't whine about not linking this GPU, yet the kext is automatically unloaded after a few minutes, without any error message at /var/log/system.log
 
Anyone ever get anywhere with this? I have the same problem with a GT 710 under Sierra. Using the FakeSMC_*.kexts from RehabMan
 
Status
Not open for further replies.
Back
Top