Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Status
The first post of this thread is a WikiPost and can be edited by anyone with the appropiate permissions. Your edits will be public.

h9826790

macrumors P6
Apr 3, 2014
16,656
8,587
Hong Kong
I wish there was a
temperature controller application for Linux based on sensors temperature like MFC. This way I could have completely switched to KVM solution with open core inside passing the GPU trough. You can have Linux on the main machine and OSX and windows in the KVM machines using AMD and Nvidia at the same time.

If you don’t mind to let the fans just run faster. We can set the minimum fans speed higher in MacOS. Then that will stay until you set another values again, or perform a SMC reset.

Of course, not as good as variable speed base on some sensors. However, we can at least set it at a safe RPM for everything.
 

startergo

macrumors 603
Sep 20, 2018
5,020
2,282
If you don’t mind to let the fans just run faster. We can set the minimum fans speed higher in MacOS. Then that will stay until you set another values again, or perform a SMC reset.

Of course, not as good as variable speed base on some sensors. However, we can at least set it at a safe RPM for everything.
What is the holdup in Linux? Is SMC not controllable?
Code:
sudo apt-get install lm-sensors fancontrol
I think I used one application forgot which maybe fancontrol, and because it was not optimized for cMP it just increased the speed of the fans. So I am pretty sure SMC is controllable we just need to configure the sensor control.
 
Last edited:

h9826790

macrumors P6
Apr 3, 2014
16,656
8,587
Hong Kong
What is the holdup in Linux? Is SMC not controllable?
Code:
sudo apt-get install lm-sensors fancontrol
I think I used one application forgot which maybe fancontrol, and because it was not optimized for cMP it just increased the speed of the fans. So I am pretty sure SMC is controllable we just need to configure the sensor control.

IMO, the best way is by using SMCfanControl in macOS.

Just spin up the fan, that will alter the associated minimum fan speed accordingly. Which keep the SMC has ultimate control of the target fan speed, which is important when something approaching overheat.

I am quite sure all associated SMC key still working in Linux, however, just no one write any software to use read / write them.

A simple test would be stressing the GPU, we should able to hear the PCIe and PSU fan spin up in Linux (no need to wait for GPU temperature to build up. SMC does that base on PCIe card's power draw. It work best if the card is connected to the mini 6pins. But also work with Pixlas modded cMP, just not that "sensitive"). Since the GPU has it's own fan onboard to cool down the GPU. So, almost no risk even the PCIe fan doesn't spin as expected.

But if the PCIe fan spin up accordingly, then we can assume the SMC is working properly, and it will protect the hardware from overheating.
 

startergo

macrumors 603
Sep 20, 2018
5,020
2,282
"This is an enhanced version of Allan McRae mbpfan


mbpfan is a daemon that uses input from coretemp module and sets the fan speed using the applesmc module. This enhanced version assumes any number of processors and fans (max. 10).


  • It only uses the temperatures from the processors as input.
  • It requires coretemp and applesmc kernel modules to be loaded.
  • It requires root use
  • It daemonizes or stays in foreground
  • Verbose mode for both syslog and stdout
  • Users can configure it using the file /etc/mbpfan.conf"
  • Tested Mac Pro Models
    • Mac Pro 3,1
    • Mac Pro 4,1 (with 5,1 firmware and hex-core Xeon X5690 mod - Linux 4.14)
[automerge]1578149972[/automerge]
I am quite sure all associated SMC key still working in Linux, however, just no one write any software to use read / write them
It looks like there is a software, but only uses input from coretemp module and sets the fan speed using the applesmc module.
 

h9826790

macrumors P6
Apr 3, 2014
16,656
8,587
Hong Kong
"This is an enhanced version of Allan McRae mbpfan


mbpfan is a daemon that uses input from coretemp module and sets the fan speed using the applesmc module. This enhanced version assumes any number of processors and fans (max. 10).


  • It only uses the temperatures from the processors as input.
  • It requires coretemp and applesmc kernel modules to be loaded.
  • It requires root use
  • It daemonizes or stays in foreground
  • Verbose mode for both syslog and stdout
  • Users can configure it using the file /etc/mbpfan.conf"
  • Tested Mac Pro Models
    • Mac Pro 3,1
    • Mac Pro 4,1 (with 5,1 firmware and hex-core Xeon X5690 mod - Linux 4.14)
[automerge]1578149972[/automerge]

It looks like there is a software, but only uses input from coretemp module and sets the fan speed using the applesmc module.

Thanks, I don't know that. It seems applesmc is all we need to control the fans in Linux.

Of course, if we want a more automatic way, we need another software to control the fans via applesmc.
 

startergo

macrumors 603
Sep 20, 2018
5,020
2,282
Of course, if we want a more automatic way, we need another software to control the fans via applesmc.
I believe this software can be enhanced if apart from coretemp other temperatures are integrated in the source code. I will request linux app from MFC I wiil be glad to pay for that software.
 

startergo

macrumors 603
Sep 20, 2018
5,020
2,282
As far as the CPU power management... we are on our own. That is why going towards the KVM route looks more reasonable for me. I think we need to focus on SMC control from Linux.
Already created feature request from MFC developers.
 

startergo

macrumors 603
Sep 20, 2018
5,020
2,282
Guys maybe we need to create a post and vote?
"I'm sorry it's not planned due to a very low % of possible linux users. I'm not closing it though."
[automerge]1578177214[/automerge]
Maybe funding a project on kickstarter or similar?
 

papadj3

macrumors regular
Nov 23, 2018
113
11
To Expert w1z:
papadj3 a Real Newbie Question:

I have an internal windows disk on an SSD in an actual sled (#2) and a 7101A in my Pcie slot 2 with my Catalina OS- Do I have to worry about your "Windows Scenario regarding the 7101A as described in your post #6?? I am having trouble getting back from windows to mac Catalina? I have to use the option button to restart in Catalina? It will not reboot to mac from the bootcamp control panel? Should I have my windows bootcamp in a particular sled? Thought you might be able to help, referred by Expert cdf-
 
Last edited:

papadj3

macrumors regular
Nov 23, 2018
113
11
Expert @cdf-
I believe I installed OC correctly, 4,1 to 5,1 cMP! (Catalina installed on a 7101A Pcie)
It is running well, I Think!
One Issue, I do not have a software update install button??

Screen Shot 2020-01-07 at 1.46.54 PM.png
Is that correct if OC boots the catalina os drive?
 

papadj3

macrumors regular
Nov 23, 2018
113
11
OH, So when there is a new update the "install/upgrade now button" will appear?!
I have all the advanced options checked-Is that OK?
Screen Shot 2020-01-07 at 2.39.15 PM.png
I Can't Thank You All Enough!!
I have my cMP running Catalina (without patches) with the normal way to Update-UNBELEIVABLE!!
 
  • Like
Reactions: cdf

papadj3

macrumors regular
Nov 23, 2018
113
11
startergo:
How can you test/be sure the VMM flag is on? During the OC procedure, I remember a check for OC working but not for the VMM flag specifically-?
 

papadj3

macrumors regular
Nov 23, 2018
113
11
Will a clone (using CCC) of Catalina that has OC/VMM flag installed work as a backup start-up disk?
If original start-up disk fails will the clone work the same? (does the clone have to be modified?)
 

MacUser2525

Suspended
Mar 17, 2007
2,097
377
Canada
Will a clone (using CCC) of Catalina that has OC/VMM flag installed work as a backup start-up disk?
If original start-up disk fails will the clone work the same? (does the clone have to be modified?)

You would need to ensure the open core files are on the new drive to load it exactly the same.
 

cdf

macrumors 68020
Original poster
Jul 27, 2012
2,256
2,583
Will a clone (using CCC) of Catalina that has OC/VMM flag installed work as a backup start-up disk?

If you installed OC by following the guide, then from your original OC-booted Catalina system, you can simply select the clone in Startup Disk and reboot. In this case, you are still booting through OC, which is installed on the EFI partition of the original disk. For a successful boot, you may have to carry out step 5d in the guide.

If original start-up disk fails will the clone work the same? (does the clone have to be modified?)

I don't think that CCC clones the EFI partition. In any case, the clone can still boot without OC. For a successful boot, you need to make sure that you have properly set -no_compat_check (see step 8 in the guide).
 

Ludacrisvp

macrumors 6502a
May 14, 2008
797
363
Would you mind linking to the version you are using and attaching your config file?
I'm using Debug 0.5.2 ... when I tried 0.5.3 it breaks the ability to boot, not sure what the cause is just yet but here is a working config.plist for a 3,1 with 0.5.2 debug with debug log writing to EFI partition.
This is also using a mix of VMWare and iMacPro info to get HEVC encoding and potentially to help movement to catalina?
 

Attachments

  • config.plist-052-debug-MP31-HEVC-ENC-VMWARE.zip
    3.2 KB · Views: 97

Ludacrisvp

macrumors 6502a
May 14, 2008
797
363
Tried to boot the Catalina installer that has been laying around for a while on my 3,1 dual X5482, which was made from an old dosdude patcher ... anyways booted it via OpenCore which panics on that AppleIntelMCEReporter kext. What is odd to me is that if I don't use OpenCore to get there it does boot to the installer, mind you the install fails to start with some generic error but it gets to a usable recovery / installer environment with no panic and no removal of that kext.

So I'm not 100% sure on the whole loss of dual socket message that was mentioned. And the issue really comes from OpenCore spoofing setup, either because OC only sees one socket or because we are using an iMacPro board ID. Perhaps it is simply this kext isn't being loaded when not using certain board-IDs and that is why no panic without opencore but panic with it.

Code:
Anonymous UUID:       427601D5-A2A6-71E2-7A30-A36B5994CE92

Thu Jan  9 23:59:00 2020

*** Panic Report ***
mp_kdp_enter() can't get x86_topo_lock! Debugging anyway! #YOLO
panic(cpu 0 caller 0xffffff8007c6ab4a): Kernel trap at 0xffffff8007b520e1, type 14=page fault, registers:
CR0: 0x000000008001003b, CR2: 0x0000000000000000, CR3: 0x000000000aaf9000, CR4: 0x00000000000026e0
RAX: 0x0000000000000000, RBX: 0x0000000000000000, RCX: 0x0000000000000000, RDX: 0x000000005f210634
RSP: 0xffffff83bc483ae0, RBP: 0xffffff83bc483b00, RSI: 0xffffff80432a8000, RDI: 0x0000000000000000
R8:  0x0000000000000000, R9:  0x0000000000000000, R10: 0xffffff833cd6a200, R11: 0x0000000000000000
R12: 0xffffff8042db6900, R13: 0xffffff83bc483c88, R14: 0x0000000000000000, R15: 0x0000000000000000
RFL: 0x0000000000010006, RIP: 0xffffff8007b520e1, CS:  0x0000000000000008, SS:  0x0000000000000010
Fault CR2: 0x0000000000000000, Error code: 0x0000000000000000, Fault CPU: 0x0, PL: 3, VF: 5

Backtrace (CPU 0), Frame : Return Address
0xffffff83bc483540 : 0xffffff8007b419fb
0xffffff83bc483590 : 0xffffff8007c78d35
0xffffff83bc4835d0 : 0xffffff8007c6a79e
0xffffff83bc483620 : 0xffffff8007ae8a40
0xffffff83bc483640 : 0xffffff8007b410e7
0xffffff83bc483740 : 0xffffff8007b414cb
0xffffff83bc483790 : 0xffffff80082d2a49
0xffffff83bc483800 : 0xffffff8007c6ab4a
0xffffff83bc483980 : 0xffffff8007c6a848
0xffffff83bc4839d0 : 0xffffff8007ae8a40
0xffffff83bc4839f0 : 0xffffff8007b520e1
0xffffff83bc483b00 : 0xffffff8007c5fc35
0xffffff83bc483b40 : 0xffffff7f8a2cea4c
0xffffff83bc483b60 : 0xffffff7f8a2ce9c1
0xffffff83bc483ba0 : 0xffffff8007c74bc5
0xffffff83bc483c70 : 0xffffff7f8a2ce9a0
0xffffff83bc483c90 : 0xffffff7f8a2ce87b
0xffffff83bc483cc0 : 0xffffff7f8a2ce786
0xffffff83bc483d60 : 0xffffff7f8a2cf83a
0xffffff83bc483de0 : 0xffffff800821d986
0xffffff83bc483e40 : 0xffffff800821d531
0xffffff83bc483ef0 : 0xffffff800821c87e
0xffffff83bc483f50 : 0xffffff800821f01d
0xffffff83bc483fa0 : 0xffffff8007ae813e
      Kernel Extensions in backtrace:
         com.apple.driver.AppleIntelMCEReporter(115.0)[6772488F-9069-3B6F-80AE-A5E1184C1AD2]@0xffffff7f8a2c9000->0xffffff7f8a2dbfff
            dependency: com.apple.iokit.IOACPIFamily(1.4)[69BFC9C0-F2F3-3C97-AEF4-41DDD9834BCB]@0xffffff7f886e1000
            dependency: com.apple.iokit.IOPCIFamily(2.9)[F2770E7E-657B-3500-8427-77DF6A190D63]@0xffffff7f886f3000

BSD process name corresponding to current thread: kernel_task
Boot args: -v debug=0x144 -wegtree -no_compat_check

Mac OS version:
19B88

Kernel version:
Darwin Kernel Version 19.0.0: Tue Sep 17 01:59:51 PDT 2019; root:xnu-6153.11.24~17/RELEASE_X86_64
Kernel UUID: 40B0A1FB-7EFD-373C-AB47-D7E8D18DA3B2
Kernel slide:     0x0000000007800000
Kernel text base: 0xffffff8007a00000
__HIB  text base: 0xffffff8007900000
System model name: VMware7,1 (Mac-7BA5B2D9E42DDD94)
System shutdown begun: NO
Panic diags file unavailable, panic occurred prior to initialization

System uptime in nanoseconds: 37913492224

EOF
 

Ludacrisvp

macrumors 6502a
May 14, 2008
797
363
Also, need to do some more testing but I was able to boot via OC with edited DSDT/SSDT with EFI flashed GTX 650 into mojave with the RX 580 taking over during boot and disabling the GTX 650 providing a fully usable RX 580.

Compared to what normally happens in this configuration of having a disabled RX 580 and a useless GTX 650 (slow UI, or unable to load UI) Now I have also removed all the NVDA*kext and GeForce* files from /S/L/E, but even with them removed, without the OC DSDT/SSDT override added it just ended up with display only on GTX 650 and not able to use it (white / grey screen).

No more time to test now, tried taking some videos but those are hard to tell whats going on.

I'd also like to try with OC not attempting GOP because currently what happens is you see the Mac EFI (alt/option) on GTX 650 and you can pick what you want to boot, i select the EFI Boot that contains OC and you see the OC loading lines and then when the boot picker from OC should show the display turns off and loads the display for RX 580 which remains black but active until a minute or two passes (opencore boot log is about 110 seconds, sometimes longer like 160 seconds not sure why the duration changes so much) at which point it starts the verbose boot of 10.14.6 which loads only the RX 580 to the desktop.

So it would seem the DSDT/SSDT patch is implemented before you tell the OS to boot, which is my theory as to why the screen on the GTX 650 goes to sleep when the boot picker should load and the RX 580 wakes up.

Added some lspci output showing both cards in the system at the same time, perhaps this will be of use... GTX 650 is 'off' and the RX 580 is 'on'.

The goal here is to allow someone to keep a cheap mac flashed unsupported GPU in slot 3 (or presumably slot 4) and be able to have a modern GPU in slot 1/2. This give you the ability to potentially have the best of both worlds.

The idea stemmed from using DSDT/SSDT mods to disable Nvidia discrete GPU in hack laptops.
 

Attachments

  • nvidia_and_amd.txt
    8.5 KB · Views: 111
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.