MP 1,1-5,1 OpenCore and the 2008 Mac Pro 3,1

[scorpio333]

I found a 7950 3GB on a shelf here. I think I pulled it because it doesn't have a boot screen when the seller said it would, as well as, non working ports. I recall messaging the seller and not getting any answers. Anyhow, put the card in and boot to El Cap. Do the whole dosdude mojave patcher, boot to it and started the Mojave install. Same result. So I've tried four different cards (two R9 280x, a 5570 and a 7950) and get the same result. The only way I've got Mojave booted is using a HDD I created in my 5,1 and booting into Boot EFI with the 5770, it's very slow and sluggish.

I initiated a return for the R9 280x and the seller came back with a bunch of questions and claims he tested it on his 3,1. He also said there's a bug in the mojave patcher that was never resolved, I'm not aware of that.

I think something isn't right with this machine or something is going over my head. I'm the original owner, the machine was paid for by a previous employer, it's given me years of good service. Maybe it's time to just let this go and keep an eye out for a 4,1 or 5,1 especially with the outrageous GPU prices. Installing OC and BS on my 5,1 took a day or two and has been mostly smooth sailing since. This 3,1 is getting to the point where it just doesn't feel worth it.

 

[scorpio333]

-------I AM AN IDIOT-------​

The problem was between the screen and keyboard. After a few messages back n forth from the seller I got it working.

Starting installing Mojave as usual using dosdude's patcher. When it rebooted I jumped in with holding down option. Select BOOT EFI. Once back at the Mojave installer, I ran the post install and rebooted. I was pleased to see the Language selection screen. Had three patcher updates, installed those and now I've got a working Mojave. I'm going to wrap the 3,1 in caution tape and take a looooong nap.

This is right up there with the time I drained the tranny on my wife's Subaru instead of the oil.

 

[StormLord]

Anyone knows if I can install Big Sur on an NVME and enable 5GT/s speeds?
On Mojave I use pciutils to enable 5GT/s speed after boot with a launch deamon

 

[StormLord]

Jacklukem’s Big Sur basesystemfix usb installer fixed audio on my 3,1 using Big Sur. Wifi and USB 2 issues also. On his github

do you use it as patch to an existing Big Sur oclp installation? Or you install the whole patched big sure with it?

 

[freqrider]

do you use it as patch to an existing Big Sur oclp installation? Or you install the whole patched big sure with it?

I patched an existing BS installed via Starplayrx’s BigMac.

 

[freqrider]

Anyone knows if I can install Big Sur on an NVME and enable 5GT/s speeds?
On Mojave I use pciutils to enable 5GT/s speed after boot with a launch deamon

I had to load DirecHW.kext via openCore to get pcitools to work. DirectHw.kext wouldn’t load in Big Sur for me. Just add it to EFI/OC/KEXTS AND add it to the config.plist. Assuming you are using openCore.

 

[joevt]

I had to load DirecHW.kext via openCore to get pcitools to work. DirectHw.kext wouldn’t load in Big Sur for me. Just add it to EFI/OC/KEXTS AND add it to the config.plist. Assuming you are using openCore.

DirectHW.kext is not needed for Intel Macs running macOS later than El Capitan.
Just add debug=0x144 to boot-args
https://gist.github.com/joevt/e3cd4ff08aae06279134969c98ca3ab7

 

[freqrider]

DirectHW.kext is not needed for Intel Macs running macOS later than El Capitan.
Just add debug=0x144 to boot-args
https://gist.github.com/joevt/e3cd4ff08aae06279134969c98ca3ab7

Really? Are you simply just obtaining pci tree info or can you actually switch/force the pci slot to operate at pci2 (5GT/s) speeds instead of pci 1 (2.5GT/s) without DirectHW.kext?

 

[joevt]

Really? Are you simply just obtaining pci tree info or can you actually switch/force the pci slot to operate at pci2 (5GT/s) speeds instead of pci 1 (2.5GT/s) without DirectHW.kext?

All the pciutils commands work including setpci which is used to write to the PCIe registers to switch PCIe link rate.

pciutils has a command to list all the access methods:

% lspci -A help
Known PCI access methods:

intel-conf1
intel-conf2
dump
darwin

In my version of pciutils, I added an option to test all the access methods:

% lspci -A detect
Trying method darwin......cannot open AppleACPIPlatformExpert (add boot arg debug=0x144 & run as root)...No.
Trying method intel-conf1......no I/O permission...No.
Trying method intel-conf2......no I/O permission...No.
lspci: Cannot find any working access method.

% nvram boot-args
boot-args    -v keepsyms=1 debug=0x144 -no_compat_check maxmem=63488 nv_spanmodepolicy=1 nvda_drv=1

% sudo lspci -A detect
Trying method darwin......using AppleACPIPlatformExpert...OK
Trying method intel-conf1...DirectHW.kext not loaded.
...no I/O permission...No.
Trying method intel-conf2......no I/O permission...No.

% sudo kextload ~/Library/Application\ Support/ROMTool/DirectHW.kext

% sudo lspci -A detect
Trying method darwin......using AppleACPIPlatformExpert...OK
Trying method intel-conf1......sanity check...outside the Asylum at 0/00/0...OK
Trying method intel-conf2......No.

The darwin method (uses AppleACPIPlatformExpert.kext for macOS versions greater than El Capitan) is better than the intel-conf1 (uses DirectHW.kext) method because it can access PCIe registers at config offsets greater than 0x100.
For example, my MacPro3,1 has 81 PCI or PCIe devices, 75 have extended PCIe registers which cannot be accessed by the intel-conf1 method.

% sudo setpci -A darwin -s '*:*.*' 100.W | wc -l
      81
% sudo setpci -A intel-conf1 -s '*:*.*' 100.W | wc -l
      81
% sudo setpci -A darwin -s '*:*.*' 100.W | sed '/ffff/d' | wc -l
      75
% sudo setpci -A intel-conf1 -s '*:*.*' 100.W | sed '/ffff/d' | wc -l
       0

DirectHW.kext should be modified to be able to provide the same info that AppleACPIPlatformExpert.kext can for macOS El Capitan and earlier. It should also be modified to work for M1 Macs which don't use AppleACPIPlatformExpert.kext. With the modified kext, we can add a darwin2 access method.

 

[scorpio333]

Ran into another snag. Found my 3,1 sitting at "waiting for remote debugger connection". When I last left it, it was running. I tried multiple reboots and each time booting it stops at that line.

Perhaps related, I just left the dosdude USB flash drive connected. Figured I can't lose it if it's plugged in. Every time the computer woke from sleep I would have a blah blah drive wasn't ejected properly. USB functionality on wake up is a hassle. I have to unplug and plug in my mouse and keyboard to get them back. A USB DAC never has issues, I can tell because it pops through the amp/speakers.

 

[christiann]

I got OCLP working great on my 3,1 with my RX 560. Full graphics support, V 0.0.20. Big Sur 11.2.3

 

[Jonathan50]

Oclp 0.0.22 installed 20D80 and "ota" updated to 20D91 (2.44GB update) with onboard audio working. Big thanks to everyone involved.

 

[Dayo]

onboard audio working.

Saw that they finally cracked this.
One of their devs shared this with me just before the OC 0.6.8 release. They were hard at work on it:

Will roll the fix into MyBootMgr's ConfigFactory app

With respect to recent Mac OS, the MP31 falls under what we can call "Generally Supported" as compared to the MP51 which is "Essentially Supported".

"Generally Supported": most things work but needs "Firm Patching" (filesystem writes) for some things.
"Essentially Supported": basically everything works after "Soft Patching" (on the fly changes in the memory space).

Looking back on this, it doesn't actually seem there is anything that can be done OpenCore wise with a cMP5,1 that is not available on an MP31. Especially with legacy audio now fixed without needing "Rigid Patching". So we can say the MP31 is also "Essentially Supported" at this point as well I think.

EDIT: APFS Support can be activated in OpenCore (and RefindPlus) for MP31 but currently, you do need to flash your firmware for NVMe support. So, "Generally Supported" on that count afterall. Only changes to "Essentially Supported" once the firmware is flashed for NVMe support.

 

[amstel78]

Saw that they finally cracked this.
One of their devs shared this with me just before the OC 0.6.8 release. They were hard at work on it:
View attachment 1755890
Will roll the fix into MyBootMgr's ConfigFactory app


Looking back on this, it doesn't actually seem there is anything that can be done OpenCore wise with a cMP5,1 that is not available on a cMP3,1. Especially with legacy audio now fixed without needing "Rigid Patching". So we can say cMP3,1 is also "Essentially Supported" at this point as well I think.

EDIT: APFS Support can be activated in OpenCore (and RefindPlus) for cMP3,1 but currently, you do need to flash your firmware for NVMe support. So, "Generally Supported" on that count afterall. Only changes to "Essentially Supported" once the firmware is flashed for NVMe support.

This is very good news, Dayo! I look forward to these fixes rolled into the next iteration of ConfigFactory. Would you know if this is to be implemented in OC v0.6.9?

 

[Dayo]

Would you know if this is to be implemented in OC v0.6.9?

it is a kext (AppleALC.kext) and not something that goes into OpenCore proper.

 

[Henninges]

using the AppleALC 1.5.9 combined with OC 0.6.9, latest lilu and weg but still got no sound.
calling alcid=x but neither one of the 1,12 or 15 layouts function.

 

[Henninges]

as i extract the AppleALC 1.6.0 from an Legacy Project EFI folder, i have now nativ audio on my 3,1...

 

[trifero]

Is there any benefits to run OC instead of a patched system in Mac pro 3,1?

 

[joevt]

Is there any benefits to run OC instead of a patched system in Mac pro 3,1?

Easier macOS updates (security updates, etc.). You can do patched system, then use OC when you want to check for updates.

 

[FarmerBob]

I was all over this for the ease of updates as mentioned. Then last night I wiped a partition on my MBP 7,1 and reformatted it as APFS. I loaded the DD1 version of Mojavè 10.14.6 (18G103) when I was done with all the patches, the Apple Updater came up with 3 updates: Supplemental Update (Safari 14.0.x), Safari 14.0.3 and SecUpdt 2021-002 (18G8022) in one step. In the background it updated Gatekeeper (twice) & MTR.

WOW! I have been struggling with all this on an HFS+ volume. Redid the one partition to APFS and "Magic". The 3,1 is next. Just need to make sure that I can migrate Office 2011 to the new partition on a new drive.

 

[joevt]

WOW! I have been struggling with all this on an HFS+ volume. Redid the one partition to APFS and "Magic". The 3,1 is next. Just need to make sure that I can migrate Office 2011 to the new partition on a new drive.

Office 2011 let you move your profile folder anywhere using an alias to the folder.
Current Office doesn't let you do that.

 

[Jonathan50]

OCLP 0.1.1 installed 20D80, updated OTA to 20E232 (5.71GB).
When using either of the old GPUs (8800GT, GT120) onboard audio is present and working.
When using metal supporting GPU (GT710) onboard audio doesn't show.
Kextstat shows AppleALC 1.6.0 loaded. Any ideas?

 

[joevt]

OCLP 0.1.1 installed 20D80, updated OTA to 20E232 (5.71GB).
When using either of the old GPUs (8800GT, GT120) onboard audio is present and working.
When using metal supporting GPU (GT710) onboard audio doesn't show.
Kextstat shows AppleALC 1.6.0 loaded. Any ideas?

Do the old GPUs have a HDA device like the new GPUs?

 

[Jonathan50]

Do the old GPUs have a HDA device like the new GPUs

An 8800GT has two DVI ports and a GT120 has a DVI port and a mini displayport. Do any of those ports output audio? Doubtful.

 

[Syncretic]

Do the old GPUs have a HDA device like the new GPUs?

If I recall correctly, NVIDIA didn't introduce onboard audio hardware until the GeForce 200/GT2xx series.

An 8800GT has two DVI ports and a GT120 has a DVI port and a mini displayport. Do any of those ports output audio? Doubtful.

While neither the 8800GT nor the GT120 has onboard audio hardware, both support audio over DVI via a S/PDIF connection (there's a header to connect S/PDIF from another source (e.g. sound card or motherboard), which then gets sent to the DVI port; a DVI->HDMI cable then allows a TV/monitor to get the full A/V stream). Newer cards with onboard audio hardware support this directly (no S/PDIF connection required); at one point, I was using this DVI->HDMI setup (with audio) using a GT710.