Ryzen 7000 Testing

[PoMpIs]

Few things spring to mind having reviewed the EFI and config.plist.

1. Order of Lilu.kext and VirtualSMC.kext is interrupted by NVMeFix.kext and RestrictEvents.kext, i.e. these two Lilu plugin kexts are loaded after Lilu, but before VirtualSMC.kext.
I always thought VirtualSMC.kext had to be the first plugin following Lilu. Is that not the case?

2. Have you tried setting SecureBootModel to Disabled instead of Default.
I know you are using the MacPro7,1 SMBIOS and it is normally set with SecureBootModel > Default. But as this is an unsupported setup, running Secure Boot mode set as 'medium' might cause a KP, when expected settings are not present.

3. Adding RadeonSensor.kext and SMCRadeonGPU.kext makes little sense to me as these kexts aren't usually essential to having a working RX6000 series GPU.
As both kexts contain 'executable' code within them, wouldn't it be better to wait till the system is up and running before adding these two non-essential kexts.

Otherwise it looks good.

They are almost all disabled

 

[ExtremeXT]

I have tried several times... reset the bios, with Resibar enable and disable... nothing works

the first time I booted with AmdTscSync it showed this:

View attachment 7437


on the following attempts it seems different:

View attachment 7438

This issue is happening because the CPU cores cannot communicate properly, because of this XNU panics instead of continuing. I can try making a patch to disable the panic, but I'm not sure if it'll cause problems later on... I think our best bet is to wait for AMD to fix the issues (maybe we can submit a bug report? You should try running Linux, see if it works fine...). You can try booting with cpus=1 which should make it boot with only one core/thread. At least we'll know if we encounter any other problems if we fix this one...

Don't bother with changing BIOS, leave the following:
Above4G and Resizable BAR Support: Enabled
SVM: Enabled

Plist:
no npci in boot-args
ResizeAppleGpuBars -> 0

I will do more research about it tomorrow, but I don't really think we can fix it ourselves.

Few things spring to mind having reviewed the EFI and config.plist.

1. Order of Lilu.kext and VirtualSMC.kext is interrupted by NVMeFix.kext and RestrictEvents.kext, i.e. these two Lilu plugin kexts are loaded after Lilu, but before VirtualSMC.kext.
I always thought VirtualSMC.kext had to be the first plugin following Lilu. Is that not the case?

2. Have you tried setting SecureBootModel to Disabled instead of Default.
I know you are using the MacPro7,1 SMBIOS and it is normally set with SecureBootModel > Default. But as this is an unsupported setup, running Secure Boot mode set as 'medium' might cause a KP, when expected settings are not present.

3. Adding RadeonSensor.kext and SMCRadeonGPU.kext makes little sense to me as these kexts aren't usually essential to having a working RX6000 series GPU.
As both kexts contain 'executable' code within them, wouldn't it be better to wait till the system is up and running before adding these two non-essential kexts.

Otherwise it looks good.

VirtualSMC doesn't have to be loaded before Kexts that don't depend on it, like NVMeFix. I don't think SecureBootModel or GPU Kexts would cause this issue, but I guess they could try disabling them...

- ExtremeXT

 

[Shaneee]

Out of curiosity can you try booting with a single RAM stick. If that doesn't change disable XMP or EXPO.

 

[PoMpIs]

Out of curiosity can you try booting with a single RAM stick. If that doesn't change disable XMP or EXPO.

I have already tried that many times, having the memory profile active or turned off does nothing

cpus=1 doesn't work either

nothing works

 

[PoMpIs]

@Shaneee

I have a computer running 24/7 with unraid as nas, and I have it configured in netboot... I have activated LAN boot in ryzen and I have loaded UBUNTU

Everything works fine and you are right cpu family is 25

Can you help us with something Linux?

Because here the Ryzen works everything, network, audio, RX6800XT...

Spoiler

 

[Middleman]

Okay @PoMpIs,

I have done some changes to @ExtremeXT's EFI file which should help such as enabled the (expected) PTXH USB renames, added -no_compat_check (to get past the EB|LOG:EXITBS:START), rearranged the kexts (Lilu/WEG, AMDRyzenCPUPowerManagement was not loading first before SMCAMDProcessor), SetupVirtualMap was enabled, deleted OCValidate.exe etc.

Try it now.

 

[PoMpIs]

Okay @PoMpIs,

I have done some changes to @ExtremeXT's EFI file which should help such as enabled the (expected) PTXH USB renames, added -no_compat_check (to get past the EB|LOG:EXITBS:START), rearranged the kexts (Lilu/WEG, AMDRyzenCPUPowerManagement was not loading first before SMCAMDProcessor), SetupVirtualMap was enabled, deleted OCValidate.exe etc.

Try it now.

Come on, let's try it right now..

in 5 minutes I give you the result

 

[PoMpIs]

@Middleman

Well no luck...

It fails to load the system:

Ventura:




Monterey:

 

[Middleman]

@Middleman

Well no luck...

It fails to load the system:

Ventura:

View attachment 7443


Monterey:

View attachment 7444

Okay then, try enabling SetupVirtualMap but using my EFI.

 

[Middleman]

Also set SecureBootModel to Disabled.

 

[PoMpIs]

@Middleman

Well, he didn't like changing SecureBootModel

 

[Middleman]

@PoMpIs,

One thing I do want to mention to you. How I managed to get the Atari VCS Picasso APU to boot was that I used the following device properties plus enabled Whatevergreen. I may have enabled npci=0x2000 but that was about it - and it booted straight into Monterey.



So looking at your situ, AMD IGPU PCIRoot address we're using is our main concern, because it is currently the same (thus clashing) with your RX6800 on Slot-2. Because it is not an Intel IGPU, the OS is looking for the correct platform-id to match as it boots up. Because the Picasso was added sometime ago into the macOS AMD GPU stack (as Apple was likely experimenting with alternative AMD GPU based builds for laptop), that's why it was able to boot.

Unfortunately it seems @ExtremeXT's disable-gpu setting under the IGPU doesn't seem to work.

Anyways I have just created a new config.plist from @ExtremeXT's latest config.plist and added the AMD Raphael IGPU details into it using PCIRoot PciRoot(0x0)/Pci(0x8,0x1)/Pci(0x0,0x0) and also changed the SMBIOS to iMacPro1,1. See if it boots.

 

[PoMpIs]

@PoMpIs,

One thing I do want to mention to you. How I managed to get the Atari VCS Picasso APU to boot was that I used the following device properties plus enabled Whatevergreen. I may have enabled npci=0x2000 but that was about it - and it booted straight into Monterey.

View attachment 7445

So looking at your situ, AMD IGPU PCIRoot address we're using is our main concern, because it is currently the same (thus clashing) with your RX6800 on Slot-2. Because it is not an Intel IGPU, the OS is looking for the correct platform-id to match as it boots up. Because the Picasso was added sometime ago into the macOS AMD GPU stack (as Apple was likely experimenting with alternative AMD GPU based builds for laptop), that's why it was able to boot.

Unfortunately it seems @ExtremeXT's disable-gpu setting under the IGPU doesn't seem to work.

Anyways I have just created a new config.plist from @ExtremeXT's latest config.plist and added the AMD Raphael IGPU details into it using PCIRoot PciRoot(0x0)/Pci(0x8,0x1)/Pci(0x0,0x0) and also changed the SMBIOS to iMacPro1,1. See if it boots.

Tested...

There is no way to boot macOS on this Ryzen



Now I'm going to sleep for a few hours, then we'll continue fighting with this ryzen and MacOS

Thank you very much everyone for the help, it is a pleasure to do tests like this

 

[ExtremeXT]

@PoMpIs Can you try this DEBUG EFI? Boot macOS and when it gets stuck again take a screenshot, boot back to another OS and zip the log file + the SysReport file at the root of the EFI partition.

- ExtremeXT

 

[Shaneee]

Definitely seems to be TSC related I think. Give this a go.

 

[ExtremeXT]

Definitely seems to be TSC related I think. Give this a go.

Hmm, what did you change? I see no changes related to TSC.

 

[Shaneee]

Hmm, what did you change? I see no changes related to TSC.

Kernel patches. Added a few older ones back. Same method that got the Lenovo builds to initially boot without cpus=1

 

[ExtremeXT]

Give this a try, it's the old EFI + @Shaneee's TSC kernel patches + some new fixes I added, hopefully this gets it booting...

- ExtremeXT

 

[PoMpIs]

Hello!!!

I'm ready again to keep trying

I'm going to try those EFI and I'll tell you

 

[craighazan]

Hello!!!

I'm ready again to keep trying

I'm going to try those EFI and I'll tell you

Don't give up, progress is being made!.