How to extend the iMac Pro to X99 [Successful Build/Extended Guide]

[nmano]

What exactly do you want to fix? You mean that your CPU is not properly recognized?

Did you use the correct Haswell-E "FakeCPUID" entry and the correct Haswell-E "xcpm_bootstrap" KernelToPatch entry?

Following PMHeart and Sherlocks, you do not need the "xcpm_cpuid_set_info" KernelToPatch entry for Haswell-E..

The supposedly correct values you can find in my 10.13 Desktop Guide.

I guess it is also clear that you still have to create the SSDT.aml for your 5960X with Piker-Alpha's ssdtPRGen.sh and subsequently inject an appropriate frequency vector to your Mac-F60DEB81FF30ACF6.plist. I hope you also use SMBIOS Mac Pro 6.1 etc... You are anyway very familiar with the entire XCPM procedure... Anyway everything is also chronologically and stepwise described in much detail throughout my 10.13 Desktop guide.

Would you do me the favour to verify my guide with your system? I do not posses any Haswell-E CPU to do it myself.. I would appreciate it much.

Cheers man!

KGP

I don't use any fakecpuID
XCPM work well.
only issue about this Mac.
look my config cpu type not work maybe clover issue.

 

[nmano]

Can you check this is right cpu id set info for Haswell-E

xcpm_cpuid_set_info © Pike R. Alpha

Find 72 3CD07750 0FB6C0

Replace

6F 3CD07750 0FB6C0

 

[nmano]

I thing your xcpm_cpuid_set_info © Pike R. Alpha
wrong value.
check kernel in OS X 17A315i

 

[kgp]

I don't use any fakecpuID
XCPM work well.
only issue about this Mac.
look my config cpu type not work maybe clover issue.

O.K. I see what you mean. I also encountered the "3 GHz Unknown" report in "About this Mac" at the very beginning. But as I say, after adding the correct FakeCPUID and after properly configuring XCPM, my CPU was reported more or less correctly as a "3 GHz 10-Core Intel Xeon E5". Try to use Haswell-E FakeCPUID for test purposes and lets see...

By the way did you realize that Geekbench in your case detects the wrong CPU Architecture: Intel (32-bit) ?

 

[kgp]

I thing your xcpm_cpuid_set_info © Pike R. Alpha
wrong value.
check kernel in OS X 17A315i

Well I use:

Find: 75000472 3cd07750 0fb6c0
Replace: 7500046a 3cd07750 0fb6c0

It is the entry provided by PMHeart for PM2/DP3:

Find: 72 3c d0 77 50 0f b6 c0
Replace: 6a 3c d0 77 50 0f b6 c0

I double checked the entry with Sherlocks also for PM2/DP3. Why do you think it is wrong in PM3/DP4? Because of the 75 00 04 at the beginning?

XCPM is working perfectly (see also the AppleInfoKext dump in the XCPM section of my 10.13 desktop guide)...

Pike states:
_cpuid_set_info:
ffffff800039bd8c add al, 0x72

and Stinga states:
_cpuid_set_info, replace 72
Find: 04 72 3C D0
Replace: 04 XX 3C D0

My plugin type in IOReg for CPU00@0 is "0X1"
and the cpu-type is "<02 0a>"

I never understand how to dissamble the kernel and check for the right KernelToPatch entries. Can you please explain it step by step? I would appreciate it much!

One more question: To properly boot the PB3/DP4 installer I needed to add the following KernelToPatch entry:

Find: 07000000 31d2e88f fcffff
Replace: 07000000 31d29090 909090

Do you know the function of this patch and what would be the proper comment? macOS 10.13, 10.12, 10.11 or only 10.13?

Cheers,

KGP

 

[kgp]

Can you check this is right cpu id set info for Haswell-E

xcpm_cpuid_set_info © Pike R. Alpha

Find 72 3CD07750 0FB6C0

Replace 6F 3CD07750 0FB6C0

So one does need the "cpu id set info" KernelToPatch Entry also for Haswell-E?

Can any other Haswell-E user confirm that

Find: 72 3CD07750 0FB6C0
Replace: 6F 3CD07750 0FB6C0

is the correct "cpu id set info" KernelToPatch Entry for Haswell-E?

What about Ivy-Brige CPUs? Do you also have a correct "cpu id set info" KernelToPatch Entry for Ivy-Bridge?

Thanks,

KGP

 

[Subsystems]

Although your post and questions seem to be addressed to @MacGrummel let me just provide a view comments:

1.) Do you use the 10.12.6 Web Driver workaround?
2.) If so, did you apply the AGDPFix or config.plist patch?
3.) Do you get a signal (login screen) with the boot flag nv_disable=1?
4.) Do you have a single or multi-monitor configuration?

Cheers,

KGP

The questions were addressed to you as well mate or to anyone that could give more information about this matter. Thanks for the reply though.

I've installed DP4 and I'm using only one monitor.
With the boot flag nv_disable=1 the boot to the OS runs just fine.
My attempts to get my graphics card (Pascal) pass the black screen without that boot flag are being somewhat annoying.
I've tried all your workarounds and so far with no luck. I'm going to try once more, hopefully it gets the graphics card recognized.

I'll get back to you after this. Thanks once again mate!

 

[kgp]

The questions were addressed to you as well mate or to anyone that could give more information about this matter. Thanks for the reply though.

I've installed DP4 and I'm using only one monitor.
With the boot flag nv_disable=1 the boot to the OS runs just fine.
My attempts to get my graphics card (Pascal) pass the black screen without that boot flag are being somewhat annoying.
I've tried all your workarounds and so far with no luck. I'm going to try once more, hopefully it gets the graphics card recognized.

I'll get back to you after this. Thanks once again mate!

The AGDPFix attached should work... If not forget about the 10.12.6 driver workaround and just work with nv_disable=1. The difference is anyway minimal, as there is neither OpenGL nor Metal support with Pascal and Maxwell CPUs with the current workaround.

Cheers,

KGP

 

[Subsystems]

The AGDPFix attached should work... If not forget about the 10.12.6 driver workaround and just work with nv_disable=1. The difference is anyway minimal, as there is neither OpenGL nor Metal support with Pascal and Maxwell CPUs with the current workaround.

Cheers,

KGP

Will do. Thank you!

 

[Subsystems]

The AGDPFix attached should work... If not forget about the 10.12.6 driver workaround and just work with nv_disable=1. The difference is anyway minimal, as there is neither OpenGL nor Metal support with Pascal and Maxwell CPUs with the current workaround.

Cheers,

KGP

Well, basically it's still not recognizing my GPU. I'm giving up for the time being and waiting for further progress on NVIDIA web drivers. Thanks for all your help!

Subsystems