Kernel panic after upgrading to dual E5-2667 v2

[xuegy]

Hello all,

My CPU was dual E5-2670 v1, running 10.15.4. I upgraded to 2667 v2. Everything is OK in Linux(CPUs are good). But I got kernel panics booting into macOS. The error message is like

panic(cpu 0 caller 0x......): "x86_validate topology() 32 threads but 33 registered from MADT~0/..."

I found that enabling "patchAPIC" option in clover can temporarily fix this problem. I can boot into the system. But as long as I run AppleIntelInfo.kext, the system hangs up immediately.
Then I checked my DSDT. I have 32 cpus in each socket(SCK0~SCK3), which means 128 CPUs in total??? I thought I only have 32 when I was using 2670v1.
Any idea how to fix this issue?
Thanks in advance.

 

[huananzhiiiiii]

i have a similar issue...
anyone got an idea?

 

[RomaKoks]

So, I have a similar issues too. We all have same configuration. Several things that I've figured out:
1) With only one cpu it is stable.
2) Installation of OS is better to do with one CPU, it would be more stable, but still not completely.
3) WIndows works normally
4) Works normally with 2x E5 2697 v2
So, may be for this CPU Windows is the only stable solution

 

[xuegy]

i have a similar issue...
anyone got an idea?

So, I have a similar issues too. We all have same configuration. Several things that I've figured out:
1) With only one cpu it is stable.
2) Installation of OS is better to do with one CPU, it would be more stable, but still not completely.
3) WIndows works normally
4) Works normally with 2x E5 2697 v2
So, may be for this CPU Windows is the only stable solution

I've resolved this issue a long time ago. https://github.com/dortania/bugtracker/issues/48#issuecomment-774943662
You need to manually patch your APIC table because dual CPU messed it up.

 

[theroadw]

The problem with this one is tricky, It may be the same that I had on my HP Z820 2x E5 2697 V2.
Your DSDT or CPU SSDT is enumerating your CPU's incorrectly, so you need to manually enumerate them in the proper order.

The correct SSDT-1-CpuDef should be:

Processor (C000, 0x00, 0x00000410, 0x06)

Processor (C001, 0x01, 0x00000410, 0x06)

Processor (C002, 0x02, 0x00000410, 0x06)

Processor (C003, 0x03, 0x00000410, 0x06)

Processor (C004, 0x04, 0x00000410, 0x06)

Processor (C005, 0x05, 0x00000410, 0x06)

Processor (C006, 0x06, 0x00000410, 0x06)

Processor (C007, 0x07, 0x00000410, 0x06)

Processor (C008, 0x08, 0x00000410, 0x06)

Processor (C009, 0x09, 0x00000410, 0x06)

Processor (C00A, 0x0A, 0x00000410, 0x06)

Processor (C00B, 0x0B, 0x00000410, 0x06)

The second block , "0x00" is the order number, in hex. In decimal they are 1,2,3,4,5, and so on

But, the native SSDT-1-CpuDef in my Z820 has an inconsistent enumeration of cpu-devices:

Processor (C000, 0x80, 0x00000410, 0x06)

Processor (C001, 0x81, 0x00000410, 0x06)

Processor (C002, 0x00, 0x00000410, 0x06)

Processor (C003, 0x01, 0x00000410, 0x06)

Processor (C004, 0x02, 0x00000410, 0x06)

Processor (C005, 0x03, 0x00000410, 0x06)

Processor (C006, 0x04, 0x00000410, 0x06)

Processor (C007, 0x05, 0x00000410, 0x06)

Processor (C008, 0x06, 0x00000410, 0x06)

Processor (C009, 0x07, 0x00000410, 0x06)

Processor (C00A, 0x08, 0x00000410, 0x06)

Processor (C00B, 0x09, 0x00000410, 0x06)

So the first core of the first processor gets assigned to 0x80, instead of 0x00. Windows and Linux don't mind, but MacOS doesn't like that at all.

 

[RomaKoks]

I've resolved this issue a long time ago. https://github.com/dortania/bugtracker/issues/48#issuecomment-774943662
You need to manually patch your APIC table because dual CPU messed it up.

My issue is similar, but not the same - it is more linux related, but thank you, I will look at patching of APIC table. Btw in my case, the system messes not only CPUs number but also USB ports (but without panic).
Forgot to mention before: memtest86+ always freezes on Test 1 (Address test, own address, 1 CPU). But, all RAM modules passes it with single CPU (with any one of E5 2667 v2 istalled)