Windows 10VM, consistent bluescreen in VMWare (Win11) no bluescreen in VMWare (Win10)

[itsdeadjim]

I have a Windows 10 Guest VM which is consistently crashing at launch of passmark as soon as I launch it and hit continue on one host, it's fine on another host.

VM Working fine, Windows 10, i3 8109U host.
https://imgur.com/A83Ks72

VM Crashing, Windows 11, i5, 12500T host.
https://imgur.com/oOVHPFr


I realise there's a very fair argument here, pointing at the problem being the host.

However others have suffered this issue previously, so if I could at least understand how it was solved for others and the cause, it may help me diagnose the issue on the Win11 VMHost, rather than the Win10 VM.

[David (PassMark)]

Can you send us or post the debug log files from the VM where it crashes.

See,
https://www.passmark.com/support/per.../debug-log.php

[itsdeadjim]

David,

Thank you so much, this is very appreciated at a busy / rough time for me.
It's PROBABLY not your software but it may well help me diagnose what is misconfigured or broken on the VM in some way

Any private info in these logs? or should I just post here?

[itsdeadjim]

Hi David,

One more unfortunate thing.
I've managed to make an exact copy of the VM on another machine - AMD based, Also Windows 11.
Same VMWare version and options for the VM / mitigations / emulation of counters etc and the VM works fine.

The VM and likely your software isn't the issue but if you have any tips on what might cause a DirectIO64.sys error, I would dearly appreciate it.
I will test on another, identical machine to the one I have faulting over coming days.



EDIT:

I can confirm the same directio64.sys crash on an identical HP 800 G9 12'th Gen Alder system which is quite a fresh Windows 11 build.

I can confirm the same directio64.sys crash on a totally different Dell Laptop, with 12th Gen Alder processor.....
(Dell 3571)

[David (PassMark)]

Shouldn't be anything private in the logs, but if you prefer Email that is OK.

Also make sure you have the latest version of our software as there was a recent BSOD issue with a particular Lenovo BIOS config.

With your Email can you also confirm which version of VM ware you are using and what the specs are for the VM and the host details.

[itsdeadjim]

If there's nothing in it no biggie, here we go.
https://www.dropbox.com/scl/fi/493kd...=czt9ihd7&dl=0
SysinfoLog.txt

https://www.dropbox.com/scl/fi/bk3ik...=op2d161u&dl=0
PerfTestLog.txt

Interesting note, the Dell machine is a 'business' grade machine with Windows 10 - the only thing here which is common between it and the 2 failing HPs is the 12'th Gen Alder processor.

Honestly if this is a passmark only bug, I'll be happy, it's purely terrifying to me, to think something else might trip up, this important VM for BSODs!
Thank you, amazing David

[itsdeadjim]

Failing system:
VM Workstation: 17.5.2 build-23775571
VM Config is the same obviously - same VM same VMDX file etc.
2 Processors, 2 cores
4.7GB memory
VMX file is in either VMWare 11 or "converted up" to VMWare 17, doesn't change it.

[keith]

Can you give the latest PerformanceTest build a try (build 1015):

https://www.passmark.com/products/pe...t/download.php

There were some updates related to specific CPU chipsets, including Alder Lake.

[keith]

Just to follow up, we looked into where the code appears to fail and found some bugs related to measuring the clock speed.

If Build 1015 still fails, please give this internal build a try:
https://www.passmark.com/temp/petst-...015-ADLVM1.exe

[David (PassMark)]

Log shows you are using build 1014, which is slightly old. Can you try build 1015
https://www.passmark.com/products/pe...st/history.php

[itsdeadjim]

Add to list, crashing: Surface Pro 9, 12th Gen processor, VMWare on Windows 11 recommends using the HyperV engine not the VMWare engine.

[David (PassMark)]

Pointless reporting additional systems if not using the latest software.

[itsdeadjim]

Pointless reporting additional systems if not using the latest software.

Apologies, I thought I had the latest, I'd not grabbed it long ago.
I managed to crash 1015, trying the test build now.

I want to thank you both a lot for this, I'm not having a good week for a variety of reasons and it's very rare to encounter decent software support from intelligent people who actually listen nowadays. So thank you.

[itsdeadjim]

Just to follow up, we looked into where the code appears to fail and found some bugs related to measuring the clock speed.

If Build 1015 still fails, please give this internal build a try:
https://www.passmark.com/temp/petst-...015-ADLVM1.exe

Ok so this is working!!
Again you 2 have been amazing, seriously, I know this seems silly but this means a lot to me right now.


So this is a bit frustrating, because whatever the issue is, it doesn't just kill your software, it bluescreen crashes a pretty important VM for me.
(I am not blaming you in the slightest!)


If your software can crash a VM under these conditions, what else can crash the VM?
The exact same Windows 10 VM which works flawlessly with Windows 10 host, VMware Workstation 15.5.1.



I have done some googling and VMWare 16 / 17 and Windows 11 have started using HyperV at the backend, rather than the old 'engine' they used to, something like that.
(You'll see a plethora of posts on this topic along the lines of):
virtualize intel vt-x/ept or amd-v/rvi
virtualize cpu performance counters
unrecoverable error: (vcpu-1)
etc A whole heap of issues moving from Windows 10 and or upgrading the VMWare workstation build.


I suspect this is where it's tripping up.

[David (PassMark)]

OK great that it is fixed.

There is a fair bit of code in PerformanceTest that collects system information. (Clock speeds, temperatures, disk interface types, ECC capability, RAM model numbers, etc...). A lot of this involves very low level code tailored to specific CPU & motherboard models. (CPU MSR registers, MMIO, PCIe bus config, SMBus, etc..). Very little of this is well documented by the CPU vendors and much of it changes with each new CPU family. A VM tries to emulate a lot of this low level hardware functionality and doesn't always get it correct, and we don't always get it correct either as there is often a fair amount of trial and error involved.