virtualbox.org

According to at least one AMD forum, other users are reporting similar
problems with VMWare as well as VirtualBox.

Workarounds so far have included switching off I/O APIC and/or AMD-V.

Opinion on the AMD forum seems to be there is a problem with AMD-V, which
would be consistent with users here switching off AMD-V and the guest working.

If there was a problem with AMD-V, a lot more than Windows XP would fail to work. The CPU the guest OS sees when AMD-V is on vs. off is fairly significantly different.

What I want to know is whether anyone successfully booted XP on Ryzen without any virtualization involved. Does that actually work? If it does, then we can start talking about AMD-V problems.

I was trying to point out to users here that the reported problems
were not necessarily unique to VirtualBox, as (very) similar symptoms
had been reported with XP Guests when using VMWare.

I apologise if that was not clear.

I think it was clear -- I'm just more interested in finding out what the problem really is rather than whether other products are affected by it

There is a possible workaround: VBoxManage modifyvm <vm> --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd

This masks off the VME bit in CPUID. It is currently unclear why that should make a difference, but it reportedly does.

Hello Guys!!
Greg from France
Asus Crosshair Hero6 bios 1107+R7 1700X+2x8GoCorsair+samsung960pro+nvidia660(don't laugh for graphic card please)
For 3 weeks, I try too desesperately to run winxp 32bits (to be able to access to my old 16bit saab car technical review) on my win10 64bit host with Ryzen.
SVM is enable in the bios, I have activated and deactivated hyper-v to unlock harware virtualization in virtualbox.
Try tick, untick all what I was able to->black screen after first reboot or bsod with "vga" after rereboot
Try to install linux mint 64bit guest, and winxp guest->i can finish installation but Windows don't start
Try to install linux mint 32bit guest, and winxp guest->i can finish installation but Windows don't start

So can somebody confirm (not "I heard about..." but some facts) and explain (what to tick exactly) if it's possible run a virtualized Windows XP with AMD Ryzen

Best regards

You tried the CPUID workaround I mentioned in the post right before yours, right? Did it make any difference? No one wants to say if it makes a difference for them or not, and I don't have a Ryzen machine.

Command line as admin, result:
C:\Program Files\Oracle\VirtualBox>VBoxManage.exe modifyvm <vm> --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd
Le fichier spécifié est introuvable.

C:\Program Files\Oracle\VirtualBox>VBoxManage modifyvm <vm> --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd
Le fichier spécifié est introuvable.

Try again...

Gregouze wrote:Command line as admin, result:

Why? Did anyone tell you to run as admin? VirtualBox runs on the user context and you will have to run the command line as the owner of the VM.

Also, you need to replace "<vm>" with the actual name of the VM. In quotes. Example:

VBoxManage modifyvm "winxp 32bits" --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd

Thanks to Michaln & Socratis!!!



I have just apply this command line (with the good syntax) and Windows XP host and my saab 16bit technical review are working for me now.
I let automatic settings for Windows 32 bit host in virtualbox installed on my win10 pro with ryzen, you have saved my life of little emulator.



Thanks again

I'm pretty sure that michaln is going to be much more pleased than I am
Glad you got it going.

Sorry for my poor english,

read : I let automatic settings for Windows 32 bit GUEST in virtualbox installed on my win10 pro HOST with ryzen, and apply the command line

Thanks for confirming that the workaround works. It looks like AMD broke something, we have a good idea what exactly but need to investigate a little more.

michaln wrote:There is a possible workaround: VBoxManage modifyvm <vm> --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd

This masks off the VME bit in CPUID. It is currently unclear why that should make a difference, but it reportedly does.

Hi michaln, thanks for the workaround. It worked in my case.

For anyone interested, I have also found this on the Internet. It has a detailed analysis and explanation of this issue.

Now I wonder, is this fixable via updates in the future? It's not an issue with the hardware implementation, right?

zypA13510 wrote:For anyone interested, I have also found this on the Internet. It has a detailed analysis and explanation of this issue.

Nice finding! Nice reading!

zypA13510 wrote:Now I wonder, is this fixable via updates in the future? It's not an issue with the hardware implementation, right?

To me, from reading the article, it would seem that it is indeed a problem in the hardware implementation, but I guess it's something that can be worked around in software. From the article:

As incredible as it is, Ryzen has buggy VME implementation ... How did AMD miss it?