virtualbox.org

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

Did you compare the name of the blog author with the user here providing the workaround?

On https://www.virtualbox.org/wiki/Testbuilds there are new test builds which automatically enable a workaround (clear the CPUID VME bit) on AMD Ryzen CPUs. With those, no manual tweaking should be necessary.

I have a Ryzen 5 1500 x and i faced the same problem. I am using VirtualBox and i have virtualized 2 machines 32-bits windows XP and i always have black screen after boot. I have solved changing the config of virtual machine(Configuration/System/Aceleration and Toggle off in VT-X+x/AMD-V) and now the virtual machines works.

Hi,

is "Ryzen" the same reason why i can install only 32bit guests? Drives me crazy because i shutdown my intel-host and rebuild it with Win 10 64 Pro and a Ryzen - and now i can only install/use 32bit guests. I virtualize my old Intel-Win10-64 Systems and for testing purpose it works on my old rig.

Have you tried using one of the test builds listed above? Have you enabled hardware virtualization (AMD-V) in the host machine's BIOS?

BillG wrote:Have you tried using one of the test builds listed above? Have you enabled hardware virtualization (AMD-V) in the host machine's BIOS?

I'm using a Ryzen 7 with SVM and IOMMU enabled in my BIOS. I have installed the 5.1.23 r115417 test build in Windows 10 x64. I'm not having the black screen issue; I use all 64-bit guests. I am however having periodic freezes in my Windows 8.1 x64 guest OS. I have done a little testing in Xubuntu 14.04 64-bit and didn't notice any of these issues. However my Windows guest seems to have the hardest time. Initially I was trying to run it over a USB 3.0 hard drive and the freezes were unbearable, then I moved it to my NVMe drive and performance marginally improved. After upgrading to the 5.1.23 test build, I've had much better performance with my virtual disk on the NVMe drive, but after about 10 minutes or so the freezes began to be too much to handle again. I still have my old FX-8320 PC handy so I'm willing to try anything on the Ryzen machine.

Hula wrote:is "Ryzen" the same reason why i can install only 32bit guests?

No.

dbjungle wrote:I have installed the 5.1.23 r115417 test build in Windows 10 x64. I'm not having the black screen issue; I use all 64-bit guests. I am however having periodic freezes in my Windows 8.1 x64 guest OS.

Completely unrelated to this discussion.

Im Sorry, failure sitting in front of the monitor. My first time with a uefi bios and i did not find the AMD-V option. My fault. Got it late in the night. Sorry again

Hula wrote:My first time with a uefi bios and i did not find the AMD-V option.

I'd blame the vendor for disabling AMD-V by default

I am surprised that Oracle has no AMD contacts to report bugs like this. There is "bugref:8852" in the commit that has the workaround.

@yohong

• Your first (and last so far) post was on 2011-09-12. Wow, that's what I call a regular
• How do you know they don't already have a contact? Just curious... It's not that if they report an issue with hardware, AMD is going to issue a patch.

I am asking especially because a hack for this has already been added into VirtualBox.

"Hacks" like this have to be implemented all the time in software to bypass hardware limitations/bugs. You can't get around a hardware bug by issuing a hardware patch. Your only option is to implement a software "hack".

socratis wrote:"Hacks" like this have to be implemented all the time in software to bypass hardware limitations/bugs. You can't get around a hardware bug by issuing a hardware patch. Your only option is to implement a software "hack".

Microcode update.