Windows 11 hangs with 100% CPU usage

[joao0]

I recently ordered an EPYC server where I've installed the latest proxmox version and added a Windows 11 VM. The VM would hang at boot after enabling Hyper-V but I finnally managed to fix it by upgrading kernel to 5.15. Now the VM will hang with 100% CPU usage and becomes unusable -- this normally happens at around the 10 minute mark after booting (I don't need even need to sign in).

CPU: AMD EPYC 7282 16-Core
Kernel Version: Linux 5.15.5-1-pve #1 SMP PVE 5.15.5-1 (Mon, 29 Nov 2021 18:49:57 +0100)
PVE Manager Version: pve-manager/7.1-8/5b267f33

qm.conf
acpi: 1
agent: 1
bios: ovmf
boot: order=ide0;scsi0
cores: 4
cpu: host
efidisk0: local-lvm:vm-100-disk-1,efitype=4m,pre-enrolled-keys=1,size=4M
ide0: none,media=cdrom
kvm: 1
machine: pc-q35-6.1
memory: 16384
meta: creation-qemu=6.1.0,ctime=1639170682
name: Windows11Enterprise
net0: virtio=1E:1A:04:A8:A4:43,bridge=vmbr0,firewall=1
numa: 0
onboot: 1
ostype: win11
scsi0: local-lvm:vm-100-disk-0,discard=on,size=64G
scsihw: virtio-scsi-pci
smbios1: uuid=dcd0fb75-7852-4284-bf25-b2dd1081d1da
sockets: 1
tablet: 0
tpmstate0: local-lvm:vm-100-disk-2,size=4M,version=v2.0
vga: virtio
vmgenid: 63ddbb5a-6fa9-43fe-b867-eb532025a82b

Windows 10 has no issue with this machine. Windows 11 works fine on another intel server I've got with all the same software.

What could be causing this 100% CPU usage?

 

[Nyoxy]

Set ssd flag on your hard disk.
scsi0: local-lvm:vm-100-disk-0,discard=on,size=64G,ssd=1

 

[spirit]

do you have tested with cpu model != host ? (maybe try epyc v2 model, just to be sure than unsupported qemu cpu flag are not passthrough to guest os)

 

[joao0]

None of the above suggestions helped unfortunately. I did try several different configurations and none have helped so far. The weird thing is that it usually happens at around the 10 minutes mark which could give some clues but I have no idea. Sometimes it stays up for several hours though. I may have to replace this epyc server with an Intel one...

 

[puertorico]

same issue here with Windows 11 on:

CPU(s) 12 x AMD Ryzen 5 5600X 6-Core Processor (1 Socket)

Kernel Version Linux 5.13.19-2-pve #1 SMP PVE 5.13.19-4 (Mon, 29 Nov 2021 12:10:09 +0100)

PVE Manager Version pve-manager/7.1-8/5b267f33

i was running with cpu = host - i have now changed it to KVM64 and will se if that helps.

 

[SimoF]

Same issue here with Ryzen 3600. Kernel 5.13.19-2-pve. Pve-manager 7.1-7

 

[vali20]

I experience the same. Ryzen 9 5900X. Windows 11 VM on Proxmox 7.1.2 and kernel 5.15. 5.13 was unusable for me, the NVIDIA Quadro GPU was not being passthrough correctly, giving the famous code 43 in Windows.

This platform is doomed. I have bought this CPU way back in the summer and have not been able to use it properly since then. Biggest mistake ever not to go with Intel or to wait for their next generation. AMD has interesting products, but lackluster software support.

Since Windows 11 was launched, the Windows Hypervisor Platform is bugged on this platform, in the way that only VMs with 1 vCPU are able to start. Thus, VMware and VirtualBox are unusable since launch day if you also want to use Hyper-V, which works fine but lacks a 3D video driver. Also, back in the summer, Proxmox was unusable on this CPU and a Windows VM, as nested virtualization was not working, spiting errors 1/2 boots, randomly. I put the idea to rest for the past months, and only recently decided to recreate the setup and see what's the current status. Indeed, things have improved, in the sense that it now almost works fully. VMware still has a bug where it crashes dwm.exe inside the guest unless you set dwm.exe to run on a single CPU, or boot the VM with just 1 vCPU. But at least with that, you can get decent nested virtualization.

The problem is this bug though. Randomly, after a couple of minutes, RDP drops, any file transfer drops, the web console shows 100% CPU usage and the VM hangs until restarted. Same issue as described here. I hope someone figures this soon, it's too bad these CPUs are let down by some many software issues...

 

[puertorico]

Other than this specific error, the Ryzen platform has been very stable for me the last year on proxmox, except Windows 11, but i made the decision early on to buy Asrock X570 server boards. Consumer boards has been a disater for me.

 

[vali20]

Other than this specific error, the Ryzen platform has been very stable for me the last year on proxmox, except Windows 11, but i made the decision early on to buy Asrock X570 server boards. Consumer boards has been a disater for me.

Yeah, indeed, I think I have a consumer X570-based Asrock motherboard, if I remember correctly (the system is at a remote site, and it’s the holidays, don’t fancy going there at this exact moment) and it is very crap indeed. Fan control is stupid, it’s either low RPM, or as soon as the CPU is pushed in any minor way, jet engine mode. The UEFI is bugged, they got the boot options indexing wrong, it always shows one less item in the F11 menu then if you look in the same list in the UEFI menu itself. POST is incredibly slow even with quick boot etc.

As I said, this platform as a whole is a mess. I also have a Core i9-9920x based proxmox setup with 2 Windows 10 VMs and it works flawlessly. Idk, indeed, too new stuff is dangerous, it could be Windows 11 for what is worth, Microsoft seemes to have screwed some stuff with it and are slow to release a fix for whatever reason.

I just wish we find a workaround for this problem as well, it’s the final hurdle to have a setup that works for my use case. Convoluted, but at least I can use this CPU and regret less investing in it.

 

[puertorico]

Yeah, indeed, I think I have a consumer X570-based Asrock motherboard, if I remember correctly (the system is at a remote site, and it’s the holidays, don’t fancy going there at this exact moment) and it is very crap indeed. Fan control is stupid, it’s either low RPM, or as soon as the CPU is pushed in any minor way, jet engine mode. The UEFI is bugged, they got the boot options indexing wrong, it always shows one less item in the F11 menu then if you look in the same list in the UEFI menu itself. POST is incredibly slow even with quick boot etc.

As I said, this platform as a whole is a mess. I also have a Core i9-9920x based proxmox setup with 2 Windows 10 VMs and it works flawlessly. Idk, indeed, too new stuff is dangerous, it could be Windows 11 for what is worth, Microsoft seemes to have screwed some stuff with it and are slow to release a fix for whatever reason.

I just wish we find a workaround for this problem as well, it’s the final hurdle to have a setup that works for my use case. Convoluted, but at least I can use this CPU and regret less investing in it.

Sound alot like those problems i ran in to, wired network was unstable network dropped out when hit with a load, fan control was a nighmare would not always restart but hang. I was ready to throw it in the garbage but the reason why i wanted the ryzen platform was beacuse off ecc to zfs and the performance in a small idle power since i run multiple servers in cluster. I have the same experience with intel, it just works, but no ecc was a dealbreaker


tried setting my windows 11 vm to cpu kvm64 it has not locked up so far

 

[joao0]

setting cpu=kvm64 disables hyper-v... only cpu=host works for me, except for the hanging

 

[vali20]

I need Hyper-V for the workloads that I run. I also don’t know where to look, dmesg doesn’t show anything interesting when the VM locks up. I don’t even know what’s to blame, that’s the thing…

 

[Deleted member 116138]

I run 2 systems on Ryzen 5900X on ASUS PRO WS ACE X570 boards with non-critical VMs (W10, W11, Server 2019/2022) and don’t have any of the observed problems. One PVE runs the 5.13 kernel, the other one 5.15.

What caused a lot of trouble in the past were ACPI and C-States. As soon as I disabled this within the Bios everything ran smoothly.

Maybe this is worth a try.

 

[joao0]

I run 2 systems on Ryzen 5900X on ASUS PRO WS ACE X570 boards with non-critical VMs (W10, W11, Server 2019/2022) and don’t have any of the observed problems. One PVE runs the 5.13 kernel, the other one 5.15.

What caused a lot of trouble in the past were ACPI and C-States. As soon as I disabled this within the Bios everything ran smoothly.

Maybe this is worth a try.

Do you have Hyper-V enabled on your W11 VM? this is what triggers the issue for me.

 

[SimoF]

Do you have Hyper-V enabled on your W11 VM? this is what triggers the issue for me.

To me it happens with any kind of virtualization, not only HyperV. Bluestacks or VMWare cause the same problem. I really don't know how to solve it...

 

[SimoF]

I've updated the kernel and everything seems to be finally working. Now I'm on
Linux pve 5.15.7-1-pve #1 SMP PVE 5.15.7-1 (Tue, 14 Dec 2021 16:42:34 +0100) x86_64 GNU/Linux

 

[joao0]

I've updated the kernel and everything seems to be finally working. Now I'm on
Linux pve 5.15.7-1-pve #1 SMP PVE 5.15.7-1 (Tue, 14 Dec 2021 16:42:34 +0100) x86_64 GNU/Linux

The latest kernel did not help me.

 

[SimoF]

The latest kernel did not help me.

Here is my config... Maybe it can help:

balloon: 2048
bios: ovmf
boot: order=scsi0;ide2;net0;ide0
cores: 6
cpu: host,hidden=1,flags=+pcid
efidisk0: zfs-vm:vm-102-disk-1,efitype=4m,pre-enrolled-keys=1,size=1M
hostpci0: 0000:09:00,pcie=1,x-vga=1
ide0: local:iso/virtio-win.iso,media=cdrom,size=543390K
ide2: local:iso/win11.iso,media=cdrom
machine: pc-q35-6.1
memory: 8192
meta: creation-qemu=6.1.0,ctime=1640341991
name: Win11
net0: virtio=32:EB:CD:CF:6F:A2,bridge=vmbr0,firewall=1
numa: 0
onboot: 1
ostype: win11
scsi0: zfs-vm:vm-102-disk-0,size=128G
scsihw: virtio-scsi-single
smbios1: uuid=d3a81362-2416-4261-a9e8-76e2beec477f
sockets: 1
tpmstate0: zfs-vm:vm-102-disk-2,size=4M,version=v2.0
vga: none
vmgenid: 751ab865-977b-43a3-9f08-aaac1e931a23

 

[parhedberg]

Same problem here. Running kernel 5.13.19-2-pve #1 SMP PVE 5.13.19-4
Having AMD Ryzen7 3700X.

No problem with Windows10 och 8 linux VMs.

Only Windows freezes up after aprox 10 minutes even if only idle..

 

[Krullix]

Windows 11 was stable with wsl1 without Hyper-V installed. It started locking up with high CPU when I went to install an android emulator and then again when I added Hyper-V for wsl2. It would not boot at first.

I changed the CPU from KVM64 to Host (AMD 3900x) - Did not resolve the issue.
Updated to Ver 7.1-10 kernel 5.15.12-3 (Booted)

The system would boot, but it would still at times become unstable CPU use 98-100% and would not respond to interrupts like CTRL-ALT-DEL or other commands had to use qm stop <id> to kill the system. The 10 minute observation seems close to what was occurring for me after resolving the boot issues.

Latest Changes:
HD: Added SSD emulation (ssd=1)
CPU: KVM64

I think I ran into a couple issues, the new kernel resolved some issues but there is also a problem with emulation/virtualization on the guest and it may be something more specific to AMD host cpu's.

I was able to toggle the CPU back to Host and Windows 11 has been stable overnight and hopefully it remains that way.

--Update--
The changes have not been stable, with little to no user activity the system went back into a high CPU state where the system / console stopped responding.