Server 2025 High CPU

[mubbasher.hussain]

Hi all,

I am currently testing Proxmox as a potential VMware replacement and have built a new Windows Server 2025 VM from ISO. I am experiencing unusually high CPU usage inside the VM.

This is not a migrated VM! it is a clean installation of Windows Server 2025.

The VM was installed successfully, joined to the domain, and the only Windows role enabled so far is File Server role. No additional applications have been installed.

The issue is that CPU usage remains unusually high compared to what I would expect from an idle file server. Norm interaction with the VM (RDP session, opening Server Manager, browsing folders, etc.) causes noticeable CPU spikes.

Below is my Environment

• Proxmox VE: 9.x
• Guest OS: Windows Server 2025
• CPU: Intel Xeon Gold 6252
• vCPU assigned: 48
• Storage: HPE MSA 2050 SAS SAN via multipath
• Storage type: LVM-thin
• Disk controller: VirtIO SCSI
• BIOS: OVMF (UEFI)
• Machine type: Q35

Troubleshooting completed​

• Installed latest VirtIO drivers.
• Tested different VirtIO driver versions now using virtio-win 0.1.271 (as i know 0.1.285 has a issue with server 2025)
• Tested different CPU types:
  • x86-64-v2-AES
  • x86-64-v3
  • x86-64-v4
• Tested different disk cache options.
• Enabled IO Thread.
• Verified SAN performance and health.

The same underlying SAN is currently used by our VMware environment without any performance issues.
At this stage I am trying to understand if there are any recommended Proxmox settings specifically for Windows Server 2025 guests.

Are there any known issues or recommended settings around:

• Windows Server 2025 and KVM/QEMU
• VirtIO driver versions
• CPU type selection
• Hyper-V enlightenments
• VBS / Core Isolation
• QEMU Guest Agent
• NUMA configuration

Any advice would be appreciated as we are evaluating Proxmox for our production environment.


Thanks.

Bash

 

[longer]

hmmm there currently known issue explained in:

Post in thread 'Windows Server 2025 CPU suddenly at 100%'

Jul 13, 2026

Since several users have encountered this issue, these two workarounds have proven to be a current solutions:

The easiest option is to select a generic CPU for the VM, for example, x86-64-v2-AES [0].
The second option would be to create a custom CPU model based on the host and then remove the vmx-apicv-vid flag.
This should allow the VMs to work normally again. Feel free to test.


[0] https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_qemu_cpu_types
[1] https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_custom_cpu_types

• mariol

Where its recommended to use CPU type other than host CPU. So you can roll out this and keep CPU type as x86-64-v2-AES.

Please share the vm config:
root@pve ~# qm config [VMID]


Also if you can share journalctl log from observed issue time from the pve host:
https://learning.lpi.org/en/learning-materials/102-500/108/108.2/108.2_02/

 

[Johannes S]

In theory host with disabled nested virtualization shoukd get the best performance. x86-64-v2-AES is a quite old architecture and afik not even offically supported on recent windows systems. so I would definitively try out the others.
Following tool is quite handy to determine the maximum possible generic type: https://github.com/credativ/ProxCLMC

 

[crawrj]

Are you actually using 48 vCPU? Or is that a typo? If you are using 48, that is probably the issue. I am not an expert, so I could be wrong, but overkilling the vCPU causes issues, from what I understand. In contrast, I have a 2025 file server with 4 vCPU that runs flawlessly and stays at 6-8%.