imported esxi vm slower than created one with CPU "Host"

K

kriev98

New Member
Feb 14, 2025
8
0
1
Hey,

I'm importing VMs from ESXI to Proxmox. For the imported VMs, if i select "host" for CPU they are slower than VMs created in Proxmox. Most notable impact is on speedtests. Both VMs have 4 cores, 4gb RAM, Windows 2022. The imported VMs get half the speed results from the same test server.

If i change the cpu type to AMD Genoa (the host is a AMD Epyc 9454) on the imported VM, i get the same performance as the proxmox created VM with CPU "Host"

Is this because the original VM was created with an Intel CPU and i need to emulate the AMD CPU for these VMs?
 
It could be because when you installed Windows on an Intel CPU it decided that certain features were available and said features are not existing or differently implemented on AMD. So it is falling back to a generic version that is slower.
 
Last edited:
BobhWasatch said:
It could be because when you installed Windows on an Intel CPU it decided that certain features were available and said features are not existing or differently implemented on AMD. So it is falling back to a generic version that is slower.
Click to expand...
BTW, PVE does not emulate AMD instructions when running on Intel or vice versa. The CPU flags are really just visibility flags. They can hide features but not create them. The purpose of this is to facilitate live migration between not-identical hardware.
 
BobhWasatch said:
installed Windows on an Intel CPU it decided that certain features were available and said features are not existing or differently implemented on AMD. So it is falling back to a generic version that is slower.
Click to expand...
where you find this information ?
There is another explication.

@kriev98 :
post VM imported vs clean.
speedtests depends of vNIC type where virtio is faster than virtual Intel.
Post others metrics like cpu benchmark imported vs clean.
 
Last edited:
You must log in or register to reply here.
Top Bottom