luckylinux
Limp Gawd
- Joined
- Mar 19, 2012
- Messages
- 225
Dear all,
I know it might be a strange request but please bear with me ...
I started using VMWARE ESXi about 5 years ago as my virtualization platform of choice in my home lab. Since then, due to easier backup, scripting and base OS (Debian), I switched to Proxmox VE (without subscription). All latest updates have been applied and machine has been rebooted.
For Linux / FreeBSD / OpenBSD VMs there is no problem at all. With Windows 10 VM however, it seems that it leads to very high CPU usage. I wanted to upgrade and virtualize my Desktop using Server grade hardware (Supermicro X9dri-LN4F+, 2 x E5-2697 v2, 256GB ECC RAM) and even without other VMs on the server, Windows 10 takes about ~5 minutes to boot, stays at 100% CPU usage on all 8 virtual cores during that time - 16GB virtual RAM, after which it drops to about ~30% CPU at idle.
Somehow libvirt users seem to have fixed this kind of behaviour with the "hpet" option, but since proxmox VE defaults to "-no-hpet" and a few other options have been introduced to supposidly fix the problem for Windows 10 (hyperv-stimer and another one), I am a bit at a loss about what to do ...
I would like to stay with Proxmox VE however this is definitively not usable. I tried to passthrough my GTX 1060 but also that didn't work so far with Windows 10. On other machines I could passthrough GTX 1070/1080 without issues to Linux VMs without any special option. For Windows 10 I tried to add the romfile option but also that didn't help ...
Would you suggest me to switch over to ESXi for this application? I still have one Napp-IT subscription I can use however I didn't like OmniOS too much (seemed like a dead project at one point, no more updates). Or ... just nest Proxmox VE inside ESXi and setup NFS server on Debian/Proxmox :S :S :S ?
EDIT: ESXi Limitation of 8 vCPU per VM is too tight IMHO. Other options with some management tool? XCP-NG maybe with Xen Orchestra ?
I know it might be a strange request but please bear with me ...
I started using VMWARE ESXi about 5 years ago as my virtualization platform of choice in my home lab. Since then, due to easier backup, scripting and base OS (Debian), I switched to Proxmox VE (without subscription). All latest updates have been applied and machine has been rebooted.
For Linux / FreeBSD / OpenBSD VMs there is no problem at all. With Windows 10 VM however, it seems that it leads to very high CPU usage. I wanted to upgrade and virtualize my Desktop using Server grade hardware (Supermicro X9dri-LN4F+, 2 x E5-2697 v2, 256GB ECC RAM) and even without other VMs on the server, Windows 10 takes about ~5 minutes to boot, stays at 100% CPU usage on all 8 virtual cores during that time - 16GB virtual RAM, after which it drops to about ~30% CPU at idle.
Somehow libvirt users seem to have fixed this kind of behaviour with the "hpet" option, but since proxmox VE defaults to "-no-hpet" and a few other options have been introduced to supposidly fix the problem for Windows 10 (hyperv-stimer and another one), I am a bit at a loss about what to do ...
I would like to stay with Proxmox VE however this is definitively not usable. I tried to passthrough my GTX 1060 but also that didn't work so far with Windows 10. On other machines I could passthrough GTX 1070/1080 without issues to Linux VMs without any special option. For Windows 10 I tried to add the romfile option but also that didn't help ...
Would you suggest me to switch over to ESXi for this application? I still have one Napp-IT subscription I can use however I didn't like OmniOS too much (seemed like a dead project at one point, no more updates). Or ... just nest Proxmox VE inside ESXi and setup NFS server on Debian/Proxmox :S :S :S ?
EDIT: ESXi Limitation of 8 vCPU per VM is too tight IMHO. Other options with some management tool? XCP-NG maybe with Xen Orchestra ?
Last edited: