Hi, I have an old system running PVE for about 5 years now, with 5 containers and one Win10 VM. All have worked flawlessly. But recently I wanted to test GPU pass through and managed to get Windows to detect it as the primary graphics card. But its incredibly slow now EG click the start menu and takes 10 seconds to draw the menu. Let alone navigating windows. This is prior to using any apps. I know my RAM is probably too small, but seemed to work fine without the GPU.
Could someone advise why the bottleneck, is it my system is really to under specced for the GPU, as Win10 works fine without it.
I connect via Windows RDP or Linux KRDC over local LAN.
Attached is what Win10 VM and my PVE node is showing in terms of performance. (Yes RAM is maxed but everything else seems un taxed)
I also powered down all the other VM's to see if it made any difference it does not. Disconnect the GPU from the VM and it speeds back up to normal.
I am not using GRUB, so updated /etc/kernel/cmdline to be "root=ZFS=rpool/ROOT/pve-1 boot=zfs quiet intel_iommu=on iommu=pt"
System
OS: Proxmox VE 7.4-17 x86_64
Kernel: 5.15.126-1-pve
Uptime: 18 hours, 13 mins
Packages: 838 (dpkg)
Shell: bash 5.1.4
Resolution: 1680x1050
Terminal: /dev/pts/0
CPU: Intel i7-3770 (8) @ 3.900GHz
GPU: Intel HD Graphics
GPU: NVIDIA GeForce GTX 1650 SUPER
Memory: 14941MiB / 15883MiB
VM Conf
#Win 10 with Fusion 360
agent: 1
bios: ovmf
boot: order=scsi0;net0
cores: 4
efidisk0: local-zfs:vm-105-disk-1,efitype=4m,pre-enrolled-keys=1,size=1M
hostpci0: 0000:08:00,pcie=1
machine: pc-q35-6.1
memory: 8092
meta: creation-qemu=6.1.0,ctime=1639389574
name: WindowsProxmox
net0: e1000=7A:BE:A7:32:A8:1A,bridge=vmbr0,firewall=1
numa: 0
ostype: win10
sata0: HDD-RAID:vm-105-disk-0,backup=0,size=200G
scsi0: local-zfs:vm-105-disk-0,cache=writeback,discard=on,size=64G
scsihw: virtio-scsi-pci
smbios1: uuid=b424157a-d617-44b3-80a4-371e74ccc9c4
sockets: 2
tpmstate0: local-zfs:vm-105-disk-2,size=4M,version=v2.0
vga: none
vmgenid: 567fe49b-e8d5-4423-9f58-38d37eb1f2b2
Could someone advise why the bottleneck, is it my system is really to under specced for the GPU, as Win10 works fine without it.
I connect via Windows RDP or Linux KRDC over local LAN.
Attached is what Win10 VM and my PVE node is showing in terms of performance. (Yes RAM is maxed but everything else seems un taxed)
I also powered down all the other VM's to see if it made any difference it does not. Disconnect the GPU from the VM and it speeds back up to normal.
I am not using GRUB, so updated /etc/kernel/cmdline to be "root=ZFS=rpool/ROOT/pve-1 boot=zfs quiet intel_iommu=on iommu=pt"
System
OS: Proxmox VE 7.4-17 x86_64
Kernel: 5.15.126-1-pve
Uptime: 18 hours, 13 mins
Packages: 838 (dpkg)
Shell: bash 5.1.4
Resolution: 1680x1050
Terminal: /dev/pts/0
CPU: Intel i7-3770 (8) @ 3.900GHz
GPU: Intel HD Graphics
GPU: NVIDIA GeForce GTX 1650 SUPER
Memory: 14941MiB / 15883MiB
VM Conf
#Win 10 with Fusion 360
agent: 1
bios: ovmf
boot: order=scsi0;net0
cores: 4
efidisk0: local-zfs:vm-105-disk-1,efitype=4m,pre-enrolled-keys=1,size=1M
hostpci0: 0000:08:00,pcie=1
machine: pc-q35-6.1
memory: 8092
meta: creation-qemu=6.1.0,ctime=1639389574
name: WindowsProxmox
net0: e1000=7A:BE:A7:32:A8:1A,bridge=vmbr0,firewall=1
numa: 0
ostype: win10
sata0: HDD-RAID:vm-105-disk-0,backup=0,size=200G
scsi0: local-zfs:vm-105-disk-0,cache=writeback,discard=on,size=64G
scsihw: virtio-scsi-pci
smbios1: uuid=b424157a-d617-44b3-80a4-371e74ccc9c4
sockets: 2
tpmstate0: local-zfs:vm-105-disk-2,size=4M,version=v2.0
vga: none
vmgenid: 567fe49b-e8d5-4423-9f58-38d37eb1f2b2