GPU Passthrough Performance issues inside Windows 10 VM

[lukgth]

Hello,
I'm having GPU Passthrough Performance issues inside a Windows 10 VM. On any benchmarks the GPU will reach 90-100% Usage but on most games, I usually see 40-60% usage, but in some games I get 90+% most of time. What could cause this?

Hardware:
HP Proliant DL380e Gen8
2x Intel Xeon E5-2430
64GB RAM (16GB allocated to guest)
GPUs: Nvidia Geforce GT 710, Geforce GTX 1050 Ti (assigned to guest)
Static High Performance Mode is selected. I also tried using a vBIOS that i dumped myself.

VM Config:

agent: 1
bios: ovmf
boot: order=virtio0;net0;sata2;sata0
cores: 5
cpu: host
efidisk0: local:110/vm-110-disk-3.qcow2,efitype=4m,pre-enrolled-keys=1,size=528K
hookscript: local:snippets/exec-cmds
hostpci0: 0000:0b:00,pcie=1,x-vga=1,romfile=gtx1050ti.rom
machine: pc-q35-6.2
memory: 16384
meta: creation-qemu=6.1.0,ctime=1638048727
name: <REDACTED>
net0: virtio=BA:03:90:3D:B7:70,bridge=vmbr3,firewall=1
numa: 1
onboot: 1
ostype: win10
sata0: none,media=cdrom
sata2: isos:iso/virtio-win.iso,media=cdrom,size=519096K
scsihw: virtio-scsi-pci
smbios1: uuid=cf92e406-7926-4977-a123-544a3ded84b2
sockets: 2
tpmstate0: local:110/vm-110-disk-4.raw,size=4M,version=v2.0
vga: none
virtio0: local:110/vm-110-disk-0.qcow2,cache=writeback,discard=on,size=80G
virtio1: <REDACTED>:110/vm-110-disk-0.qcow2,cache=unsafe,discard=on,size=64G
virtio2: <REDACTED>:110/vm-110-disk-1.qcow2,cache=unsafe,discard=on,size=120G
virtio3: <REDACTED>:110/vm-110-disk-0.qcow2,discard=on,size=400G
vmgenid: 372329f2-3f43-4a4f-86c5-0f6aa1f14c04

pveversion -v:

proxmox-ve: 7.2-1 (running kernel: 5.15.35-1-pve-relaxablermrr)
pve-manager: 7.2-3 (running version: 7.2-3/c743d6c1)
pve-kernel-5.15: 7.2-3
pve-kernel-helper: 7.2-3
pve-kernel-5.13: 7.1-9
pve-kernel-5.15.35-1-pve: 5.15.35-3
pve-kernel-libc-dev: 5.15.35-3
pve-kernel-5.15.35-1-pve-relaxablermrr: 5.15.35-1
pve-kernel-5.15.30-2-pve: 5.15.30-3
pve-kernel-5.13.19-6-pve: 5.13.19-15
pve-kernel-5.13.19-6-pve-relaxablermrr: 5.13.19-14
pve-kernel-5.13.19-5-pve: 5.13.19-13
pve-kernel-5.13.19-4-pve: 5.13.19-9
pve-kernel-5.13.19-3-pve: 5.13.19-7
pve-kernel-5.13.19-2-pve: 5.13.19-4
pve-kernel-5.13.19-2-pve-relaxablermrr: 5.13.19-4
pve-kernel-5.13.19-1-pve: 5.13.19-3
ceph-fuse: 14.2.21-1
corosync: 3.1.5-pve2
criu: 3.15-1+pve-1
glusterfs-client: 9.2-1
ifupdown: residual config
ifupdown2: 3.1.0-1+pmx3
ksmtuned: 4.20150326
libjs-extjs: 7.0.0-1
libknet1: 1.22-pve2
libproxmox-acme-perl: 1.4.2
libproxmox-backup-qemu0: 1.2.0-1
libpve-access-control: 7.1-8
libpve-apiclient-perl: 3.2-1
libpve-common-perl: 7.1-6
libpve-guest-common-perl: 4.1-2
libpve-http-server-perl: 4.1-1
libpve-storage-perl: 7.2-2
libspice-server1: 0.14.3-2.1
lvm2: 2.03.11-2.1
lxc-pve: 4.0.12-1
lxcfs: 4.0.12-pve1
novnc-pve: 1.3.0-3
proxmox-backup-client: 2.1.8-1
proxmox-backup-file-restore: 2.1.8-1
proxmox-mini-journalreader: 1.3-1
proxmox-widget-toolkit: 3.4-10
pve-cluster: 7.2-1
pve-container: 4.2-1
pve-docs: 7.2-2
pve-edk2-firmware: 3.20210831-2
pve-firewall: 4.2-5
pve-firmware: 3.4-2
pve-ha-manager: 3.3-4
pve-i18n: 2.7-1
pve-qemu-kvm: 6.2.0-6
pve-xtermjs: 4.16.0-1
qemu-server: 7.2-2
smartmontools: 7.2-pve3
spiceterm: 3.2-2
swtpm: 0.7.1~bpo11+1
vncterm: 1.7-1
zfsutils-linux: 2.1.4-pve1

 

[dcsapak]

are you sure you're not cpu limited in those games? you run on an older (sandy bridge; launched 10 years ago) platform with a relatively low turbo clock (2.70 GHz)

 

[lukgth]

are you sure you're not cpu limited in those games? you run on an older (sandy bridge; launched 10 years ago) platform with a relatively low turbo clock (2.70 GHz)

No, CPU Usage is around 30% with none of the cores at 100% but rather 25-40%.

 

[dcsapak]

i mean if the cpu usage can reach 100% in the vm, i guess the virtualization overhead is not the issue, leaves just another bottle neck (cpu/ram) or badly coded games (there are plenty...)
which games are those in particular?

 

[lukgth]

i mean if the cpu usage can reach 100% in the vm, i guess the virtualization overhead is not the issue, leaves just another bottle neck (cpu/ram) or badly coded games (there are plenty...)
which games are those in particular?

DiRT 3, DiRT Rally, GTA 5, Saints Row IV, Saints Row the Third, Euro Truck Simulator 2 and Teardown all seem to have this issue. Sometimes even lowering quality settings can result in lower FPS and (obviously) GPU Usage. While some games (e.g. Assetto Corsa, PC Building Simulator) do not have this issue and work as expected.

 

[dcsapak]

those games all strike me as cpu intensive (or at least not very optimized for multi core). your cpu can also be the bottleneck with less than 100% usage on a core (where did you measure that? in the guest or the host?)
depending on cache layout/branch prediction etc.

 

[lukgth]

(where did you measure that? in the guest or the host?)

Measured on Host and Guest. Host has two cores at around 60-80%, rest at 5-40%. On guest CPU usage is around 25-40% per core and total.