proxmox does not allow 2 rtx 3070 to use 1 internal error

[halil]

I'm sorry for my bad english.
My problem is, as I mentioned above, when I run one of the 2 GPUs, it gives an error. Currently, there are virtual machines working with 1 gtx1660ti and rtx3070. I do not have any problems with these, but I get an "internal error" error on the 2nd rtx 3070. I wonder why?

options vfio-pci ids=10de:2484,10de:228b,10de:2484,10de:228b,10de:2182,10de:1aeb disable_vga=1

[B][I]args: -cpu 'host,+kvm_pv_unhalt,+kvm_pv_eoi,hv_vendor_id=NV43FIX,kvm=off' boot: order=sata0;net0;hostpci0;usb0;usb1 cores: 2 cpu: kvm64,hidden=1,flags=+pcid hostpci0: 0000:02:00,pcie=1 machine: q35 memory: 6144 meta: creation-qemu=6.2.0,ctime=1664666433 name: win10.4coreRTX3070 net0: e1000=BE:D7:E7:CB:60:65,bridge=vmbr0,firewall=1 numa: 0 ostype: win10 sata0: local-lvm:vm-101-disk-0,size=65G scsihw: virtio-scsi-single smbios1: uuid=998fa525-dadc-48f6-9869-871de9a134ac sockets: 2 usb0: host=1-3,usb3=1 usb1: host=1-4,usb3=1[/I][/B] [I][B]vmgenid: a56430f3-62c8-479e-a532-963cb424d445[/B][/I]


I have 2 questions.
Can 1 gpu be deployed for 2 vms?

 

[leesteken]

My problem is, as I mentioned above, when I run one of the 2 GPUs, it gives an error. Currently, there are virtual machines working with 1 gtx1660ti and rtx3070. I do not have any problems with these, but I get an "internal error" error on the 2nd rtx 3070. I wonder why?

My guess is that he third GPU is not alone in its IOMMU group. It would help if you can tell us the make and model of your motherboard d show the IOMMU groups.

options vfio-pci ids=10de:2484,10de:228b,10de:2484,10de:228b,10de:2182,10de:1aeb disable_vga=1

There is no need to use the same ID multiple times. All devices with that ID are bound to vfio-pci (which can be a problem if you want to use for Proxmox and one for a VM).

[B][I]args: -cpu 'host,+kvm_pv_unhalt,+kvm_pv_eoi,hv_vendor_id=NV43FIX,kvm=off' boot: order=sata0;net0;hostpci0;usb0;usb1 cores: 2 cpu: kvm64,hidden=1,flags=+pcid hostpci0: 0000:02:00,pcie=1 machine: q35 memory: 6144 meta: creation-qemu=6.2.0,ctime=1664666433 name: win10.4coreRTX3070 net0: e1000=BE:D7:E7:CB:60:65,bridge=vmbr0,firewall=1 numa: 0 ostype: win10 sata0: local-lvm:vm-101-disk-0,size=65G scsihw: virtio-scsi-single smbios1: uuid=998fa525-dadc-48f6-9869-871de9a134ac sockets: 2 usb0: host=1-3,usb3=1 usb1: host=1-4,usb3=1[/I][/B] [I][B]vmgenid: a56430f3-62c8-479e-a532-963cb424d445[/B][/I]

Instead of those args: I would just enable Primary GPU (and set CPU to host) but this appears to work for you, so let's not change this.

I have 2 questions.
Can 1 gpu be deployed for 2 vms?

Not with PCIe passthrough as you are doing now. You need a GPU that support mediated device passthrough, like Intel integrated graphics of NVidia GRID.

 

[halil]

asus tuf b460 pro wifi, i7 10700 cpu, 48gb ram,
I removed one of the 3070, thinking that I could not continue the process. I connected GTX1660s, rx580 and now "welcome to grub!" letter before me. I turned off Iommu, it still continues