Hello, Effectively, at the moment you can only assign 1 vgpu to VM, I think it's a limitation on the hypervisor, I'm not sure. A vGPU is just a subset of a single GPU defined in the CloudStack compute offering. it also require Nvidia licensing so you can enable nvidia grid, if I'm not mistaking. The number and size of vgpu depend on how you define them in CloudStack compute offering and you cannot mix different offering types on a single GPU, if the host have multiple GPU cards, you can mix GPU pass-through and vGPU on the same host.
Cloudstack code is relatively trivial to update for recent GPUs support. Code has been updated for T4 and v100 if I'm not mistaking. Cheers, On Wed, Jan 6, 2021 at 11:34 AM open infra <openinfr...@gmail.com> wrote: > Hi there, > > As per the documentation [1], I noticed that there is a limitation of > number os vGPU per VM. > 01. Is this applicable to other hypervisors such as KVM Hyper-V, etc as > well? > 02. When we can expect a CloudStack release that facilitate multiple vGPUs? > 03. Are there similar limitations for other devices such as SSD, Ethernet, > Sound card (attaching via PCI pass-through? > > > *'XenServer supports only one VGPU creation per VM, so admin/user cannot > create more than one VGPU for a VM.'* > > > [1] > > https://cwiki.apache.org/confluence/display/CLOUDSTACK/GPU+and+vGPU+support+for+CloudStack+Guest+VMs > > Thanks >
