Dear NVIDIA Enterprise Support Team,
Good Morning.
We are currently evaluating NVIDIA H200 GPUs for our compute-based virtualization (vGPU / mdev) proof of concept using Proxmox (KVM).
Our goal is to enable GPU slicing (MIG/vGPU) so that multiple virtual machines can share H200 resources for AI/ML workloads in our internal HPC environment.
During the evaluation, we noticed that H200 vGPU (mdev) is not currently supported under the vGPU C-Series 17 (R580) release.
We would appreciate your clarification on the following points to help us plan our deployment and licensing:
Key Queries
-
Please confirm the list of currently supported GPUs for vGPU (C-Series) under the R580 branch (e.g., A100, H100, L40, etc.).
-
When is vGPU/mdev support expected for the H200 (Hopper) platform — will it be available in C-Series 18 (R590) or later?
-
Are there any firmware or driver prerequisites required for H200 vGPU enablement once released?
-
Could you share the licensing details and cost model for NVIDIA vGPU Enterprise (C-Series) — specifically for compute workloads on KVM environments?
-
Is NVIDIA AI Enterprise licensing mandatory for vGPU usage on Hopper GPUs such as H100/H200?
We would also appreciate any official documentation or roadmap that outlines future H200 vGPU support timelines.
Thank you for your guidance and support.
Regards,
Manu