Discrete Device Assignment compatibility boundary for NVMe and GPU passthrough across Windows Server 2016 and later releases

Hyper-V Discrete Device Assignment (DDA) allows a virtual machine to directly control a physical PCIe device, such as an NVMe controller or GPU, rather than using emulated virtual hardware. The feature was introduced with Windows Server 2016 and Windows 10, and its supported device classes and management options can differ across later releases such as Windows Server 2019, 2022, and Windows 11.

When planning a deployment that spans multiple host versions, the compatibility boundary between releases becomes a design concern. A device that meets reset and isolation requirements on one host version may not be assignable on another, and the loss of live migration or save/restore for VMs with assigned devices creates additional operational constraints. The host must also release the device from host driver ownership before assignment, which affects host availability during the transition.

What documented differences in assignable device classes or management behavior exist between Windows Server 2016, 2019, 2022, and Windows 10 or 11 for DDA? How should administrators verify that a specific NVMe controller or GPU remains compatible across these versions before production assignment? Are there documented cases where a device assignable on an earlier release becomes restricted on a later release due to driver or isolation requirements?