iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
GPU partitioning (GPU-P) in Hyper-V is a Windows Server feature, not a client-Windows feature. Microsoft’s documented implementation requires a host running Windows Server 2025 or later, a partition-capable server GPU, compatible vendor drivers, and firmware with SR-IOV and VT-d/IOMMU enabled. Windows 10/11 Pro or Enterprise can run the Hyper-V role, but Microsoft’s GPU-P troubleshooting guidance excludes those client hosts and desktop-class hardware.
GPU-P divides one physical GPU into hardware-defined partitions that can be assigned to virtual machines. That can increase VM density compared with dedicating one whole GPU to each VM, but it does not guarantee a particular level of performance. The supported GPU model, partition profile, driver, license, workload and server configuration all matter.
What GPU-P does—and what it does not do
Microsoft describes GPU partitioning as sharing one physical GPU device with multiple virtual machines. Each VM receives a portion of the GPU’s available resources rather than owning the whole adapter. The partition sizes and counts are exposed by the GPU and its OEM configuration; an administrator cannot invent an arbitrary profile.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesGPU-P is different from Discrete Device Assignment (DDA). DDA passes one or more complete physical GPUs through to a single VM. GPU-P is intended for workloads where several VMs need GPU access and higher density is more useful than exclusive hardware. DDA remains the alternative when a VM needs a dedicated device or the compatibility and performance profile of a full adapter.
#1 Best Overall
- Axial-tech fans now feature a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
- 2.5-slot design allows for greater build compatibility while maintaining cooling performance
- 0dB technology lets you enjoy light gaming in relative silence
- Dual BIOS switch lets you toggle between Quiet and Performance BIOS profiles
- Dual ball fan bearings last up to twice as long as sleeve bearing designs
| Decision axis | GPU-P | DDA |
|---|---|---|
| Resource model | One physical GPU is divided into VM partitions. | One or more complete physical GPUs are dedicated to one VM. |
| VM density | Potentially higher because multiple VMs share a GPU. | Lower because a GPU is assigned to one VM. |
| Best fit | Shared-GPU workloads where VM density is important. | Workloads that need a dedicated GPU and the DDA compatibility or performance profile. |
| Graphics and compute APIs | Vendor capabilities such as DirectX 12, OpenGL and CUDA are available when supported by the vendor’s GPU and driver. | Microsoft’s comparison also lists vendor capabilities such as DirectX 12, OpenGL and CUDA, subject to vendor support. |
GPU-P and DDA are mutually exclusive for a particular physical GPU. A device assigned through DDA cannot simultaneously provide partitions to GPU-P VMs.
Host support: the most important compatibility check
Required host operating system
For the Microsoft-documented feature, use Windows Server 2025 or later as the Hyper-V host. Microsoft’s troubleshooting guidance explicitly identifies GPU-P as unsupported on Windows 10/11 Pro client operating systems and on desktop-class hardware.
This distinction prevents a common mistake: Hyper-V being available on a client edition does not mean GPU-P is supported there. Windows 10 and Windows 11 Pro or Enterprise can install the Hyper-V role, but that general Hyper-V capability is separate from the server-only GPU-P support boundary.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Server and firmware requirements
- Install the Hyper-V role on Windows Server 2025 or later.
- Use a server-class platform and a GPU that supports partitioning.
- Enable processor virtualization support in BIOS or UEFI.
- Enable SR-IOV and VT-d/IOMMU as required by the platform.
- For clustered hosts, use the same GPU make, model and size on every server, with matching partition configuration.
- Cluster processors must support IOMMU DMA bit tracking. Microsoft gives AMD EPYC 7003 and later (Milan) and 4th Generation Intel Xeon Scalable processors (Sapphire Rapids) as examples for live-migration-capable systems.
AMD EPYC 7002 (Rome) supports GPU partitioning but not live migration with GPU partitioning, according to Microsoft’s live-migration guidance. If migration or failover is a requirement, check the processor generation rather than assuming that basic GPU-P support is enough.
Rank #2
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5070 Ti
- Integrated with 16GB GDDR7 256bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
Supported GPUs and software prerequisites
GPU models on Microsoft’s current list
Microsoft’s GPU-P overview names these supported models:
- NVIDIA A2
- NVIDIA A10
- NVIDIA A16
- NVIDIA A40
- NVIDIA L2
- NVIDIA L4
- NVIDIA L40
- NVIDIA L40S
- NVIDIA RTX Pro 6000 Blackwell Server Edition
- AMD Radeon PRO V710
The list is a starting point, not a blanket approval for every server, firmware revision, driver, license or workload. Confirm the exact server-and-GPU combination with the server OEM and GPU vendor. Check power draw, physical clearance, auxiliary power, cooling, firmware, driver branch and licensing before purchasing hardware.
Drivers, licensing and management tools
Install the GPU vendor’s partition-capable driver on every physical host and the corresponding guest driver inside each VM. A driver or licensing problem can prevent Windows from showing a device as available for GPU-P even when the hardware model appears on the supported list.
The Windows Admin Center GPUs extension must be version 2.8.0 or later for the current assignment procedure. PowerShell is an alternative, and administrative PowerShell commands must run elevated. NVIDIA GPU-P live migration specifically requires the driver included in NVIDIA vGPU Software v18.x or later and a license that supports GPU-P. Licensing terms vary by vendor and function, so verify them before deployment.
Rank #3
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5060
- Integrated with 8GB GDDR7 128bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
Guest operating-system compatibility
Guest support is a separate check from host support. Microsoft’s GPU-P troubleshooting guidance covers Generation 2 virtual machines running one of the following:
- Windows 10 or later
- Windows 10 Enterprise multi-session or later
- Windows Server 2019 or later
- Ubuntu 18.04, 20.04 or 22.04 LTS
Create a Generation 2 VM, then install the vendor’s guest GPU driver. Validate the exact image, kernel, driver branch and application combination against the current vendor documentation; a supported guest operating system alone does not prove that a particular application will work.
How to enable and assign GPU-P
The following workflow follows Microsoft’s Windows Server guidance. Treat it as a deployment sequence, not a substitute for the GPU vendor’s driver and licensing instructions.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →- Confirm the host boundary. Verify that the machine runs Windows Server 2025 or later, is server-class and is not a Windows 10/11 Pro client or desktop system. Install the Hyper-V role through Server Manager or the applicable Windows Server management method, then restart if Windows requests it.
- Configure firmware. Enter BIOS or UEFI and enable CPU virtualization, SR-IOV and VT-d/IOMMU. Save the changes and reboot. Firmware labels differ by manufacturer, so use the platform manual when an option is not obvious.
- Install the physical GPU and host driver. Fit a supported partition-capable GPU, install the vendor-provided host driver and reboot when required. On a cluster, repeat the process with the same GPU make, model and size on every server.
- Prepare a Generation 2 guest. Create or select a supported Windows or Ubuntu guest, install its current vendor GPU driver and shut down the VM before assignment if the management tool requires it.
- Check partitionability. Open an elevated PowerShell window and run
Get-VMHostPartitionableGpu | FL Name,ValidPartitionCounts. A partition-capable adapter should expose its name and the valid partition counts. In Windows Admin Center, the GPU status must be Partitioned. A status of Ready for DDA assignment is not sufficient for GPU-P. - Choose an exposed partition count. Use one of the counts returned by the GPU. The GPU OEM defines the valid set; do not type a profile that is not listed. The count determines how many partition slices the device can provide, while the exact resource allocation is defined by the hardware and driver.
- Assign the partition. In Windows Admin Center, use the GPUs extension to select the host GPU and assign a partition to the target VM, or use the corresponding elevated PowerShell workflow. Save application state first: Microsoft’s assignment guidance notes that the VM may need to be shut down and restarted. The current assignment guidance permits one GPU partition per VM, and the VM and partition must be on the same host.
- Validate inside the guest. Start the VM, confirm that the vendor device appears in the guest, check that the guest driver loads without an error, and exercise the intended graphics or compute workload. Confirm licensing before putting the VM into service.
- Validate every cluster node. For failover or migration, confirm that each host has the same GPU model, partition configuration, compatible driver and required firmware capabilities. A VM cannot migrate successfully to a node that lacks an equivalent supported configuration.
Partition limits and operational rules
Hardware-defined profiles
Each partitionable GPU exposes a predefined set of valid partition counts. Those choices are not interchangeable across models, and the number alone does not describe guaranteed application performance. Use the values reported by Get-VMHostPartitionableGpu and the GPU vendor’s profile documentation.
Rank #4
- Powered by Radeon RX 9070 XT
- WINDFORCE Cooling System
- Hawk Fan
- Server-grade Thermal Conductive Gel
- RGB Lighting
Automatic physical placement
GPU-P assigns a valid partition automatically. The administrator does not select a particular physical slice on the card. This simplifies provisioning but means that placement and capacity planning must follow the platform’s supported model rather than manual pinning.
One partition per VM in the current assignment workflow
Microsoft’s assignment and troubleshooting pages currently state that only one GPU partition can be assigned to a VM. The overview describes how multiple assigned partitions would appear as additional GPUs, but the explicit assignment limit is the practical constraint for the documented Windows Server workflow. Check the current Microsoft documentation before designing a configuration that depends on multiple partitions in one VM.
Cluster homogeneity
GPU-P is not supported with heterogeneous GPU configurations in a cluster. Match GPU make, model, size and partition counts across servers. Mixing unlike adapters can block placement, failover or live migration even when each individual host can run GPU-P.
Recommended Free Tools
Live migration: supported, but not free
Windows Server 2025 supports GPU-P live migration when the host, processor, GPU, driver, license and cluster configuration meet the required conditions. Microsoft’s implementation falls back to TCP/IP with compression. That can increase host CPU consumption and extend migration time compared with a migration that does not carry GPU state.
Best Value
- Axial-tech fans now feature a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
- Phase-change GPU thermal pad helps ensure optimal heat transfer, lowering GPU temperatures for enhanced performance and reliability
- 2.5-slot design allows for greater build compatibility while maintaining cooling performance
- Dual-ball fan bearings last up to twice as long as standard conventional sleeve bearings designs
- 0dB technology lets you enjoy light gaming in relative silence
Plan migration windows around the workload’s GPU memory and activity. Test the complete path between the actual source and destination hosts, including network capacity and CPU headroom. A cluster that can assign a partition locally is not automatically a cluster that can live-migrate that VM.
Troubleshooting common failures
The GPU does not appear as partitionable
- Confirm the host is Windows Server 2025 or later, not Windows 10/11 Pro.
- Check that the machine is server-class and that the exact GPU model is on the supported list.
- Recheck BIOS/UEFI settings for SR-IOV and VT-d/IOMMU, then reboot after changing them.
- Install the vendor’s GPU-P-capable host driver and check for licensing requirements.
- Run
Get-VMHostPartitionableGpu | FL Name,ValidPartitionCountsfrom an elevated PowerShell session and inspect whether any valid counts are returned.
Windows Admin Center shows the wrong status
Proceed only when the GPUs extension reports Partitioned. Ready for DDA assignment indicates a different assignment mode and does not establish GPU-P readiness.
The assignment fails or the VM will not start
- Shut down the VM and retry the assignment.
- Use a partition count exposed by the GPU rather than a manually chosen value.
- Ensure the VM and partition are on the same host.
- Remove any conflicting DDA assignment; one physical GPU cannot be used by DDA and GPU-P simultaneously.
- Check that the guest is Generation 2 and that its vendor driver is installed.
Migration fails between cluster nodes
Compare the source and destination GPU make, model, size, partition configuration, driver, firmware and licensing. Verify IOMMU DMA bit-tracking support on the processors. AMD EPYC 7002 systems can provide GPU partitioning but do not provide GPU-P live migration, so a Rome-based cluster cannot satisfy that migration requirement.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhen GPU-P is the right choice
Choose GPU-P when several compatible VMs need concurrent GPU access, the hardware’s predefined profiles fit the workloads and higher VM density matters. Choose DDA when one VM needs an entire GPU, when partitioning is unavailable for the selected adapter or when the application requires the dedicated-device behavior that DDA provides.
Do not select GPU-P solely because Hyper-V is available on a workstation. The supported host operating system, server platform, partition-capable GPU, firmware, drivers, licenses and guest image must all align before the design is supportable.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
