Event banner
Supercharge your datacenters with Hyper-V and virtualized GPUs
Event details
This session will dive into new GPU features on Windows Server and Azure Stack HCI, including GPU failover clustering and GPU partitioning (GPU-P). GPU-P, or GPU virtualization, is a new feature that allows users to share a physical GPU-P device with multiple virtual machines. With GPU-P, each virtual machine (VM) gets a dedicated portion of the GPU capacity instead of the entire GPU. This translates into more cost efficiency for your organization as you can assign GPU resources where needed. Join us as we share more on this new feature and run through multiple exciting demos.
Speakers: Nicole Bourain, Afia Boakye
Thanks for tuning in to the Windows Server Summit on demand!
9 Comments
- Char_CheesmanBronze Contributor
Thank you for joining us this week for the Windows Server Summit! Q&A is now closed, but all sessions are available on demand so you can watch and learn when it is convenient for you. We hope you enjoyed the event.
- ciscenCopper ContributorRegarding GPU partitioning, can we partition a single GPU into multiple profiles, such as creating 2GPU and 4GPU profiles from the same GPU?
- Nicole_Bourain
Microsoft
Yes, that is correct. There is a max number of partition counts that can be created, and that depends on the specific GPU driver.
- py300meCopper ContributorWill GPU-P someday be formally avaliable on Client Hyper-V as a feature with a configuation UI? Indeed many home users are using it through scripts, and Windows Sandbox seems to be using it. Moreover, recent Readon & Geforce consumer GPUs are also supporting partitioning.
- Nicole_Bourain
Microsoft
Thanks for your question and insights. Currently we are focused on Azure Stack HCI & Windows Server for GPU-P support.
- Char_CheesmanBronze Contributor
Welcome! Supercharge your datacenters with Hyper-V and virtualized GPUs is starting now. If you have any questions or feedback for our product teams, please post them here in the Comments.
- Will this work with any nVidia GPU or are there minimum requirements for the architecture of the GPU?
- GoaMetzCopper Contributor
Short answer: Yes.
I implemented P-GPU countless times, with everything from a 1050 up to 4090 and RTX-A Series and i got it working always. Consumer GPUs might not be officially supported or recommended, but they work just fine. Even the CPU integrated GPUs of Ryzen 7 & 9 can be used this way, it's not much, but better than nothing and it makes a big difference for video decoding, office workloads and the very GPU heavy internet browser.