Vigyata.AI
Is this your channel?

Building a Homelab : Episode 6 vGPU Mastery - NVIDIA Grid P40 Sharing 4GB to 6 VMs on Proxmox

810 views· 34 likes· 48:35· Aug 9, 2025

🛍️ Products Mentioned (3)

Episode 6:Building a Homelab vGPU Mastery - NVIDIA Grid P40 Sharing 4GB to 6 VMs on Proxmox Project FastAPI-DLS https://git.collinwebdesigns.de/oscar.krause/fastapi-dls CREDITS: "Subscribe Button" by MrNumber112 https://youtu.be/Fps5vWgKdl0 Music I Use: https://www.bensound.com/free-music-for-videos License code: MS6H2K6QACHPUSXU Visit our sister channel Unkyjoe's Aquatics: https://www.youtube.com/channel/UCWHrQHkTZ1iACO8VxIXkBzw Thanks for watching! I hope you all enjoy... Join us on LBRY:https://lbry.tv/$/invite/@unkyjoesplayhouse:9 Facebook: https://www.facebook.com/UnkyjoesPlayhouse/ Twitter :https://twitter.com/ujplayhouse Email:unkyjoesplayhouse@gmail.com For PayPal or Patreon donations to Unkyjoe's Playhouse, please visit the "About" section on my channel. All cash donations are directly put back into Unkyjoe's Playhouse channel projects. I cannot respond to all emails, but give it a go! *PLEASE NOTE* I do not respond to YouTube or Google+ private messages. Please contact me via the official Facebook page or via my email address to get in touch.

About This Video

Well greetings, people of the internet—Uncle Joe here. In Episode 6 of my “Building a Homelab” series, I finally put the lab to work with a real-world goal: figuring out how to do remote AutoCAD and graphics work using virtual machines while sharing one big GPU across multiple VMs. I’m using my Dell R730 running Proxmox, and I’m taking one of my NVIDIA Tesla/Grid P40 cards (24GB VRAM) and slicing it into six vGPUs at 4GB each so six Windows 11 VMs can use it at the same time. The other P40 I leave dedicated to a single VM, just to keep a baseline. I walk through how I build a clean Windows 11 Enterprise 24H2 template, clone it, fix the VLAN/networking, add the PCI device with the right MDEV type, and then finish the Windows side: drivers, Remote Desktop, renaming, joining the domain, and pushing Group Policy by moving the machine into the right OU. I also show the “home lab reality” part—NVIDIA’s licensing is expensive, so I’m using an open-source Grid licensing server (FastAPI-DLS) running in a Docker VM with Portainer to demonstrate what’s possible. Once it’s licensed, I verify the vGPU is actually working with a quick FurMark run, and you can see the Grid P4-4Q show up and do its job.

Frequently Asked Questions

🎬 More from Unkyjoe's Playhouse