Split a GPU Between Multiple Pods in Kubernetes

[@j0hnpedr0]

The first line alone got a like 😂

[@Jims-Garage]

Haha, thanks for the sympathy vote :D

[@monointent]

Completely under rated. The amount of times I have rewatched multiple videos multiple times to ingest what is actually going on. This man has helped me go from only ever really using hyperV to having a 3 node proxmox cluster. Thank you.

[@Jims-Garage]

@@monointent thanks! Glad I've been able to help you along the way. Sounds like a great setup.

[@MarcMcMillin]

this is great news!

[@Jims-Garage]

I agree, changed my lab (again!!!)

[@toddselby443]

Nice produce in your thumbnail. Wish I could find produce like that.

[@Jims-Garage]

It's rarely found in the reduced section 😂

[@richguy011]

Hey Jim, love your videos. I've seen that there are some issues with Intel Raptor Lake CPUs which made me pause into following your footsteps with the MS-01. Hope you end up not running into any problems and if you do good thing you are clustering :)

[@Jims-Garage]

@@richguy011 thanks, it has caused me some concerns but so far nothing to report.

[@richguy011]

@@Jims-Garage I look forward to seeing another video if you end up being impacted and how well your lab survives

[@bramvanneerven]

Nice video! Do you think the same is possible for Coral TPUs, such that you can - for example - use them for Frigate hardware acceleration and some other kind of AI service that benefits from them too?

[@Jims-Garage]

@@bramvanneerven probably but I haven't tried, I'm not sure if it's officially supported. I have since switched to the iGPU for frigate.

[@spexpl]

And there is a solution for nVidia or AMD?

[@Jims-Garage]

Yes, both. The Nvidia instructions are here: github.com/NVIDIA/k8s-device-plugin