Planning hardware changes - CP.AI on separate server?

[subterminal]

I've been a lazy admin and have let my Blue Iris server sit without any updates or attention for a few years. Since I'm planning to add more cameras, it's time to add more hardware.

My main question is: Does anyone run CP.AI on a separate server from your Blue Iris server? If so, does the 'use GPU' setting work well [with the GPU being in the CP.AI server]?

My current setup:
Single physical server running KVM with a dedicated Windows10 VM for Blue Iris. The CPU hates life because the onboard APU is already being passed-through to a different VM, so CPU load spikes when doing transcoding/saving clips.

Plan:
Build a new 1u server specifically dedicated to Blue Iris. Then, add a GPU to the original server and spin up a new VM for CP.AI. Point Blue Iris server towards CP.AI server and...profit?

 

[Edcfish]

I run CP AI on a separate PC from my BI install.
My AI PC also runs my media server, ad blocker, wifi controller and a few other odds and ends.
8 camera using AI, and I do not use GPU.
My AI PC cpu rarely goes above 40% and my BI PC is usually around 20% with substreams.

 

[CanCuba]

If you have a Raspberry Pi not doing anything, you can install CPAI on that and buy a Google Coral TPU. Faster than any GPU and only draws 2W.