When is it necessary to offload AI to GPU?

[Ri22o]

I am wanting to start using more CP.AI in BI and wanted to get ahead of it and possibly look into offloading the AI to a GPU but wasn't sure when it is necessary. Currently I am running AI for my LPRs and will get some massive CPU spikes, but that is also while I am doing some other setup/footage viewing on my BI PC.

When is it a good idea to consider a GPU? Are there any reasons not to?

PC: i7-10700k on an MSI Micro-ATX mobo with 32gb of ram.
Cameras: 25 4MP and 1 4K (going to add a couple more 4MPs and possibly another 4K)

UI3 usually reports the status below (or lower) except when CP.AI is running it can spike CPU to high 90s.

 

[fenderman]

I am wanting to start using more CP.AI in BI and wanted to get ahead of it and possibly look into offloading the AI to a GPU but wasn't sure when it is necessary. Currently I am running AI for my LPRs and will get some massive CPU spikes, but that is also while I am doing some other setup/footage viewing on my BI PC.

When is it a good idea to consider a GPU? Are there any reasons not to?

PC: i7-10700k on an MSI Micro-ATX mobo with 32gb of ram.
Cameras: 25 4MP and 1 4K (going to add a couple more 4MPs and possibly another 4K)

UI3 usually reports the status below (or lower) except when CP.AI is running it can spike CPU to high 90s.

View attachment 159589

The temp spike is fine. The reason for avoiding a GPU is the increased power consumption and cost of the gpu. Depending on your power supply you may need to upgrade that as well. They key thing to remember with AI if you are using it for alerts (Which is what most use it for) is that of you have several cameras covering a particular area you only really need one cam to utilize the AI.

 

[Ri22o]

I overbuilt the PC, so I am sure the PSU is sufficient, but I had also considered the issue with overlapping cameras all running AI at the same time for the same image and it being unnecessary.

As I get into messing with the AI and alerts I guess I will keep an eye on CP usage and whether or not it seems to be excessive.

 

[wittaj]

It also seems like many of the problems with CodeProject that people have are with the GPU version, so it your CPU can handle it, you are better off that way.

 

[MikeLud1]

You can try the Object Detection (YOLOv5 .NET) module, it will use the GPU that the CPU has (Intel® UHD Graphics 630 ). If you enable the Object Detection (YOLOv5 .NET) module you must disable the Object Detection (YOLOv5 6.2) module.