what GPU do i need for CodeProject.AI with my HP EliteDesk 800 G3 SFF?
- Thread starter tung256
- Start date
What cameras do you have?
Most of us find the camera AI to be more than adequate.
Many have found the Coral USB to be problematic.
Many have found the CPU version of CodeProject to be easier to use and less hassles.
How many cameras do you plan to use CodeProject on? If just a couple try the CPU version first.
The GPU is not necessary unless you plan to run it on lots of cameras and the CPU is spiking.
Many here don't have a GPU.
Most of us find the camera AI to be more than adequate.
Many have found the Coral USB to be problematic.
Many have found the CPU version of CodeProject to be easier to use and less hassles.
How many cameras do you plan to use CodeProject on? If just a couple try the CPU version first.
The GPU is not necessary unless you plan to run it on lots of cameras and the CPU is spiking.
Many here don't have a GPU.
For running CodeProject.ai, a dedicated GPU can significantly enhance performance, especially for AI tasks. The GIGABYTE GeForce RTX 3050 OC you mentioned should work well with your HP EliteDesk 800 G3, assuming your PSU supports it and you have sufficient space. Regarding using your Coral USB stick, you can run TensorFlow Lite models efficiently, leveraging your existing setup. If you're primarily focusing on lightweight models or inference tasks, your Coral device might suffice without a dedicated GPU. However, for more intensive training or larger models, upgrading with the RTX 3050 would be beneficial. 
looking at this video, , gigabyte recommends 300w psu. meanwhile, my HP is only 180w.For running CodeProject.ai, a dedicated GPU can significantly enhance performance, especially for AI tasks. The GIGABYTE GeForce RTX 3050 OC you mentioned should work well with your HP EliteDesk 800 G3, assuming your PSU supports it and you have sufficient space. Regarding using your Coral USB stick, you can run TensorFlow Lite models efficiently, leveraging your existing setup. If you're primarily focusing on lightweight models or inference tasks, your Coral device might suffice without a dedicated GPU. However, for more intensive training or larger models, upgrading with the RTX 3050 would be beneficial.
so you are saying i can still install CodeProject.ai but choose a TensorFlow Lite model for Coral? i just need person detection. nothing fancy like facial/car/dogs/cats/etc...
i have a variety of IP cams...
1 Lorex 4k, wired
1 reolink 4k, wired
1 amcrest 2k doorbell, wifi
4 Tapo cams 2k, wifi
2 Galayou 2k, wifi
For what it is worth, the 3050 6GB will be a great choice for Code project AI and Blue Iris.
My personal requirement is 16 HD cameras running as efficiently as possible.
Low power usage to keep the power bill reasonable and help the UPS last longer.
The 3050 6GB in my system with 16 cameras handles the load just fine while using less than 70 watts.
Using CP AI explorer and my test images, inferences are around 30ms on Yolov5 6.2
I replaced EVGA 1650 cards which had inference times were around 200ms with the same test images.
Double check the specs on any card you consider to make sure it is only 70 watts and only uses the PCIe power without an additional connector.
This is assuming that you value power efficiency.
My personal requirement is 16 HD cameras running as efficiently as possible.
Low power usage to keep the power bill reasonable and help the UPS last longer.
The 3050 6GB in my system with 16 cameras handles the load just fine while using less than 70 watts.
Using CP AI explorer and my test images, inferences are around 30ms on Yolov5 6.2
I replaced EVGA 1650 cards which had inference times were around 200ms with the same test images.
Double check the specs on any card you consider to make sure it is only 70 watts and only uses the PCIe power without an additional connector.
This is assuming that you value power efficiency.
Last edited: