Suggestions for good hardware for object detection using Yolov7

Question

I need some suggestions on what is the best hardware to do object detection using YoloV7. I have been using Nvidia's Jetson hardware like Nano and Xavier. I have tried running and optimizing Yolov7 tiny models but no luck in acheiving good speed, so have been using ssd-mobilenet models. These ssd models are fast but a bit less accurate and thus now I want to use my own custom built machine so that I can easily use Yolov7. My ultimate goal is to make 3 versions of the machine i.e.

1. Standard : which can do object detection from 2 cctv cameras
2. Medium: which can do object detection from 4 cctv cameras
3. Heavy: which can do object detection from 6 cctv cameras

But to start with, I just want suggestions for standard. So please if anyone can suggest below specs:

1. CPU
2. GPU
3. RAM

I have been paying ~500GBP for Jetson Nano and it was only able to interface 2 cameras with FPS of 20-23FPS using SSD mobilenet. So for standard PC, anything below 500GBP would work. Thanks