How much YOLO v8n GPU VRAM will be consumed with two running model inference on 1080p camera feed. on each inference two YOLO model will be running. I will run this on rtx 3060ti and rtx 3080ti. And what will be the maximum number of inference I will be able to run on each one of these GPUs where each inference will use two models at least?
The best way to answer this question is to test, as there are many factors that can influence the usage of vRAM. Things like video feed, encoding, GPU driver, etc. that will change the amount of vRAM used/available.