Here is a basic explanation of what the CPU is used for. The CPU is used for the following
Decoding the incoming RTSP stream
Encoding any outgoing annotated RTSP
Resizing frames before being passed to the analytic engine
Preprocessing before the frame is passed to the GPU for the DL tracker to process
The first 3 points are currently performed for all trackers, the standard motion object tracker and the DL trackers.
The last point is performed when using the DL trackers and requires some additional resources from the CPU. As a result, the number of channels that can be supported on a particular CPU is reduced when using the DL trackers.
This chart is giving some overview how VCA can perform with GPU. Different scenarios may affect performance.
Memory Bandwidth (GB/sec)
Actual channels DLOT tested
GeForce RTX 3070
GeForce RTX 2080 Ti
GeForce GTX 1660 SUPER
GeForce GTX 1650