why does tensorrt use so much memory?

Hello,
How much memory tensorrt use when calls the function deserializeCudaEngine to deserialize the engine? My test results show that each process will use about 1.4G memory when inferring the network, is this normal? If so,tensorrt requested so much memory, what was it used for?
Tanks for your reply.

I also want to know why tensorrt request so much cpu memory?