There is a common experience: Transmitting a large number of fragments is a waste of time. For example: when backing up and restoring website files, you generally don’t upload and download them directly, but habitually pack them into a package first...

This gameplay is the same. If you direct the cp command, the essence is to read a file and then write a file. Through tar pipeline relay, you can make full use of the size of the pipe buffer, which essentially means reading the size of a buffer and then writing the size of the buffer. In this way, small files can be read in batches at a time and written in batches at a time.

tar can do this because the tar package is a format that facilitates sequential inflow and sequential outflow. The tar format is basically: [TAR文件头][文件实际内容][TAR文件头][文件实际内容] …… [TAR文件头][文件实际内容][TAR文件尾], so the later level of tar can flow in a little bit of data for processing, without having to wait for the entire tar package to arrive before unpacking it.

Performance issues. The
cp command will read the files one by one and then write them to the new location. The
tar command will read as much of the file contents as possible before batch writing to the new location.

I think the only special thing is that tar can display the copied files on the screen, while cp displays nothing

I think another reason is that using tar processing can preserve file permissions. If it is cp, it is hard to say what permissions the final file will have.