The backupserver, and more importantly the sybmultbuf processes that are spawned to do the actual dump/load/compression work, run as separate OS processes outside of the ASE dataserver ... so you need to monitor CPU usage at the OS/host level.
The backupserver (and really the sybmultbuf processes) do not access ASE dataserver cache/memory. The backupserver will query sysdatabases, sysusages and sysdevices to figure out where the database fragments reside on the various devices. The sybmultbuf processes then read/write directly between the database devices and the dump devices ... completely bypassing the ASE dataserver cache. Net result is that the database dump does *not* churn the ASE dataserver cache.