Few basic points could alleviate your fear.
Running autovaccum will eventually improve the performance of queries. It is not something to harm the performance. This is because autovaccum does the garbage collection of dead tuples in table. so queries need to scan though less number of tuples.
In your question, it is not clear how you analysed that autovaccum is creating performance problem. if possible please provide details.
Autovaccum algorithms/logic in Postgres is continuouly improved in new versions. for example, Version 9.2 received a patch for skipping a table if there is lock. so autovaccum will do better job in handling bloat. This logic is further improved to skip blocks which are having locks. in latest version 9.6 (which is in development) received a patch to completely skip all the blocks where all tuples are already freezed.
Robert Hass posted about this recently: http://rhaas.blogspot.in/2016/03/no-more-full-table-vacuums.html
You may want to fine tune autovaccum related parameters to suit your system. if possible take the help of expert DBA. Each parameter is documented very well in Postgres documentation : http://www.postgresql.org/docs/9.5/static/runtime-config-autovacuum.html
So Regarding your question : "Is there a way to make it not consume as much memory/CPU so that client queries can run more efficiently?"
Answer is yes. by fine tuning the parameters mentioned in the above link you can achieve the best results for your system
A general advice will be to always upgrade to latest stable version of PostgreSQL to take advantage of all the autovaccum improvements.
Always we should keep in mind that autovaccum is very necessary for keeping the performance of the database. If there is CPU or Memory starvation in your system, consider getting better machine.
otherwise tables will become bloated and performance goes down over a period of time.