Posting activity monitor probably wouldn't help. But, if you can post the code that exhibits this behavior would be helpful. Even better would be if you can post the DDL for the tables and scripts to populate those tables along with the query so someone can copy it and run it in their environment.
This type of random poor performance can happen for a number of reasons:
There is extra load on the server
There is resource contention - for example, a table the query needs is locked by another process/query
Poor query execution plan for certain combinations of input parameters (aka parameter sniffing).