When creating a nci on a clustered table, trail and error approach is needed to determine how selective the query will be. High selectivity = good; Low selectivity = bad. For example:
select col1, col2, col3 from tableA where col1 >= 60000
You can then look at the estimated execution plan and see if it uses an index seek or scan. Seek is what you are looking for along with maybe a key lookup on the clustered index.
To calculate the selectivity of a query, take the number of returned rows and divide it by the total rows. In addition, make sure you turn on SET STATISTICS IO ON to see the logical reads, physical reads, and read-ahead reads.
For more information, this site seem like a good read.