If all the columns are used to build the index, this become a copy of original table with column values stored differently, column oriented instead of row oriented.
Data in table is treated as readyonly, good for data warehousing analytic queries, not good for everyday OLTP since table with columnstore index can't be updated.
For a query that already uses most of the columns in a table, performance gaining can be limited. in this case, columnstore index's management overhead (recombining the rows etc) can be worse than the benefits it brings in.
