Table partitions have new and improved functionality in SQL Server 2005, and it is now actually very easy to define them.
A simple take is that partitions enable you spread table data over multiple locations.
In one scenario, your application may have 10 years worth of data in a ‘Sales’ table. If you create a partition for each year and place them each on a separate disk, you will in theory have fewer reads/writes on each disk resulting in more efficient queries.
In addition, many bulk operations can be further optimised by dropping indexes and other constraints on individual table partitions.
It should be noted that your database will have to be very large to make this a worthwhile exercise. Even then, other factors might produce little or even worse performance benefit.
If you are interested, take a look at the following links to get started:
Partitioned Tables and Indexes in SQL Server 2005
How to Implement an Automatic Sliding Window in a Partitioned Table on SQL Server 2005
Top 10 SQL Server 2005 Performance Issues for Data Warehouse and Reporting Applications
Partition Elimination in SQL Server 2005
And for some more detail: