Fine-grained row-level security allows many different and useful security policies to limit users access to other users data in a centralized and maintainable fashion.

Provides simultaneous commits on master and slaves and includes the ability to implement on a session or transaction basis for tremendous flexibility.

Improves performance on read-only workloads for machines with more CPUs and improves write scalability for machines whose CPUs have large core counts.

Provides high performance for selected tables in-memory that don't require durability such as session information or other cacheable data.

Examines SQL statements and makes recommendations for new indexes on tables to improve SQL response time. Works for single and composite indexes.

Now you can begin integrating an open source based solution into your M$ infrastructure and save costs! And replication is now bi-directional.

CREATE TABLE… PARTITION BY…, ALTER TABLE… ADD PARTITION/SUBPARTITION, DROP, SPLIT AND TRUNCATE are supported along with a number of dictionary views.

Allows you to force the optimizer to use your knowledge of the database over its evaluation when it comes to joining tables in sophisticated queries.

Collation is no longer fixed at database creation time. Collation can now be set per column, domain, index, or expression, via the SQL standard COLLATE clause.