Today at TechEd 2014, many people had questions about the in-memory database features in SQL Server 2014. A common question is how an in-memory database is different from having a database on a SQL Server with an amount of ram far greater than the size of the database. In-memory or memory optimized tables have different data structures and are accessed differently using a latch free and lock free approach that greatly improves performance. This provides part of the performance improvement.
The rest of the performance improvement comes from natively compiled stored procedures that can only access memory optimized tables. Conventional stored procedures can access either conventional or memory optimized tables. While it is true that conventional stored procedures are compiled, they do not compile all of the way down to native machine code. Natively compiled stored procedures are faster than conventional stored procedures.
For obtaining an in-depth understanding of in-memory database features, I recommend that you read the excellent whitepaper written by my friend Kalen Delaney which can be downloaded from Microsoft here.