<t>First off, if you're starting a new project, go with Entity Framework ("EF") - it now generates much better SQL (more like Linq to SQL does) and is easier to maintain and more powerful than Linq to SQL ("L2S"). As of the release of .NET 4.0, I consider Linq to SQL to be an obsolete technology. MS has been very open about not continuing L2S development further.<br/>
<br/>
1) Performance<br/>
<br/>
This is tricky to answer. For most single-entity operations (CRUD) you will find just about equivalent performance with all three technologies. You do have to know how EF and Linq to SQL work in order to use them to their fullest. For high-volume operations like polling queries, you may want to have EF/L2S "compile" your entity query such that the framework doesn't have to constantly regenerate the SQL, or you can run into scalability issues. (see edits)<br/>
<br/>
For bulk updates where you're updating massive amounts of data, raw SQL or a stored procedure will always perform better than an ORM solution because you don't have to marshal the data over the wire to the ORM to perform updates. <br/>
<br/>
2) Speed of Development<br/>
<br/>
In most scenarios, EF will blow away naked SQL/stored procs when it comes to speed of development. The EF designer can update your model from your database as it changes (upon request), so you don't run into synchronization issues between your object code and your database code. The only time I would not consider using an ORM is when you're doing a reporting/dashboard type application where you aren't doing any updating, or when you're creating an application just to do raw data maintenance operations on a database. <br/>
<br/>
3) Neat/Maintainable code<br/>
<br/>
Hands down, EF beats SQL/sprocs. Because your relationships are modeled, joins in your code are relatively infrequent. The relationships of the entities are almost self-evident to the reader for most queries. Nothing is worse than having to go from tier to tier <br/>
<br/>
(Réponse tronquée)</t>
