TBH in most cases linq or SQL aren't exactly the issue. Your performance will be related to how much data you are inserting, the amount of data currently in your table and the indexes you are maintaining.
Secondly whether you need to do cross checking and/or integrity checks across multiple columns on your data. I have had situations where adding an index and rebuilding a table has taken insert time down from minutes to milliseconds, just due to bad fragmentation and lack of an algorithm.
Linq is an effective way to generate SQL for insertion and modification logic. However you will always end up with the pattern:
- Fetch data from database
- Modify data using Linq
- Submit changes to database.
If you have any logic you can exploit in your insertions, you may be able to use set logic to do updates in SQL. E.g. Update Customers Set KeyCustomer = 1 where Sales > 1000000. The SQL Server will process a command like this 1000s of times faster than you could ever do it with your ORM. However as @gbn has already correctly pointed out, unless you have a team full of strong SQL coders, maintenance will often trump any perf gain in the short term.
If you have to insert a significant number of records, then you should really be looking at batch loading and/or ETL via SSIS. These APIs will use smarter algorithms and perform any constraint checks in batches rather than per insert which will give you excellent performance increases. But managing an SSIS package is far more work than clicking a button in an app. These are all design decisions you will need to consider when you architect your application.
