Not very scientific, but I found that the first time I ran, the result from the @tableparams was considerably faster when selecting approx 17,000 rows from 682,000, where there is no useful index.

(17234 row(s) affected)

SQL Server Execution Times:

CPU time = 2250 ms, elapsed time = 8095 ms.

-- this one was parameterised

(17234 row(s) affected)

SQL Server Execution Times:

CPU time = 515 ms, elapsed time = 1784 ms.

-- this using the table variable

Running exactly the same thing a second time the difference was greatly reduced.

(17234 row(s) affected)

SQL Server Execution Times:

CPU time = 343 ms, elapsed time = 958 ms.

(17234 row(s) affected)

SQL Server Execution Times:

CPU time = 314 ms, elapsed time = 264 ms.

Using a different table, the paraemterised query was quicker on the second run.