>>Always depends on the data at hand, but most uniform best times should be reached by piping the list of PK's into a temp table with index and inner joining to base select table, esp on longer PK-lists.
>
>Can't do this one but will do the benchmarks. This is a contained option and I doubt that visual notification of speed would be a factor.

I expect SQL server to execute it this way - at least for longer list len, perhaps switching strategies depending in the list given in the SQL.
Update: on reading Naomi's link I may have overestimated SQL optimizer capabilities, but basic gut instinct was correct ;-)