I've seen Read Committed run faster than Read Uncommitted in a couple of circumstances.
One is if the table is heavy on data-modification, and Read Committed Snapshot Isolation has been enabled. In that case, Read Committed will read from snapshots, which can be faster, especially if tempdb is on different drive media than the main DB (since tempdb is where the snapshots are held for RCSI). Can be faster, can be slower, depends on a lot of variables, but definitely can be faster in some situations.
The other is heavy updates causing lots of page splits. The Read Committed version will lock the pages it needs, preventing concurrent splits. This can make a big difference in I/O time on the query, especially if it's over a range of data that parallels the leading edge of one of the indexes. The RU query will have to deal with pages being split while it's reading, potentially reading the same data two or more times, which can result in the total query being slower and returning incorrect results, both at the same time.
Either one is kind of rare, but they can happen.
- Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
Property of The Thread
"Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon