I havent been the contrarian lately, so....

I tend to use NOLOCK frequently, but I don't write financial apps and I generally understand the data movement that might occur. I think a comparison of READPAST vs NOLOCK would be interesting indeed.

I think the usefulness varies and I would have guessed at maybe a 5% savings. Maybe its the same thing said a different way, but I use NOLOCK because I know that I dont need a lock and dont want to risk blocking writers, and I dont want to be blocked - which I guess translates to a performance boost because I shouldn't be waiting or causing waits, but I don't use it to get a x% decrease in resource usage (though maybe I should?).

Couple thoughts from that. One, can we craft a test that would satisfy us all, and then maybe see if the author or someone else is game to dig in deeper? Two, maybe we can get Steve to lean on MS to see if they can explain a bit about the performance implications.

READPAST has me thinking - should I be using it more?