Nick,

Perhaps I misread your post, I thought you meant to make 450,000 changes took 81 seconds total.

If you are instead saying that to make the 450,001th change took 81 seconds then it sounds like you need to use an algorithm that is not exponential in execution time. If you've ever taken a Data Structures and Algorithms class you know that O(n²) algorithms are bad for any large value of n.

>Not sure where you got that throughput from. Guess I didn't explain too clearly, I start off by creating a 60000 record cursor, I then start making random changes, after every 10000 changes I run a sys(1104), in the begining that only takes a second or so but this increases each time and after 450,000 random changes (45 Sys(1104)'s) it had increased to 81 seconds.
>
>Without the sys(1004) the continuous random changes almost grind to a halt way before it gets to the 450,000 changes level.