Do you have conlog loaded? If so, it may contain messages from the console that you can read and figure out what's causing the problem. The RAM sounds to me like enough for the size of your volumes, but are you using NSS volumes or traditional volumes? On the hardware side, check to make sure that you have the latest drivers from the vendor. I know HP, Dell, and IBM all post new hardware drivers from time to time to address these types of issues. It would be great to find out what causes the utilization to jump right before a crash. I'm running service pack 8 here, with no problems. I haven't run service pack 5, so I can't comment on that.
Iolair MacWalter