As far as memory is concerned, you need to be alert for three potential problems: not enough memory, RAM failure, or problems interrelating with paging functions. You can use the various memory related counters within Performance Monitor to diagnose memory problems. A system that needs more RAM will display a high page/sec counter (memory object). Unfortunately, this particular counter could also indicate that the hard disk is running a bit slow for the system. As with nearly any counter, knowing baseline information can significantly help you here, along with a thorough understanding about how the devices interact.

If you’re looking to determine if more RAM is necessary, take a look at the Pool of Nonpaged bytes. Nonpaged data is data that cannot be shoveled to the pagefile on the hard disk, and is essentially immovable from RAM. If this number is rather high, then you have less space for paging information to be swapped back and forth from the hard disk, which can eventually cascade to other problems…hard disk failure, and system performance slowdown to just name two.

Outright RAM failures are more frequent than a ‘creeping’ problem that you can pre-diagnose with careful monitoring of the occasional fault. Outright RAM failures usually happen as a result of power problems - surges, fluctuations, and even outright failure. Occassionaly, a RAM card will fail for no apparent reason. In either case, outright failure will result in a failure message when the system goes through POST (Power On Self Test).

Creeping RAM failure, although less frequent, will appear with an increased Cache Faults/Sec counter. Although this is the more reliable counter available, Page Faults/Sec can also be monitored, with an eye towards potential hard disk issues as well.