As you probably know, data is never really "deleted" from a file system -- the file's name is removed and the data is added to a list of "free space."
On a first gen SSD, these "free space" blocks become interleaved with "actually used in files" blocks. Unfortunately, in order to reclaim space on part of an SSD block, you have to clear and re-write the entire block. Magnetic hard drives, for which all existing drivers and file systems were designed, do not have this limitation. The end result was write pauses as drives aged, and more of their free space was interleaved with real data.
Second gen SSDs got around this by using idle time garbage collection, wherein drives that are not busy go through written blocks looking for deleted data. Each block must be checked individually, there's no "hit list," and so idle GC is an inefficient way to clear up blocks. For laptop users who frequently close their lids, there isn't as much GC as there is for a server user and thus idle GC is less useful.
TRIM takes this to the next level, proactively identifying blocks with free space on delete and making garbage collection much more efficient.
TRIM will play nicely on any "stress test" that has a server-like load -- where data is in constant motion without extended (minutes or longer) periods of rest where GC can run a full cycle. In "real world" desktop/laptop tests, TRIM is not magic or even a necessity; many users will not notice much difference beyond idle GC. Users without idle GC (such as my first gen SSD at work) will notice a HUGE difference between garbage collection and none, regardless of the collection technique (idle or TRIM). Drives using alternative storage techniques, such as the SandForce controller, should not notice any difference with TRIM since storage is allocated much differently.