Its in the way they are built. Factory cards are not designed to handle the huge data load we put on them. they are designed to play one wave or MP3 at a time to a movie or game. They can handle that quite well.
We are playing back multiple waves at a time and forcing the card to convert several midi tracks into sound and do it all in real time as well as process the various FX in the audio tracks, again in real time. A factory card simply chokes or delivers the synth sound as fast as it can but for us, that sound arrives at the speakers much to late to be useful.
About the most delay we can deal with is in the 10ms range..... start getting longer than that and musicians can hear the delay. Most good cards that we all use easily deliver 8ms latency or less.
Add enough synths, FX and tracks and you can get the good cards to display their weaker side too. Everything has a limit.