Is there anything wrong with creating a large amount of secondary buffers, as long as only a few are being mixed at any given time? For instance, when I load a level, I can either: 1. Create 50 (or whatever) secondary buffers, and play them as needed. or 2. Create 50 images of wave data in memory and copy one of those into a buffer as needed and play that buffer. The first approach seems like it would be more processor efficient, though I''m not sure if there would be a memory loss. However, will DS freak out if I create that many buffers? I can easily put in code to make sure no more than 10 or 20 are being mixed at once, but I don''t want this to crash on systems with older ISA sound cards.