Well, it _is_ embedded platform - which means it _needs_ preemption. It's
got nothing to do with audio signals, 100% about data latency. Maybe you
could make it sequence and synthesize lighter samples and instruments, but
looking at the spec sheet I don't think we can expect that much.