Depends on efficiency of compression I guess. If X bytes of data takes N time to transmit, and each slice of N takes Y CPU cycles during transmission, how many Y must your compression algorithm use, and how low must it lower N, in order to be more efficient from a CPU utilisation perspective? assumedly there's an inflection point from CPU use perspective, maybe a point that's impractical to achieve? I'm just vibe-thinking-out-loud.