The bigger the file size, the larger the chance that it has been generated in a pseudorandom way, meaning it is in theory compressible. On the flipside, in a larger file finding patterns from that pseudorandomity could take longer.