Avoiding the memory capacity crunch