All it means is that higher priority tasks will get either a larger timeslice of CPU activity, or (more properly) get more frequent timeslices. For example this might mean that for each time a low priority job gets queued to run on your CPU a medium priority job will get run twice, and a high priority job three times.

Your CPU should *never* be idle if there is work to be done. I'll stop now, I'm beginning to sound like my mother