Sorry to dig up an old post (well 10 days old) but I'm bored and thought the thread needs answering...
Let's say you have a file system with 10KB clusters. If you were to make a Word document which ended up being 33KB, it would actually take up 40KB (which in this case is 4 clusters) because Windows will only deal in 10KB jumps.
33KB divided by 10KB = 3 Clusters with 3KB left over
The 3KB needs to be homed and as Windows will only work in 10KB clusters, we would have to use 10KB wasting 7 of them! Therefore it is much more efficient to use smaller clusters.
The down point is that with a smaller cluster size, you are going to end up with loads more clusters on the drive. This means a lot more time is needed to find the cluster you want when you search for it... It is a negotiation process based on speed and efficiency.
If you right click a file and select properties, you will see a 'Size' value which is the number of bytes the file data is and a 'Size on Disk' which is the number of bytes the file has taken up due to cluster allocations.
**and on a side note, you will might see in films people talking about ghost processes that have made themselves. This is supposidly because the excess bytes from wasted cluster space groups together and creates a command by accident**