CPU does not really play into a deadlock. One operation starts next to another operation - which could both be small and perhaps both pull from disk. When they start, they both lock a resource that the other will eventually need. The first operation will complete what it needs to and then move on to the second resource, but can't use it because the first operation has that locked, so it will just wait. But, now the second operation will complete with the first resource, keeping it locked and now need the resource the first operation has locked. They will both forever wait for one another, until the back end components of SQL server come along, prioritize one of the operations, and choose a victim.

Deadlocks are easy to reproduce. There are many solutions out there, but mainly optimizations are the recommended path.

http://msdn.microsoft.com/en-us/library/ms178104(v=sql.90).aspx