That's an "it depends" answer. As Balaji notes it depends on the QoS policy, but what a given QoS policy might be also depends on the platform's QoS feature support.
On software based routers, there's generally an implicit policer which drops all LLQ traffic beyond the defined bandwidth limit.
On many switches, PQ can use all the bandwidth (i.e. there's no implied sub limit), and will then queue traffic, only dropping such traffic when the PQ queue overflows.
Again, the forgoing is general default behaviors of QoS policies on those platforms, but how you define the policies can sometimes change the foregoing general/default behaviors.
BTW, Balaji also mentions ". . . any Voice can not be Queed since its UDP, not like TCP.", but, actually, VoIP and other UDP traffic be queued, and (very, very) often is, however, depending on the nature of the application, queuing, especially "excessive" queuing, such traffic might be more detrimental to it than other application traffic using a protocol like TCP.
