I am currently trying to free-up some DCMs for use in other parts of our design, so I am trying to make some MGTs share DCM outputs. Currently I have one DCM per each of the eight MGTs which uses all of our DCMs. For reference, we are on a Virtex II-pro trying to communicate at a rate of 2.5Gbs in two-byte mode. Xilinx document UG024 talks about a case where no DCMs are needed at all, but it isn't clear on when this solution is suitable (pg.46). I thoguht we may at least be able to use one DCM for the top four MGTs and another for the bottom four. Any thoughts or experiences related to this issue or clocking MGTs on a Virtex II in general would be appreciated.