I have found the exact same issue. Does anyone have a solution yet? On production I am doing a physical_only so the time there is not a problem but I am restoring the backup to another server in order to do a CHECKDB to make sure everything is in order. I pinpointed in down to a single table and the DBCC CHECKTABLE is taking 11 hours on a 1.1 billion row table that has a clustered index on two columns (int, bigint) and a non-clustered index on a single int column. There are no relationships.

Literally 100% latch time is DBCC_OBJECT_METADATA and it is always waiting on CXPACKET. There is a lot more CPU than I/O.

This is running on WS2012 std, SQL2012 SP2 (11.0.5569.0). 4 CPU, 12G.