If you care to manage which instance runs on which CPU and manage memory so each instance's data is on the CPU's local memory then the dual 6-cores should be faster since each instance shares system memory bandwidth with half as many others. On top of that, the 6-cores Xeons would likely be clocked higher than the single 12-cores.
The single 12-cores would make more sense in scenarios that rely more heavily on shared memory and inter-process communications.