Since they run on different boards, technically you'd set up two or three nodes and run them in a parallel supercomputing setup.
Would be expensive as hell, I'd reckon.
I was thinking about this as well. It would make sense to cluster these with a front-end node. The improvements to the scheduler, among other things over the past few years, would indicate a simpler way to build the equivalent of a VAXCluster.
I was wondering if the idea would be to daisy chain these together via TB 2 and see one workstation with... say ...36 cores or whatever.
I'm trying to determine if that would be a possibility. In essence, is the Mac Pro a Thunderbolt 2 "processing" device? Or could it be made to be one?
If it were...
Man...
put me down for a couple of them. I mean you could potentially get a 48 core "workstation" in 1/2 the volume of the current Mac Pro.