I was just thinking. With AMD FX architecture; you ahve the 6 core version. 3 modules, 6 integer units. Each "pair" combines to a 256bit bus during certain instructions. Seems like it be vastly superior to run 4 cores on 2 matching modules, than it would be to run single core off 3 modules.
Dual and single core instruction rate should be superior, at least in my head. Can anyone comment on this?