you mean instead of 1 and 2 i would put the t-benches? but how would that fit in size? i thought they would be directly next to it.
No I think he means get another identical radiator and then split the flow right before the radiators and merge the flows right after. Anyways I'm not convinced of parallel loops. They require twice the component counts (two SLI links in your case) as well as a splitter for what appears to be minimal gains. See Martin (from martinsliquidlab.org)
post on this subject.