I'm basically trying to get you lot to talk me out of it
We are not silly enough to do that
Many of the DC projects are now using GPU, rather than CPU. In most cases the GPUs, if the project is correctly optimised, achieve more for the same output. These though tend to be of the number crunching and medical research type projects of which folding at home is probably the best known.
Folding at home is one of the few projects that can make use of the multicored / multithreading capabilities of modern CPUs. They refer to such work units as "SMP". Having two CPUs on a board will show some results. For a while folding did have special work units available, known as BigADV, which could only be run on specialist machines that operated 24/7 with 2 and sometimes 4 CPUs. Today those still exist but you need at least 24 cores. Don't buy a machine for such work now as this work type will cease by the end of January 2015.
But multicored, multi processor rigs are of value to many, many projects that use the Boinc platform. At present Boinc cpu work is single threaded. But the boinc manager platform will run multiple threads simultaneously. Thus if you have 8 cpu cores (say an intel i7 cpu that has 4 processors allowing each to hyperthread), you get to do up to 8 different work units at a time. That then also means that with boinc you can join more than one project at a time, and indeed even have the ability to change resources, so that perhaps one project takes 75% of resource and another the remaining 25%. There are many many boinc projects, skynet is one, and many others are astronomical, some medical or whatever. It's worth noting that some of us who have the 2p (2 cpu processor) and even 4p rigs available originally intended for folding are running boinc for the skynet, so in my case I have one machine, offering 64 simultaneous skynet processing threads. But you may have to use linux as most home windows versions cannot recognise all these cpu cores.
In theory the ratio of points awarded for a wu in boinc should correspond to different boinc projects. They won't correspond to other projects such as folding. Where projects offer cpu and gpu crunching there should be a relationship so you can for example judge by folding at home whether cpu or cpu is more efficient on a hardware cost v energy use basis.
Only you could decide if a specialist server type 2p (2 processor) rig is suited for the type of work you want to do, or if you would be better with a single cpu that can accommodate multiple gpu cards.