I think at this stage the original question has been answered well enough, and in my opinion it does not involve an expensive server based system for a newcomer. Especially one that could be matched by a single socket system before the end of the year, or early next year. Multi-socket servers are an oddity, and for most they are really not useful. Some people argue that they are confined by space, but two small systems don't take up much more room than a multi-socket server; they tend to be Extended ATX2 FF. Their real benefit is longevity - servers are designed to run continuously, at or near flat out for years. However, even that is flawed; you could buy two good systems now and replace them after a year for less, and still do more work. Given that some even require two PSU's, andzgrid's point is quite valid, and something I normally preach when contributing to GPU projects. On that note, we are likely to see a GPU project soon, so advising to over-invest in a server now is totally folly.

So, to summarize:
Get a SB CPU, a good PSU, motherboard, and 4GB or 8GB RAM. Set it up for Linux and await a GPU project.

Mysteron347:
I will FIRST have to do some spectralAnalysis of the rock that you're refering to before I can have a definitive answer as to the type of the rock. I think they call the rock type as "reality" and we have plenty of that here on planet Earth. It is with that type of rock that makes "run THEN walk" as something that makes no sense to us humans here as "learn-to-walk-first-before-running" makes you and you kind go for an expert (in what, delusion?).. from, planet, what?
;

Hi skgiven,
I agree with your recommendation. For starting to participate to Grid-computing projects, a good (and affordable) single CPU system is fully sufficient. If, in the future, such system could be expanded with one GPU is better but not mandatory.
My indication regarding a system based on two Magny-Cours was more focused on "experienced crunchers". Participating to WCG projects does not require to have powerful systems. As soon as a member notices that S/He has fun and time and money for doing more, more powerful systems could be interesting. But again, WCG is not expecting that people invest a lot of money for supporting projects. The primary idea is to use unused CPU power, not to invest in more (and expensive) additional power.
Cheers,
Yves

Roberto,
I would suggest you don't invest in GPU's, if you intend dedicate your rig exclusively to run WCG projects.

There isn't any WCG project today in conditions to use it, and probably won't in short future. I wonder if WCG tech team post something here, that point us all to a clear and real distinct direction from that...

So, GPU investment just to dedicated this resource to WCG projects, by now, will be a money lost and an frustation. And, with the technology available today to build computers, which for sure fit in your budget, you could invest on GPU's to install in your rig in the future, if and when WCG will announce a project using GPU's powerful capabilities...

Up to you and good luck in your decisions.

roberto, Get a good efficient system now (a Sandy Bridge), with a good PSU (Corsair, or similar, with 2 GPU power connectors), and a motherboard with 2 well spaced PCIE slots.
In the near future when there is an active GPU project and when we know how different GPU's perform, then buy a good performing and afordable GPU.