FreeBSD + Cluster + Folding@Home

Hello all! Recently I've decided to create a little project. I was thinking of doing a great deal of contributing to Stanfords Folding@Home project. How was I going to help? After recently falling in love with FreeBSD and reading up on as well as watching videos of the fine work FreeBSD has done in clusters, was to create a cluster of PC's to fold all together. I'm currently looking into some donations from generous persons, who are willing to give up old Dell Poweredge servers to help start this project.

If anyone wanted to help out in this project. Helping me setup a cluster or contribute anything at all to this project, help is appreciated. Let me know what you think.
You should start out with a single machine since these massively distributed projects do not actually do anything in parallel using MPI, which is what you'd need if you seriously wanted to use a distributed memory cluster.

I also looked into doing this at some point. As of now I have 7 Penguin Computing Relion 130's and one Penguin Computing Relion 240 set up and in a datacenter. I have not yet set up the clustering. But with each machine having dual 2.8Ghz Xeon HTT processors and between 3GB and 6GB of ram, I'm assuming I could get decent performance for a folding@home cluster. Last I checked the SMP performance for folding@home wasn't spectacular unless you had some massive supercomputer, but I think even a cluster of this size would be beneficial.

On the topic of SMP / Cluster performance, what do you think would yeild better results with folding@home; Using Hyperthreading on the processors (so dmesg would show 4 processors per machine) or turning off hyperthreading (only two)?

Anyway, hardware wise I think I have a good start...
I thought the same thing at first... I know the PS3 has amazing statistics due to the cell processor. But after doing some research through folding@home and such, the PS3 is limited. It may have a very fast graphics processor and a high gflop count, but it can't do some of the tests an SMP cluster could do. The PS3 looks at the data from different angles. So the cluster would work on a different aspect of the folding. According to folding@home, they said both are needed and even are encouraging people to continue to use clusters so all research aspects are covered.

The cluster can work on much larger projects and data where as the PS3 is extremely efficient at a certain type of test. So there still is a use! :)