Long time no see,
I was thinking about this recently, because I finally left academia and was able to afford to build another computer for myself. That brought back the memories of this old thread. Since I know this is one of the few places where you might find someone who understands these things, I thought I'd let everyone know what happened.
That story came to end recently, as it happens. I understand the machines are still doing scientific work, just productive work. Not what they were built for, which was development of a way to do productive work.
The group actually built another machine with 144 gb of ram using similar hardware with the fastest 6 core nehalem. That was called "the Beast". The original was called BoB (bride of the beast), remade from leftover beast parts.
After years of using them and the beating of heads on a very difficult scientific problem. This paper was finally published
http://www.jamesfolsom.com/img/efm1
There was much scientific interest, but the method was not in reach of alot of scientists who couldn't lie, cheat and steal there way to complete domination of campus computer lab machines. But I exaggerate, as it was all legit, except that one time we caused the buildings file server to reboot because we had 20 simultaneous multi 100 gb file copies going. We were forgiven though. In the end our plan played out masterfully.
Because, there was another group who had been working on improving the computational technique, and optimization of the calculation. When they realized they still couldn't beat our groups technique for raw #'s of EFMs calculated, they took our technique and added it to theirs. This suited our group nicely, because we didn't have the expertise to do anything with underlying computer code.
The solution to the problem, at least for now was recently published.
http://www.jamesfolsom.com/img/efm2
Sadly, there were those who fell along the way, we toast them now. They gave great hope, but never found their way.
http://ieeexplore.ieee.org/xpl/logi...re.ieee.org/xpls/abs_all.jsp?arnumber=6008870
But anyway, it was all a fitting end for this thread too.