Dear skgiven,
a 64bit version should actually not be faster than the 32bit version. The only difference is that the former can address more/bigger RAM, but at a level which we never reach in CEP2 (or any other BOINC project) anyways. So it makes sense that it is not a high priority for the IBM WCG team.
Best wishes
Your Harvard CEP team

https://secure.worldcommunitygrid.org/about_us/viewNewsArticle.do?articleId=157

Maybe this can answer some questions

So you are suggesting CEP2 is a floating point intensive project and would not benefit in terms of performance because 64bit processors use 32bit floating point registers.
My understanding of 64bit CPU’s is that they run x86 tasks in emulation mode. So I thought by simply running x64 apps there would be some benefit from not using emulation, and possibly some benefit from using 64bit libraries. I think that even 8% is worth having.
Perhaps there would be little or no performance difference and it’s presently low priority, but looking into the near future, “A Bulldozer core module has two 128-bit floating point units, which can do two 64-bit double-precision operations per clock or four 32-bit single precision operations. Either core in the module can see if the other core is doing floating point math. If it is not, then one core can take all 256 bits and do four double-precision or eight single-precision ops in a single clock cycle. AMD has started calling this an AVX mode”.
WRT the amount of RAM used, 3.2/3.5GB of RAM can be easily reached when running CEP2; you just need a common operating system such as Vista or Win7 that uses over 1GB RAM within seconds of logging on, a couple of desktop applications (Outlook, Firefox) running and you’re up to 2GB. Use 4 threads and you are already pushing 1.6GB bringing you over the 3.5GB mark. 6 threads for an AMD X6 or all 8 threads of an average i7 and you are in deep trouble. Throw in a couple of GPU tasks and you are throttling the drive to the point you are wasting your time. I would not expect too many people to have a 12 threads x86 system but there might be a few.
Several non WCG-Boinc projects use enough memory to test a quad on x86.
Within a few months 16 to 32 thread systems will start to become common, so do you think a 64bit app for CEP2 would be useful, or would you be happy with the project remaining 32bit for the long term?

Lawrence,

Re the bolded bit in the quote, I'd love to have them now too, but my Linux quad is only getting C4CW 64 bit work in the mix.

skgiven,

To convert or not to convert, the answer was given in the very first reply: Order of work is as I described, unless there are those who start working on the hardest nuts with the lowest effect on production, not knowing if they'll succeed. As cleanenergy replied, he'll accept from what he knows intimately, that CEP2 is of a lower priority order in this transition, the 8% mentioned as a reference of what to expect. 8% of 11-13 years, however well argued that it would benefit, still does not warrant high placement. But why argue when WCG has made the statement as to what it is doing and committed to do all future sciences this way for sure?

--//--

edit:

A PS: We should not forget that by the mere running on 64 bit OS we have already taken a substantial part of the achievable gain as countless have observed switching from 32 to 64 bit OSses.