The authors would like to thank S. Boxer, A. Brunger, D. Ensign, H. Grubmüller, and O. Troyanskaya for many helpful discussions and the center for parallel computers at KTH and Folding@Home donors worldwide for computational resources.
It looks like they were a bit optimistic about the number of WU to generate ... I've seen some reports of people not getting work, or getting GPU2 projects.
Memory has little effect on FAH performance (see overclocking tests) and the units are designed with 512 MB as a default on most boards.
I didn't test the older card because I don't have it, and I don't have a friend who has one ... it's the same for the comparison between GDDR3 and GDDR5 memory, or 512 MB vs 1 GB.
And posting the server IP is also a good excuse to highlight the work of Stanford's partners that run their own researches on the Folding@Home platform.
In my opinion, you shouldn't see major performance differences ... as my overclocking tests shown, memory has a minor influence on the FAH GPU client performances.
The more, the merrier ... it's always good to have multiple source to compare the data (if they are comparable, because of potential methodology differences).