Clearly Benny has set his adjustable WU runtime to 24hrs. It's a great way to crunch more models on the same protein and minimize your network bandwidth. This is in your Rosetta preferences. Credits are issued for the time spent on the WU (as scaled by relative processor performance), so you can set a higher runtime preference, reduce your bandwidth requirements, put less hits on the project's servers, while getting just as much credit and helping the science just as much (with LESS WUs!).

Beware: The client won't use the new runtime until you update to the project, and BOINC won't figure out that the WUs will take longer to run until it has crunched a few and updated again... so don't just jump from 3hrs to 24hrs with a 3 day cache. The client will ask for work thinking it takes 3hrs to complete, and so request 24 WUs to fill 3 days... then end UP taking a day on EACH ONE. Rule of thumb. Make changes gradually if you have a longer "connect to network every... days" setting in your general preferences.

I've been getting 240-270 credits per WU running each for 24 hours, on a 2Ghz, Athlon 64 754 pin 3000+ cpu using the default Boinc client (5.2.16). So you can get close to 5xx credits per WU with something more modern than my 2 year old cpu by running each WU for 24 hours (86,400 seconds).

If you're talking about 5xx credits per WU that have run for 10,000 seconds, then someone's got really outrageous Boinc Benchmarks..

Looking at top particapants, some person has 144k average credit. That must seriously screw up the tflop measuremnent for rosetta@home.

Why? You must be referring to DDTUNG, who has somewhere in the region of 80 dual core opterons + 30 or so dual xeons@3Ghz + other assorted P4's. He is legitimately processing that amount per day..

Bok