Rosetta Credits

[pututu]

Rosetta is pretty lousy in PPD. Over the past two weeks of crunching, I notice that there are some WUs that give higher credit averages, something like 3-5 times higher than the "normal" credit/WU/hr. Below is the stat for my rig 2683 v3 running Windows with 2 hour target CPU run time.

Thought I might share this information while the credits is still good, if you decide to crunch Rosetta.

 

[Gilthanis]

I don't think it is just the application version as some of your other Rosetta Mini v3.78 were also credited less. Are all the Rosetta v4.06 credit small while the Mini v3.78's all have another factor? I see that the v4.06 also had some higher pointers. Did you change any of the settings in the Rosetta preferences such as target run time?

Is there any difference in the batches sent out? I'm just curious because if someone were to only run the Mini v3.78's is there a way to only pull the higher crediting work? As I'm not seeing any non-manual way of filtering between the 2 CPU applications. Let alone control which will get me more points.

Edit: Do you see any differences at all on your side that will allow for manually filtering out the lesser point work?

 

[pututu]

For targetted CPU time, I set two hours. I run both Mini v3.78 and Rosetta v4.06. I took a glance at the names of the WUs but nothing stands out. Feel free to let me know if you do see something on my 2683 rig for any pattern. I'll take a look at it again this weekend. Now it is a good time to run Rosetta, if you ask me.

 

[pututu]

There are many different WU types. Some do score high and some don't. And since I'm getting a bunch of different WU types every day and it keeps changing, not sure if there is any simple way to filter these out. For e.g. any WU with the name "bnd_avian" does not score well but WU with "ab_stage_h002___robetta" appears to score well.

Just saw Skillz started one of this 4p on Rosetta today.

 

[Deleted member 88227]

I'm on it. Its sharing resources with WEP though. Gotta keep EVGA at bay on that project.

 

[pututu]

Ok. It seems that the WU task name with "*bnd_avian*" does not give good PPD on my rigs. Though I should not be doing this, I aborted all the tasks with "bnd_avian". Most of the minirosetta yields better PPD. I didn't keep a good stats but just eyeballing the numbers.

So far my two rigs daily RAC performs better than the other 90 rigs above me with more cores/threads except for one possibly i7 haswell rig with unknown CPU model name.