According to the records, my computers of various speeds used to create anywhere from 10 to 60 decoys for each 4 hour run.

However, all of my recent work units, on all machines, only result in 1 or 2 decoys.

They're validated but that's a big drop. Is that supposed to be like that?

---

yup

I went to look at your results, but stopped when I couldn't decide on which puter to pick. Wanna help me out a little??

I went to look at your results, but stopped when I couldn't decide on which puter to pick. Wanna help me out a little??

Any of them :)

Matt what follows are all the wus done by my AMD64 X2 4800 Toledo, since the 11th of October. If you look, most have different models in the same run time, some are more, some are less. Recently I've see low ones too, but for me 8-20 is low and probably corresponds to 1-2 for slower machines. Do the WU names match up? The list is sorted by WU name

by recent you mean TODAY? My today WUs won't be in my data until I collect them tomorrow. I'll have to get back with you.

I don't know about my AMD64 X2 4800 because I bought it only a few hours ago.

But yes, today. Pick a random computer, like this one:

https://boinc.bakerlab.org/rosetta/results.php?hostid=201194

Units at the bottom of the first page have decoy counts in the double digits. The newest ones, however, are only 1 or 2 per unit.

Maybe new units are more complex and take much longer per decoy. This must be true since most of my computers have suddenly decreased like that.

I don't know about my AMD64 X2 4800 because I bought it only a few hours ago.

But yes, today. Pick a random computer, like this one:

https://boinc.bakerlab.org/rosetta/results.php?hostid=201194

Units at the bottom of the first page have decoy counts in the double digits. The newest ones, however, are only 1 or 2 per unit.

Maybe new units are more complex and take much longer per decoy. This must be true since most of my computers have suddenly decreased like that.

I know my AMD64 2800 got one in ralph that ran like 14K seconds (10,600 pref) and only did ONE model.

I see tomorrow.

I looked at ralph and yes the FRA 2rio and IDL_RLX seem like ONE model wus.

I think I'll go back to my first answer to you, and that's

YUP

Hi all:

mmciastro, that list is pretty useful! Before last week, we were mainly sending out workunits that indeed produced >10 decoys per workunit. Since the update, a number of workunits are being sent out that are more computationally intensive and take advantage of new modes available in teh updated app. Some test the new "docking" mode, some are full-atom refinements of big proteins (like the FRA 2rio WU), and some are full-atom refinements with a new protocol that will allow us to be more flexible with the backbone and will likely be better optimized by the next update (the 1hz6A_BOINC_NATIVEJUMPS WU). So, don't worry! I think the credit per decoy should be much higher for these more computationally intensive jobs. We really appreciate being able to crunch these on Rosetta@home since our local clusters are just not up to the task.

I looked at ralph and yes the FRA 2rio and IDL_RLX seem like ONE model wus.

I think I'll go back to my first answer to you, and that's

YUP

---

As promised, Matt, I've taken that look, and I too have them. I've charted them. Included a couple new columns; Seconds/decoy, and Granted credit/decoy

As Rhiju stated, the granted credit/decoy/time is in line with the rest of rosettas' wus. Pay special attention to the FRA 2rio wus at the bottom. This is in Result ID order (or by time sent)