This task ended much earlier than expected (below 60% of status, just a minute after restarting Rosetta on BOINC after calculating a QMC task.

The one thing that worries me is, that as far as I can remember, the task was on a Model 2. However I see only one decoy reported.

I don't know if it's a normal behaviour of these WUs - however I have crunched a couple of them afterwards and they took more time.
This machine is certainly not overclocked and it's stable - but I wouldn't want to put some miscalculations into the database...

I do realize that the team has much more struggles nowadays but maybe this highlight helps somewhat.

This task ended much earlier than expected (below 60% of status, just a minute after restarting Rosetta on BOINC after calculating a QMC task.

The one thing that worries me is, that as far as I can remember, the task was on a Model 2. However I see only one decoy reported.

I don't know if it's a normal behaviour of these WUs - however I have crunched a couple of them afterwards and they took more time.
This machine is certainly not overclocked and it's stable - but I wouldn't want to put some miscalculations into the database...

I do realize that the team has much more struggles nowadays but maybe this highlight helps somewhat.

Looks typical for times when the first decoy in a workunit takes significantly longer than expected, so the software decides that it cannot complete a second decoy taking that long before running out of time.

I've been having problems with abinitio_norelax_homfrag workunits too, but a different problem. They arrive with significantly more runtime expected than I requested (such as 24 hours instead of 14 hours) As they run, though, the total runtime keeps dropping toward the 14 hours I requested. They finally complete, apparantly successfully, but receive a much worse ratio of granted credit to claimed credit than normal for workunits that complete that many decoys.

https://boinc.bakerlab.org/rosetta/workunit.php?wuid=199364979

https://boinc.bakerlab.org/rosetta/workunit.php?wuid=199207567

O.K... if that is the point, I hope the work left gets resent...

So far I have thought that a given WU will return exactly the same results regardless of a computer used to perform calculations (alright, I've learned the processors may differ in some calculations but AFAIR it can be smoothed out thanks to some dedicated libraries).

I know you put Robert a lot of effort into Rosetta and I have seen here a few dedicated guys - so does anyone know if it is reasonable to increase default run times in Rosetta, just to get more decoys?

Of course, there is another problem with long running models though...

Anyway, thanks for your help and it's a pity with these tasks you have shown.

O.K... if that is the point, I hope the work left gets resent...

So far I have thought that a given WU will return exactly the same results regardless of a computer used to perform calculations (alright, I've learned the processors may differ in some calculations but AFAIR it can be smoothed out thanks to some dedicated libraries).

I know you put Robert a lot of effort into Rosetta and I have seen here a few dedicated guys - so does anyone know if it is reasonable to increase default run times in Rosetta, just to get more decoys?

Of course, there is another problem with long running models though...

Anyway, thanks for your help and it's a pity with these tasks you have shown.

Increasing the default run times will almost always get more decoys per workunit. I'd consider it useful if you are having problems getting workunits as soon as you need them; instead it gets you more decoys, and therefore more credit, per workunit.

I'm seldom having problems that are treated as errors in the workunits I get, so I now request 14 hour workunits. It took some effort to find the best combination of options to make this happen, though - leave in memory, an increased upper limit for the disk space BOINC can use, and an increased upper limit on the swap space BOINC can use. Also, I let BOINC run for 24 hours most days, so I'm less likely to need to restart from a checkpoint.

I feel that 14 hour workunits is a reasonable balance between Rosetta@home's recently announced desire to send out a lower total number of workunits, and my desire to see my computer complete at least one Rosetta@home workunit a day.

Some of the more recent BOINC project software combines a number of decoys into one workunit, and only tries to complete as many of them as many as will fit into the desired time. Some of the older BOINC project software tries to make all completed workunits for a given workunit come back identical, though. That means that the validators for the newer BOINC project software has a harder time of deciding how much credit to give to each completed workunit, though.