https://boinc.bakerlab.org/rosetta/workunit.php?wuid=436452069

hmmm... very low credit... why ?

https://boinc.bakerlab.org/rosetta/workunit.php?wuid=436452069

hmmm... very low credit... why ?

Rosetta@home uses a different credit granting scheme than most other projects. You get credit based on results (number of decoys produced), rather than effort (total time spent crunching). See https://boinc.bakerlab.org/forum_thread.php?id=2194

That particular job looks like it's granting ~9 credits per decoy, so it looks like you've only returned one decoy, despite running 3.25 hours. This is consistent with the last lines of the output, which claims "This process generated 1 decoys from 1 attempts"

I don't know, however, why there are two such lines, the earlier of which is claiming 8 decoys produced. If they were combined, you should have gotten ~80 credits, much closer to your (time-based) 71 claimed credits.

My only thought is that for some reason BOINC decided to dump the results and restart the process. One thing to do is check that your "Keep applications in memory" to "YES" (https://boinc.bakerlab.org/rosetta/forum_thread.php?id=669#10374). I don't know if that will fix this issue, though, as all R@h runs should have decoy-level check-pointing (even if you restart, all the decoys you've already produced should accumulate). My rough guess on very little information is this is likely a BOINC-level issue rather than a Rosetta@home-level one. Something likely happened to BOINC to cause it to toss the results and restart. Unfortunately, I don't know what that might be.

Hmm OK thx for answer,

i uncheck keep work in RAM because i had problems with another project...
But i'll try again. This was the only unit with such a low credit... all other units are OK...

---

Speaking of low credit, I just finished one that says the claimed credit is about 81, but only received 20; ouch.

https://boinc.bakerlab.org/rosetta/result.php?resultid=488130789

All that time spent and someone else could've done it faster than this old machine...

Your task hit the limits of the watch dog's patience. It spent more than 7 hours trying to complete a single model. This is abnormal. So the watchdog did his job to clear the task from your machine so another task could be started.

I say that because of this line in the output:
BOINC:: CPU time: 25417.4s, 14400s + 10800s[2012- 3- 2 3:31: 3:] :: BOINC


The 14400s is your 4 hour runtime preference, plus 10800s is the watchdog's 3 hours of patience. Also note after the line mentioned it indicates no output file was found (because no models were completed to write to it I suspect).

---

Rosetta Moderator: Mod.Sense

It spent more than 7 hours trying to complete a single model. This is abnormal.

One potential reason for this is if the execution gets frequently interrupted. Although runs are checkpointed, if execution is interrupted more frequently than it is checkpointed you waste a lot of time recomputing things that have already been computed.

That's why we highly recommend setting "Keep applications in memory" to "YES", which should preserve application state, even between checkpoints. (Note that your OS will swap R@h out to disk when running, so setting it to "yes" shouldn't eat up RAM, or substantially affect other application performance.)

Rocco, I believe you may be mistaken on your logic here. I believe if a task is removed from memory, before reaching a checkpoint, that when it restarts, all knowledge of the CPU time done after any checkpoint is lost, and therefore the watchdog would not detect the CPU time threshold being reached.

However, if this kept up, you'd hit the other limit on how many times a task is allowed to start with no progress. The item described is showing it actually was running for 7 hours and either checkpointed 7 hours of compute time, or ran that long.

---

Rosetta Moderator: Mod.Sense

Moderating dears, in my case I have activated to leave the work in memory and alone work for R@H. I have observed that enough works that exceed the 4
hours predetermined in my computers and that give error of computation, or they have no checkpoints or the checkpoint remains in minutes or between 1 ó 2 hours, that, on having seen the result, seems that the computer only has worked the time of this checkpoint, but the reality is that I have lost in every work 7 or 8 hours that do not remain reflected in the statistics.
This spends topcoat in CASP and also in some others but punctual.
I hope that it should serve them of something my information.
Greetings,
Salvador Cutchet

I am also seeing very low credit for the recent work I have done. All jobs completed after March 16 produce 20-25% of the credit granted for jobs completed prior to that date, while run times are in the same approximate area.

A post by ArcSedna in https://boinc.bakerlab.org/rosetta/forum_thread.php?id=5928 seems to indicate a problem in the MacOS X client.

I have the same problem with the low credit. Here's an example of one of my tasks:

https://boinc.bakerlab.org/rosetta/result.php?resultid=494312724

I've been getting low average on points since Feb. ... and since March 13, it's been way down ... from averaging 850+ to just around 350. The Boinc client stats line practically showing a 45 degree line downward.

nothing has been changed on my computer ... anything particular i can check or update?

---

I have the same problem with the low credit. ... anything particular i can check or update?

There is a known issue with the speed of computation of Mac (and Darwin) computers with version 3.24. We're aware of the issue and looking into ways of remedying it. (See also the thread that m2a2b2 linked to.)

I've also low credit on windows system. Around 40 credit for 5h wu's.
Earlier i get nearly 100 credit for those.

---

I've also low credit on windows system. Around 40 credit for 5h wu's.
Earlier i get nearly 100 credit for those.

At present, it looks like the credit you are being granted is inline with the claim (which is based on your CPU performance). Was this the case previously? You might check the machine to see if other tasks are using CPU time. BOINC tasks run at low priority, so they yield if other work is active on the machine.

---

Rosetta Moderator: Mod.Sense

Hi!

I receive low credits (about 20% of earlier) on my Macs after approx. March 18. My Windows machines are not affected. Is there any progress in figuring out why that is?

Torsten

Hi.

I,ve the same problem with my Mac,s... :(

Any news about this???.

Thanks.

---

Mac's are presently seeing their own low credit issue as described and discussed in the 3.24 thread. It sounds like they are now on to a correction for that issue.

---

Rosetta Moderator: Mod.Sense

Hi.

Mac OSX low credit problems seams has been solved with version 3.30.

Thank you.

---

https://boinc.bakerlab.org/rosetta/workunit.php?wuid=462919979

Can someone explain th low credit granted for this WU

Thanks

The second message in this thread also explains the low credit in your case. You bumped into a long running model and were only granted the credits for the one decoy that was produced. It's the "luck" of the draw, so to speak.

Thankyou for the reply :)

As for luck of the draw.. Won't happen again as I will be ABORTING any I manage to catch that look as if they are going to run this Lenth again....


Cheers...

...I will be ABORTING...

Thus assuring that you receive zero credit for the effort.

---

Rosetta Moderator: Mod.Sense