link https://boinc.bakerlab.org/rosetta/result.php?resultid=189240710


Is it a bug?

Please reference the following threads:

granted credit < 10% of claimed credit
Low credit from mini version vs prior

---

Rosetta Moderator: Mod.Sense

No, this is not a bug, nor a specific miniRosetta issue.

Invoking all possible caveats I shall try to give a possible explanation:

You will notice that the wu (1 model) in question ran quite long compared to your default runtime and compared to other models in wus of the same category. This may be a clue.

Credits are based on averages. And while it is easy to assume that all models that might be generated within a certain wu need approximately the same time to complete, this is not always the case. Effective CPU runtime may differ with several hundred percent between models within the same wu. If a special type of wu consists of lots of short models and a few longer ones this will of course affect the average credits per model. The losers are those completing the small number of large models.

In all fairness, one may also get a surprise the other way when a series of long models suddenly will be interrupted by a quick decoy earning lots of "undeserved" credits.

from observation, i think part of the problem is that CPUs with a smaller cache have the difference between claimed and granted exaggerated because minirosetta seems to be more cache-dependent.

Of course CPUs with a smaller cache get a higher benchmark than they should as the cache size isn't taken into account in the benchmark, but this is the same for both minirosett and old rosetta...

---