Hey i have a big problem. If i have for example a folder with BOINC which is by 85% and i press stop for a break that BOINC is stopping working and I close it, it is the next time not anymore by 85&.

For example yesterday there was one which was by 97% and I make a break and close BOINC. So now I started it again and its only by 77% anymore. Why is it so? Thats really bull****.

can you edit the last word in your post pls.

Rosetta can only benchmark at certain points in its calculation, so if it is shut down it has to restart from its last checkpoint. Its unfortunate, but is a result of the complexity of the models being produced unfortunately.

Danny

---

Are there special points when are the checkpoints? For example at 20% is one at 40% and so on or are they allways othere?

Thank you for revising your post.

Yes, dc means "checkpoint" not "benchmark"

Checkpoints are taken after each model. Some task types are able to checkpoint within a model. The timing of writing a checkpoint to disk and actually preserving the work is further dependant upon your setting for how frequently to allow BOINC to write to disk. This in your General Preferences. Don't read this backwards. What I'm saying is that at times when the task can be checkpointed, it is possible for a further delay based on your preference.

The default preference of at most every 60 seconds basically assures the checkpoint data is written when the checkpoint is taken.

So, yes there are "certain points", but they are basically known only to the Rosetta programmers. From a user view they may seem pretty random. And some types of tasks can checkpoint more frequently then others. They are adding more checkpointing all the time.

In general, you will always lose some amount of work when you end BOINC. And on average, it should be less then 10 minutes lost. So if you lost 10 minutes with a 3 hours runtime preference, that is about 10% on the progress bar. But it's still just 10 minutes. I often check it when I shutdown, and find that I lose less then 1 minute. But if the task happens to have long models and happens to not be able to checkpoint, it can be as much as an hour or so.

---

Rosetta Moderator: Mod.Sense