If I set the preference to ' write to disk at most every 20 minutes';instead of the default 2 or so minutes .
I have 2 questions
1 ) Does this gain me any crunching speed ?

2)When Boinc switches to another project. do I stand to lose any unwritten calculation data to disk on any projects, or do all projects do a final write to disk before switching to the next project ?
3)Am I safe to assume that doing this only costs me up to 20 minutes of data if the program stops or is stopped or crashes for any reason ?

If I set the preference to ' write to disk at most every 20 minutes';instead of the default 2 or so minutes .
I have 2 questions
1 ) Does this gain me any crunching speed ?

2)When Boinc switches to another project. do I stand to lose any unwritten calculation data to disk on any projects, or do all projects do a final write to disk before switching to the next project ?
3)Am I safe to assume that doing this only costs me up to 20 minutes of data if the program stops or is stopped or crashes for any reason ?

1. Not for Rosetta. The amount written to disk is very small. It could make a difference for CPDN - but they have a hardcoded interval which can't be influenced by the user.
2. Yes as long as you don't have "Leave application in memory" activated. At the moment there is no such feature like switch only after checkpoint (which I would like to see implemented).
3. At most 20 minutes if apps are unload from memory. Nothing if apps aren't unloaded.

Finally this write to disk setting does not help at all and most projects don't even respect it.

---

3)Am I safe to assume that doing this only costs me up to 20 minutes of data if the program stops or is stopped or crashes for any reason ?

3. At most 20 minutes if apps are unload from memory. Nothing if apps aren't unloaded.

You will lose any computing done since the last checkpoint. Presently, a checkpoint is performed at the end of each model processed. For large proteins, a model can take over an hour to process, and therefore, regardless of the "write to disk" frequency, you can easily lose more than an hour of crunch time. This is why the checkpointing is a high focus in the release they are testing on Ralph. More frequent checkpointing is coming very soon to assure valuable work is saved, and doesn't have to be repeated the next time the WU becomes active.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/