Wilgard, old PC's are not suitable for crunching climate experiments!
The same applies to several other projects such as Aqua and the Lattice Project. Better to chose projects here, as most are fairly small. You did not say what CPU you have, so I cannot be more specific. It's a bit much to ask the developers to create code for your own specific needs, they have a million others to support, and sometimes it is essential that projects can run past their deadline, as the tasks just take longer than predicted.

Hello Wilgard,
The problem is that many people have their Duration Correction Factor (or whatever it is called) go boinkers. They hop onto the board complaining and we tell them how to reset it to 1, so their BOINC installation can start working normally again. If their copy of BOINC terminated all work units automatically, then it would never self-correct. The best thing is for you to only run projects with short work units and try not to download a second work unit until about 2 hours before you finish the current work unit. You can see how long projects run on average in Statistics - By Projects - (select a project).

Lawrence

In fact, my parents have the same problem and I suppose several users also. My parents have a new pc with core i7 but use their computer few hours per week about 4 / 5 hours. So, their computer terminated some work unit in time but something no because to finish all tasks in time they should more use it.
So, in my opinion the problem is not that I have a old pc and my parent a new one. The problem (which is not a problem but a waste of time) is that the Boinc agent download work unit which could be terminated after the deadline.

I suppose the dealline means that the same workunit may have been sent to another user because we don't have crunch it in the remaining time.

I take care about the "average time" of projects and I select projects that will be crunched on it.

But think about the number of workunits sent by all Boinc users after the dealline.

wilgard,

Several suggestions have been made to you how to circumvent. If you set the ''connect about every....'' interval to 0.00 and the ''additional buffer / cache'' to 0.01 days your and your parents computer will only have the Work Units the clients are actively working on plus they will only fetch a new work unit shortly before finishing the current, these thus having 10 days to finish a task (current standard WCG deadline)

These settings are presuming the computers are connected to the internet when on.

As said, WCG nor the client will cancel tasks that are being processed i.e. they are allowed to finish. The exception is when WCG knows a task is bad... then WCG servers will send an abort instruction. In your case by keeping just the most minimal cache setting waste is for the largest part avoided. It's accepted that a small amount of waste is the result with so many different needs by the thousands of volunteers.

Mind you, in the client message log there are red printed lines to advise on late tasks, so it's left till now to the volunteer to decide to manually cancel tasks that already had CPU time. The developers and grid operators are aware of the dilemma, where WCG chose to give credit when no Too Too Late. That view could always change... tool late slows down batch completion, but that is most of the time none critical.