Another one bites the dust...

Just after the 20th trickle went up, so that makes it at the 50% mark doesn't it?
I've got another long coupled model at 47%, will it get passed the 50% hurdle?

Let's see if things go better with the shorter models currently available.

Bernard, you may be doing this already but it improves your chances if you always suspend computation and exit BOINC before shutting down. Also some of these models will fail whichever computer they are run on and the data sent back up to the failure point and the fail itself is useful. It may be worth restricting your machine to the shorter regional models by going to your account page for CPDN if you run processor intensive stuff as this can cause models to fall over. Another option is to suspend activity before doing this if it is predictable.The regional models are much less sensitive to problems but like the longer ones are not always available.

Thank you Dave,
I always exit the BOINC manager before shutting down, however I do not suspend the task, but check "Stop running tasks when exiting the Boinc manager"

I will restrict the models computed on my computer and see how it goes.

Thanks

For information, I have noticed that the Spectraview application (a NEC monitor calibration program) probably caused 2 of my climate prediction tasks to crash with the "Error while computing message(running 8 cores at the time) when a Spectraview calibration reminder message appeared on the screen.

The reminder message was set to appear every 14 days. (The monitor gets colour calibrated to retain accuracy of colours for photographic work).

This event happened despite having Spectraview listed as an "Exclusive Application" in BOINC. (I had previously been suspicious that the Spectraview warning message was an issue).

I have now turned off the Spectraview calibration message and had zero further problems.