My models seem to crash consistently just after 30% completion. Exit code -5, generic crash. Granted credit 1273.90 for each of these affected WUs

Twin AMD 2400+ (on an ASUS A7M266-D), 1 gig RAM. Plenty of disc space available and allocated to CPDN. Win XP SP1. Boinc client 4.19, Hadsm3 V4.12.
CPDN is "sharing" the machine with F-secure anti virus, Zone alarm, ASUS PC probe and others.

I did have a problem with -5 exit errors happening unpredictably. I think this is resolved now, but there seems to be a hard limit that no WU will reach 31% on my machine.

Any guesses?
Iain-F

---

There have been some problems with people crashing at end of phase (right before the 24th trickle of various phases), but these have had different error codes (-1073741819).

Given that nearly all your errors are -5, I would suspect hardware faults. I would try <a href="http://www.climateprediction.net/board/viewtopic.php?t=2124">basic hardware maintenance</a> and <a href="http://www.climateprediction.net/board/viewtopic.php?t=2126">hardware specific tests</a> to troubleshoot the problem.

Good luck.

Try all of Geophi's suggestions first to see if you can get cpdn boinc to work.

I couldn't get boinc cpdn to work on my AMD. After a lot of time trying, I decided to abandon boinc and go back to Classic cpdn again, which is still in operation and therefore an option. But you don't get boinc credits for it and can't time-share with other boinc projects. It's usually extremely reliable and is just as useful to the researchers. After doing a few models, it is possible to try boinc again.

You get to the classic download from the home page, then the Open Uni course link.

---

__________________________________________________

Hi,

I had the first couple of modells crash with an error code -5 under BOINC 4.19, too.
Since I started defragmenting the hard drive(s) on a regular basis (about every 1-2 weeks), that has not happened any more and I am close to completing two more WU. (Keeping fingers crossed).
Before defragmenting the drive make sure you suspend &amp; exit BOINC first.

I have the idea that maybe BOINC 4.19 is not very patient when exiting and that it may terminate the climate modells before all the data has been properly written on the harddrive if the disk is too fragmented.
That would explain why the error code -5 always appeared right after start-up, i.e. when the model tries to load the data files they are corrupted.

Also, I always do suspend and exit BOINMC before shutting down the computer.

Just a thought, but worth trying.

Friedrich

---

I love CPDN, but I have only Windows clients.
--