I run CPDN on my office machine and this was recently changed so that I\'m now running a Dell GX520 with P4 3GHz and 1GB of RAM. However, every work unit that has downloaded since I changed machine has crashed, typically after around 30 hrs of processing.

The messages from the latest two units to give up the ghost are:

05/12/2006 15:44:15|climateprediction.net|Unrecoverable error for result hadcm3ohc_1h6d_05617276_1 ( - exit code -1073741674 (0xc0000096))

and

05/12/2006 18:43:40|climateprediction.net|Unrecoverable error for result hadcm3ohc_0xns_05591983_0 ( - exit code -1073741819 (0xc0000005))

I have also seen other exit codes, but all seem to only vary in the final digit.

Is this a machine problem, a problem transferring the code from one machine to the other or what?

Any suggestions for how to stop this so I might get to see a run complete would be appreciated.

Thanks, Brian

p.s. I appreciate that in any simulation model, there is always the possibility of the system going out of bounds, but this seems to be happening far more frequently than I owuld have expected - indeed I don\'t recall having this happen on any of the various machines I\'ve run CPDN on int he past.

---

Those numbers are Windows stop error codes.
They usually appear when something to do with the display goes wrong. e.g. the display is text, and then suddenly goes to large amounts of graphics, using more memory. More evident when the computer doesn\'t have a separate graphics card.

Sometimes updating the display driver from the card maker\'s web site helps, as does an update to the DirectX drivers.

Brian, in the README about crashes here

http://www.climateprediction.net/board/viewforum.php?f=36&sid=50e2d9652ba4de20f8db9beb32608bb3

item #5 by Mike and #6 by Thyme Lawn provide specific help in avoiding -107 graphics-related crashes.

Disable the screensaver if you haven\'t done so already. When you view your globe thro the button in your boinc mnanager, avoid maximising the globe window.

You also need to back up the whole boinc folder regularly so that if you do have another crash, you can restore the backup and continue crunching the same model. In the same crashes README, item #1 by Les explains how.

---

Cpdn news