Version 7.2.42(x64) according to About box for BOINC manager. Mint-17 KDE 64-bit with all updates applied.

I participate in quite a few projects. Recently I've upgraded the video card to one with 384 "cuda cores" so I can participate more meaningfully in GPUGRID. I have had a work unit sitting in the queue for roughly a week. Estimate shows 41 hours. Due date Sun 9 Nov 2014... It has not been started. Instead, work units from SETI and other projects, with due dates the 3rd week of December have been completing willy-nilly. This is a 6-core with an SSD, so when I leave it run BOINC at night, it can get a _lot_ done. The scheduler seems to be completely ignoring this work unit.

This partition gets booted when I'm done working for the day. It runs BOINC until I come back to work in the morning or, on those rare occasions I have a weekend off, will run it the entire weekend. Prior to upgrading to Mint-17 the scheduler always seemed to make sense. Work units with the closest due dates seemed to be processed first. GPUGRID it seems to just ignore. This isn't the first packet I've watched _never_ get scheduled.

---

Every single one of your recent GPUGrid tasks, on both machines, has failed with an error.

http://www.gpugrid.net/results.php?hostid=185997
http://www.gpugrid.net/results.php?hostid=183557

There's something more seriously wrong than just a scheduling problem: both machines refer to a driver error:

SWAN : FATAL : Cuda driver error 3 in file 'swanlibnv2.cpp' in line 446.

Cuda drivers in Linux are outside my competence: I suggest you post on the GPUGrid message boards to ask for help.

I have posted there. Not much more I can do. I installed the nvidia driver and the log identifies the cua.