As I noticed the boinc screensaver popping up, before I would let it completely do so, I hit cntl-alt-del to get to the task manager and get back to the desktop, this produced a "computation error" in the boinc advanced view next to the current chunk of CUDA work.

Several work items were completed earlier using CUDA, but now after the computation error, the client isn't trying to restart that chunk of work or fetch others.....will CUDA ever restart on another piece of work?

I'm a newbie.....I've disabled the screensaver...and rebooted, and still CUDA is detected but does not try to redo that chunk of work or fetch more. :(

I solved this myself by detaching from the SETI@HOME project in BOINC, and then reattaching. Now it recognizes CUDA again and schedules work....hopefully it will not produce another "computation error" since I've disabled the screensaver.

Unfortunately, there is a different inefficiency... Since I have a dual core machine, when CUDA runs, it uses one of the CPU cores just for interface with the system, and the rest of that core's idle time is left unused.......right now I'm also running the PrimeGrid project............perhaps running only one project will change this core utilization problem somehow? I'll see if I can answer this one myself too...

Ok, so now I've detached from PrimeGrid, and even restarted BOINC, and now BOINC is only using CUDA and no cpu (negligible) for SETI@HOME whereas previously I could use both the cpus and the CUDA GPU.

Don't know if this is the BOINC resource scheduler or a project specific issue.......just interested if anyone knows why the resources don't get utilized immediately.

If you want to use both cores + your GPU try running one of the optimised apps

http://setiathome.berkeley.edu/forum_thread.php?id=51621

The current app is a few posts down, Version 8a

You will be able to use all your core + GPU with the app.

---

Looks like the BOINC scheduler may have something to do with the issue...

http://boinc.berkeley.edu/trac/wiki/GpuWorkFetch

I'm done with this post......but thanks for listening.