- Joined
- Jan 27, 2011
- Location
- Beautiful Sunny Winfield
I've been running a fairly non-optimized Linux CUDA app (setiathome-6.11.x86_64-pc-linux-gnu__cuda32) and see that there is a newer (and hopefully improved) app (setiathome_x41g_x86_64-pc-linux-gnu_cuda32)
I'm curious if anyone has any experience with the new app. I stopped boinc, modified my app_info.xml for the new app and restarted boinc. When I saw it try to download the new app (doh!) I stopped it again and copied the new app to the correct directory. When I restarted boinc I found the messages:
So I'm curious if I have something wrong in the setup or if there is truly no work. (Also note that I'm configured to only run CUDA tasks for SETI since I use the CPU for Rosetta.)
...
While I was composing this I now see
And the new task is happily crunching away at it's first WU. Check that. First WU finished in 8 minutes and it's nearly finished on the second one. And it's keeping the GPU 50-60% busy. At present I'm only running one CUDA instance. I'm going to leave it that way for a while to compare RAC with old and new. Once I've established that, I'll try running two instances to see what happens.
As an aside, when I ran out of Rosetta work units I temporarily enabled Einstein and was surprised to see that it has combined CPU/GPU applications. It was running one of these that soaked up one CPU core while keeping the GPU at over 70% utilization. Looks like a good DC project if you want to easily utilize both GPU and CPU resources (at least for Linux.) Both SETI and F@H require some skills to get GPU tasks running on Linux.
I'm curious if anyone has any experience with the new app. I stopped boinc, modified my app_info.xml for the new app and restarted boinc. When I saw it try to download the new app (doh!) I stopped it again and copied the new app to the correct directory. When I restarted boinc I found the messages:
Code:
Mon 16 Jan 2012 09:24:29 AM CST SETI@home Sending scheduler request: To fetch work.
Mon 16 Jan 2012 09:24:29 AM CST SETI@home Requesting new tasks for GPU
Mon 16 Jan 2012 09:24:35 AM CST SETI@home Scheduler request completed: got 0 new tasks
Mon 16 Jan 2012 09:24:35 AM CST SETI@home Message from server: No tasks sent
Mon 16 Jan 2012 09:24:35 AM CST SETI@home Message from server: No tasks are available for Astropulse v5
Mon 16 Jan 2012 09:24:35 AM CST SETI@home Message from server: No tasks are available for Astropulse v505
...
While I was composing this I now see
Code:
Mon 16 Jan 2012 09:29:43 AM CST SETI@home Sending scheduler request: To fetch work.
Mon 16 Jan 2012 09:29:43 AM CST SETI@home Requesting new tasks for GPU
Mon 16 Jan 2012 09:29:48 AM CST SETI@home Scheduler request completed: got 23 new tasks
...
And the new task is happily crunching away at it's first WU. Check that. First WU finished in 8 minutes and it's nearly finished on the second one. And it's keeping the GPU 50-60% busy. At present I'm only running one CUDA instance. I'm going to leave it that way for a while to compare RAC with old and new. Once I've established that, I'll try running two instances to see what happens.
As an aside, when I ran out of Rosetta work units I temporarily enabled Einstein and was surprised to see that it has combined CPU/GPU applications. It was running one of these that soaked up one CPU core while keeping the GPU at over 70% utilization. Looks like a good DC project if you want to easily utilize both GPU and CPU resources (at least for Linux.) Both SETI and F@H require some skills to get GPU tasks running on Linux.