Joined: 28 Nov 02
When I run the Boinc/Seti task without any seti gpu processing all four cores report 25 (%?). Now when I run 1 or more gpu tasks, the 4 cores start changing amounts. To say 12, 13, 25, 25 The gpu task is currently reporting 25.
If I run this without -cpu_lock there is a nearly constant cycle between 12 and 25 on all cores.
If I run 3 cpu cores and 1 core dedicated to the gpu, without -cpu_lock, 2 cores stay at 25 and 1 core varies from 12 to 25.
So since my goal is "maximum production" of the over all system is there anything I can do that might smooth out the cpu variation? Which might speed up cpu production without slowing the gpu production much?
"You are entitled to your own opinion but not to your own facts." Senator and Professor Patrick Moynihan
Joined: 19 Aug 99
So since my goal is "maximum production" of the over all system is there anything I can do that might smooth out the cpu variation?
The variation means you are getting maximum production.
Personally I use app_config to reserve 1 CPU core per GPU WU and let it do it's thing. The GPU is never waiting on a CPU thread, and the CPU threads crunching CPU WUs can vary their load as necessary.
Joined: 16 Jun 01
Maximum production is quite complicated thing actually.
To maximize production of device one should carefully account for both speedup _and_slowdown of device's different parts.
And cause RAC can vary just because of different work supplied over time such accounting becomes increasely difficult being done online.
Look for few case studies of this problem:
And here is one possible way how to make it online:
http://lunatics.kwsn.info/index.php/topic,1806.0.html (next step will be to apply statistics to collected data and represent binned, averaged graphs with standard error), but when differencies big enough human eye is qute good for doing this processing in analog way just by looking what graph going higher/lower overall.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
Joined: 24 Jan 00
If you have your preferences set to use 100% of your processors and use an app_config.xml like this,
<app_config> <app> <name>setiathome_v8</name> <gpu_versions> <gpu_usage>1</gpu_usage> <cpu_usage>1</cpu_usage> </gpu_versions> </app> <app> <name>astropulse_v7</name> <gpu_versions> <gpu_usage>1</gpu_usage> <cpu_usage>1</cpu_usage> </gpu_versions> </app> </app_config>
you should get near on a constant 100% CPU load with 1 core dedicated to the GPU (that's 3 CPU tasks and 1 GPU task).
©2017 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.