Message boards : BOINC Manager : Boinc isn't using my CPUs it seems.
Message board moderation
Author | Message |
---|---|
Send message Joined: 20 Apr 07 Posts: 5 |
I've set Boinc to use up to 100% and I'm running the climateprediction.net project. Each of the four instances that the BOINC manager has created show time remaining in the thousands of hours. 75 days per work unit can't be right I thought so I checked my CPU usage and boinc is using 0.20% and BOINCManager is using 2%. This is on a Mac Pro 2.66 running OS X 10.4.9. I'm not doing anything particularly demanding at the moment, just browsing the internet. I have around 90% unused CPU usage. |
Send message Joined: 20 Apr 07 Posts: 5 |
Sorry, I forgot to ask my question; How do I get BOINC to use more juice from my cores? |
Send message Joined: 29 Aug 05 Posts: 15563 |
Set BOINC up to be used while the computer is in use. You can do so through the project's General Preferences. So if for CPDN: Edit preferences. Scroll to the bottom, click the "edit preferences link" Click on Yes next to the radio button of "Do work while computer is in use?" Save changes with the Update preferences button at the bottom. Open Boinc Manager. Projects tab. Select CPDN by clicking on it, press Update. That will do it. |
Send message Joined: 20 Apr 07 Posts: 5 |
Hmmm... no joy as yet. I've told it to run the SETI@home projet as well to see if the problem was project dependent with the same results. Both are set up to use all the processors and up to 100% CPU time while the computer is in use. Net result is between 0.9 and 2.5% use of my CPUs. :( I've tried restarting BOINC several times to, to no avail. |
Send message Joined: 16 Apr 06 Posts: 386 |
How much memory does your Mac have? Each model only takes around 100MB, but you also need space for the operating system, and whatever other programmes you are running. A whole climate model is very large, and even the fastest machines around will take at least 1100 hours (48d) - many of the machines on the project will take around double that. If it's only running at 0.2% CPU time it'd take decades. |
Send message Joined: 20 Apr 07 Posts: 5 |
Boinc is using 350MB of memory apparently and is still only using about 2.5% processor time, max. I've tried restarting several times. It really is highly odd! |
Send message Joined: 5 Oct 06 Posts: 5129 |
I've set Boinc to use up to 100% and I'm running the climateprediction.net project. Each of the four instances that the BOINC manager has created show time remaining in the thousands of hours. 75 days per work unit can't be right I thought so I checked my CPU usage and boinc is using 0.20% and BOINCManager is using 2%. 75 days per workunit is absolutely right. I'm running 4 CPDN workunits on a Dell Windows workstation (along with SETI and Einstein). My quads only run at 1.86Ghz, so my 110 day estimates (at 50+% done) are in line with yours. Again, BOINC using 0.20% and BOINCManager using 2% is absolutely right. BOINC doesn't actually do anything at all (like all bosses, it directs and controls other workers, but contributes nothing at all to productivity). The real question is, how much CPU %age do the CPDN worker applications get when you're not clockwatching? As others have said, BOINC might be telling the workers to down tools and make the place look tidy while the owner (you!) is around. |
Send message Joined: 20 Apr 07 Posts: 5 |
I've set Boinc to use up to 100% and I'm running the climateprediction.net project. Each of the four instances that the BOINC manager has created show time remaining in the thousands of hours. 75 days per work unit can't be right I thought so I checked my CPU usage and boinc is using 0.20% and BOINCManager is using 2%. Got it. I see they are actually using the processor time after all, for the individual processes of hadcm3transum and setiathome. Excellent! Wow, I'm amazed that the CPDN work units take so long! 1795 hours to go. |
Send message Joined: 16 Apr 06 Posts: 386 |
You'll be glad to know that the climate models upload their climate to the server as they go (the 'trickle' you see every 12 CPU hours or so), so if something happens to the work unit, it's not a disaster. |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.