Posts by dbltapp

InfoMessage
1) Message boards : BOINC client : stats not displaying new project time
Message 12506
Posted 14 Sep 2007 by dbltapp
Just added Einstein. Four tasks have been running for days, but the statistics graph only displays a point at the origin. Other graphs are fine (lhcathome, predictor). Fix?
2) Message boards : BOINC client : CPU throttling: How does it work with multiple cores and/or threads?
Message 11591
Posted 11 Jul 2007 by dbltapp

:-) Yes this method of throttling does put the system hard ware under tremendous stress. ThreadMaster works better.
regards
Masud.


Regarding Threadmaster - if BOINC has 4 apps running do they each count as a thread and the total cpu % I want to alot has to be spread over 4 apps?


Just tested it few days ago. TM treats each CPU core as a separate CPU to manage. Watch out though as TM gets the % from Windows thus it sees on a 4 core 25% as 100%. To run 75% per core you have to enter say 18 (only takes integer values afaik). You'd observe 4 processes in TM hovering around 18%. Also, you need to lower the threshold value which is like 25, which on a quad TM treats as 100% i.e. regardless the sampled 18, all would run full tilt. Play a bit as i've only set it up on a C2D.... works a treat.

Oh did not see it scanning the thread, but BOINC itself needs to be set to 100%.


Just realized it's throttling the project processes, not BOINC. Ohhhhh..
3) Message boards : BOINC client : CPU throttling: How does it work with multiple cores and/or threads?
Message 11586
Posted 10 Jul 2007 by dbltapp

:-) Yes this method of throttling does put the system hard ware under tremendous stress. ThreadMaster works better.
regards
Masud.


Regarding Threadmaster - if BOINC has 4 apps running do they each count as a thread and the total cpu % I want to alot has to be spread over 4 apps?
4) Message boards : BOINC client : Client not observing the idle and max CPU requests
Message 11585
Posted 10 Jul 2007 by dbltapp
Hi. Just in the last 48 hours BOINC appears to be giving control to the threads for 1 second in 5. My 4 cores are idle for the other 4. I tried opening up my general parameters several times in several different ways but its not paying attention.

The only thing I have done in this period is add Einstein project to the other three I have running, climate, QMC, and uFluids. Its wasting a lot of time. Suspending these projects in various combinations is not helping.



It sounds like CPU throttling got turned on in your preferences somehow.

Check them on the project website and make sure "use at most of CPU" is set to 100%. Make sure you don't have a global_prefs_override.xml file in the BOINC directory.



Yes I did have some throttling turned on intentionally but when this problem cropped up I set the global preferences on all projects to 100%. No effect. YOU PUT YOUR FINGER RIGHT ON IT though with that suggestion of yours to check for a global_prefs_override.xml. Somehow I must have created one of those by accident yesterday in the process of adding the Einstein project. It was set to 10%. No wonder it was so poky! How did I do that. You fixed it. Cranking now.

Thanks one more time!!



Interesting. I'm trying to get control over my cpu just to do it. I found the above file with the cpu set to 100 while the preferences were actually set to 10%. I deleted the override file, but the cpus are still run at 100%. I even restored the override file and set cpu time to 10%. No luck.

Suggestions?
5) Message boards : BOINC client : uncontrollable cpu use
Message 11584
Posted 10 Jul 2007 by dbltapp
What OS?
What BOINC version?

Per the author of Speedfan, it's made such that on certain laptop it only works as monitor and does not interfere with the laptop's CPU/Fan control. It throughroly ignore Speedfan and runs 100% on 2 cores at runs 63/66 C if sitting on a table.

I'm not sure about the BOINC control as it is to either run 100% or stand still and alternate that to arrive at an average. 10% would mean 1 second run, 9 seconds pause in the present throttle design.

If u used 5.10.x, even for testing, you might have created unwittingly an global_prefs_override file, where 100% might have gotten stuck in. Any setting in the override file causes settings on the web profiles to be ignored.


Version 5.8.16 on XP SP2.
6) Message boards : BOINC client : uncontrollable cpu use
Message 11582
Posted 10 Jul 2007 by dbltapp

Most computers should be able to run project workunits at 100% CPU without a problem. The main exception is that some laptops will overheat at 100% and may need to have the % reduced. Sometimes just slightly raising the whole laptop above the tabletop will be sufficient to solve the problem. Many laptops have less powerful cooling than desktops.

If your computer can only run SETI or any other project's workunits at 10% of the normal speed, I wonder whether you have an overheating problem on a desktop that should be addressed before you run workunits from any project.

You might like to download Everest or Speedfan which you can try as freebies to diagnose your CPU temperature.

After you changed the % in your preferences, did you click the Update button in boinc manager to implement the change?


It's not that the machine is heating up. It's that my settings for cpu time seem to be ignored by the core client. No matter what I set the preferences at, boinc runs the cpu usage to 100%.
7) Message boards : BOINC client : uncontrollable cpu use
Message 11577
Posted 10 Jul 2007 by dbltapp

It's STILL you. :)
The "10%" thing works by running flat out for 1 second, then stopping for 9 seconds. (Or a similar proportion. e.g. milliseconds.) So what you're seeing is the time it's running, which is at 100%.



So... why wouldn't the cpu usage drop back to around 5% for nine seconds then back up for one?
8) Message boards : BOINC client : uncontrollable cpu use
Message 11566
Posted 9 Jul 2007 by dbltapp
Just started boincing today for seti. Set general preferenced to various cpu percentages down to 10%, and seti preferences to various resource percentages down to 10%.

However, when I let it run it immediately maxes out my CPUs.

I also connected to Predictor@ home, and the same thing happens.

I'm often guilty of operator error, but this time I don't think it's me.

Any suggestions?


Copyright © 2025 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.