Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 35
|
![]() |
Author |
|
xintruderx
Cruncher Germany Joined: Apr 6, 2006 Post Count: 37 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
If I restarting the system the cpu efficiency goes down to 5-15%. I updated Linux (Version 11.04) and installed 2 packages (libstdc++5, freeglut3) before, but it brings nothing.
----------------------------------------![]() |
||
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1680 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
IMHO the root cause of your performance problem is probably not located by boinc respectively DSFL but in your system configuration.
----------------------------------------By the way, the mentioned Linux based hosts are two Phenom II x6 and one Athlon II x2. The Windows XP Pro based host is a Q6600. |
||
|
KWSN - A Shrubbery
Master Cruncher Joined: Jan 8, 2006 Post Count: 1585 Status: Offline |
I use a Intel Core2Duo with w7 64bit and there is no probleme with finishing the WU's. But my Server with AMD Phenom II X6 1090T with Linux Ubuntu 64 bit produced only Errors and i don't no why. There is no finished one, since i start with the project. All 7 DSFL WUs that I've run on my Phenom II X4 910e under Ubuntu 10.04.3 64-bit have completed successfully (6 valid, 1 PV). Also both WUs that I've gotten on my AMD Fusion machine running Ubuntu 11.04 64-bit have completed and validated. So neither AMD CPU, nor 64-bit Linux, nor their combination, torpedoes DSFL. I'm running desktop Ubuntu rather than the server version, by the way. The bolded system is exactly what I'm running and I've had zero errors on over 50 tasks thus far. It's not Linux or the CPU causing your problem. ![]() Distributed computing volunteer since September 27, 2000 |
||
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1680 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
My previous remark was not focused on Linux furthermore on the other applications and services running on the host.
----------------------------------------After nearby Yves ---------------------------------------- [Edit 1 times, last edit by KerSamson at Sep 4, 2011 2:39:10 PM] |
||
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I'll bite. I have one dual-boot machine (all my others are Linux only). I'll run exclusively DSFL on the Windows side for a few days, then switch over and do nothing but DSFL on Linux. Will report back. The variability in DSFL task runtimes has been way too high for me to try to make any comparisons. If the runtimes settle down later on, I may try. ![]() |
||
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I have Ubuntu 10.10 crunching 15 threads with no issues. The run times range from 6 to 10 hrs though. Not sure if this is normal or not. My win7 64 seems to crunch these fine also but take a little longer.
----------------------------------------
Crunching for humanity since 2007!
![]() |
||
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Going by what Uplinger said and others observed in the Looong running WUs thread, there have been changes in task length, and there will be more.
Until we know for sure that they have settled on a size range and that WU length is the same for both Linux and Windows, there is not much point making performance comparisons. As for why, I'm guessing the techs are trying to make the tasks returned stats and ranks more meaningful. I think it also helps Boinc if tasks from the one project are the same or similar length. So, in a few weeks it might be worth another look. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
![]() Almost eEvery new project has an algorithm for sizing and creating work units from the work that the project scientists want done. There is almost always a lot of experimenting to make the algorithm work well and sometimes we have to throw our hands in the air and accept a lot of variation. Lawrence |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Since badges are based on CPU time, the fact that windows may run the same work unit a little longer is a plus, in my opinion.
My main concern is why some work units run with terrible efficiency. e.g. ![]() C2Q6660 is an Intel Q6600 running fedora 32-bit with the fedora-issued BOINC 6.10.45; I have 2 other machines with that same version of fedora and BOINC, but with AMD processors, that I've never seen get 'stuck' with such poor efficiency (~12%) work units. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello ZoSo,
----------------------------------------Are your Intel CPUs using hyperthreading? This will present the appearance of terrible inefficiency, but in reality the CPU will produce about 15% more per unit time than it would without hyperthreading. Since this is done by using 2 threads per unit time, the reported efficiency will be 115% / 2 or 57% efficiency per task if both tasks are treated equally. If one task is favored, then the other task will look even worse. Lawrence Added: ![]() [Edit 1 times, last edit by Former Member at Sep 11, 2011 9:52:15 AM] |
||
|
|
![]() |