| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: Computer freezing due to BOINC computation? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 36
|
|
| Author |
|
|
Wolf Fivousix
Cruncher Joined: Mar 22, 2013 Post Count: 10 Status: Offline Project Badges: 
|
Hello everyone, right now I have multiplied the files in my HD and left only 6GB for BOINC to run, see if it makes a difference. Regarding fragmentation, the HD was only 3% fragmented and defragmenting it did not solved the problem.
----------------------------------------Mr. Kermit, Memtest ran for 30h and no problems were detected, can I safely assume that the DRAMs are not defective? (although they might be part of the problem, the hardware in itself would not be the problem.) As looking into the settings I had "Leave applications in memory while suspended" UNchecked, I'll be changing this setting once the Full HD test is performed and get back with the results. Regarding windows event viewer, the only critical events are kernel power issues, which happens when I have to hard-boot the machine once it "freezes". Wow, I knew that BOINC was CPU intensive, never thought it would be that I/O intensive. But if the HD is the issue, how come changing the Motherboard got rid of the issue for a month if the HD in itself never changed? I must say I did formatted the SSD for a clean install of the OS, but regarding BOINC all I did was deleting it from the HD and re-installing. SekeRob, the issue happens with the computer in use as well as idle, so for example, if I turn on the computer, open BOINC and let it run it will eventually freeze, but if I do not open BOINC it can stay on theoretically "forever" (I don't leave the computer on during the night, unless some kind of test is running, like memtest). One funny thing, my memory usage never got past 5GB even with the 8 tasks running, even though my settings says that it could use up to 60% while PC in use and 80% idle. Do you think this is relevant? Katoda, on my previous computer I swapped PSU with a friend (both were 500W at that time) and the problem persisted. HWM voltages for all components shown did not change from the usual at the moment the computer "froze", unless the freezing originated from a spike that crashed the system before the software could even register it. Eitherway I unfortunately do not have another PSU to make the same test again =( . PMH UK, thank you for the ideas, never tried such things, I will add to the tests list and get back with results once they have been performed. [Edit 3 times, last edit by Wolf Fivousix at Mar 19, 2015 12:01:29 AM] |
|||
|
||||
|
MrKermit
Advanced Cruncher Joined: Jun 13, 2009 Post Count: 95 Status: Offline Project Badges: 
|
Mr. Kermit, Memtest ran for 30h and no problems were detected, can I safely assume that the DRAMs are not defective? (although they might be part of the problem, the hardware in itself would not be the problem.) Yes - I am less skeptical it's Ram if memtest is clean.As looking into the settings I had "Leave applications in memory while suspended" UNchecked, I'll be changing this setting once the Full HD test is performed and get back with the results. Interesting. Wow, I knew that BOINC was CPU intensive, never thought it would be that I/O intensive. But if the HD is the issue, how come changing the Motherboard got rid of the issue for a month if the HD in itself never changed? I must say I did formatted the SSD for a clean install of the OS, but regarding BOINC all I did was deleting it from the HD and re-installing. Consider trying this again? Maybe there is a bad file or uncommon issues from having many cep2 wu running simultaneously. I've had a few in my fleet that needed a clean directory reset. Also make sure the directory / disk are not setup to be compressed automatically by Windows. I am trying to imagine background tasks like that getting into a bad state fighting with perpetually open files.SekeRob, the issue happens with the computer in use as well as idle, so for example, if I turn on the computer, open BOINC and let it run it will eventually freeze, but if I do not open BOINC it can stay on theoretically "forever" (I don't leave the computer on during the night, unless some kind of test is running, like memtest). That looks right for Ram usage Based on previous posts. The failure mode seems really specific to boinc+windows. If you run only non cep2 projects does it still freeze?One funny thing, my memory usage never got past 5GB even with the 8 tasks running, even though my settings says that it could use up to 60% while PC in use and 80% idle. Do you think this is relevant?  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Wolf Fivousix,
----------------------------------------As I mentioned earlier in this thread, starting 8 CEP2 simultaneous is disastrous on the system efficiency, the HD just buckles under the I/O load. Depending on caching etc I've seen an hour just been wasted on setting up the tasks before actual computing starts. This is the main reason why staggered starting feature is requested from developers, or manual staggered starting is done after reboot... most cumbersome. Now the important bit to mitigate the I/O buckling v.v. the "(I don't leave the computer on during the night, unless...": [If not doing so already] Since you are part time computing, strongly recommend to hibernate your computer when done for the day [easy with Windows]. At that time the jobs computing are saved 'as-is' from RAM to disk and restored on power-up to continue where they were instead of falling back to the last checkpoint. The hibernating can be automated with W7/8, starting by telling BOINC to quit computing after there has not been any user input for for instance 60 minutes [mouse or keyboard inactivity]. On power up, you will find your computer setup as if you've never left... a webbrowser will still be open on the last page you were on. On my laptop I'm doing this all the time, but desktops can be set up the same way. Edit: quite is quit! [Edit 1 times, last edit by Former Member at Mar 20, 2015 8:01:33 AM] |
||
|
|
Wolf Fivousix
Cruncher Joined: Mar 22, 2013 Post Count: 10 Status: Offline Project Badges:
|
Hello guys, quick update:
The computer have been running with the HD quite full (only 5GB left for BOINC to use), but it has yet to freeze. I tried my best to not change any behavior that I usually have using the computer, even though it is mostly not dependent on it. Anyway the "leave application in memory while suspended" is still UNchecked. One thing that I may add though is that a 9~12h long task that had been in my list since the issue started (and I could not get rid of it since everytime the computer froze it would go back to 9h) is now gone. Could particular tasks be somehow related to the issue? I have performed the HD scans as indicated by PMH UK, and both quick and extensive SMART scans passed. Mr Kermit, I have not yet tested computing other projects at 100% CPU time, but I will be performing it. I was leaving the re-formating the SSD (and then HD too) as the very last test, as it will take a good amount of time to re-install and configure everything back into the computer. (And it had not worked with my previous motherboard, but then again, it was probably defective at that point.) SekeRob, to be honest I have been with the "lost in task computing time" since I began donating, that was one of the first things I realized when watching the program, so I am quite used to and, considering the length of the tasks (which rarely are shorter than 5h), I gave up on considering it an issue. My knowledge regarding the BOINC operation is very limited, but around 5 min after windows have been loaded, HW Monitor and Windows Task Manager shows that the CPU started running BOINC at 100% in all cores. I would assume that computing already started at this point, but I have been watching the Resource Monitor on windows for some time now since I read your post and I can see that the disk "System" Write parameter is seldom bellow 4MB/s (or above 8MB/s) for the System parameter while BOINC is running. Once I "snooze" it though, boinc shows up using around 500KB/s and after a while disappears and the system goes down to about 8KB/s. Once I turn "snooze" off, 8 'wcgrid_cep2_qchem' show up with some 600~800KB/s each, the system shows a big spike of 22MB/s writting till it goes back to the first state. Unless BOINC is "snoozed", I see most action in the HD drive, with a scale of "Queue Length" 0.05. Would you help me make some sense out of this data? Up to this point I have never considered analyzing the HD I/O operations, so how "normal" or "dangerous" are this numbers? Is there anything out of normal on this values? Once again, thank you for all the help, it has been both invaluable and a learning experience. |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
'Leave application in memory when suspended' is fundamentel to CEP2 [It's recommended on the System Requirement page as well]. Each time BOINC is snoozed or paused, active jobs unload -except- when still before the first checkpoint [of possible 8].
For gaming etc, BOINC has an <exclusive_app> cc_config.xml option as well. Whenever a specified user application loads, BOINC is put in stasis. If the user need more memory for the apps, BOINC 'in memory' will be moved to the swap file. At the end of gaming etc, exit those apps and BOINC will resume without delay. |
||
|
|
Wolf Fivousix
Cruncher Joined: Mar 22, 2013 Post Count: 10 Status: Offline Project Badges:
|
Hello guys,
After filling up the HDD and leaving only 5GB for BOINC, it have not yet presented the same behavior for this couple days, even after activating "leave applications on memory". Now I have deleted all extra files and returned the HDD to its previous state, I will see if the system still complains before trying to reduce the workload from 8 to 4 tasks, as sugested. Thank you for all the help, I'll get back as soon as I either get a negative result or move on to the next test. |
||
|
|
MrKermit
Advanced Cruncher Joined: Jun 13, 2009 Post Count: 95 Status: Offline Project Badges:
|
Any news?
---------------------------------------- |
||
|
|
Wolf Fivousix
Cruncher Joined: Mar 22, 2013 Post Count: 10 Status: Offline Project Badges:
|
Hi Mr. Kermit, I started running 4 tasks yesterday, but no problems so far.
I am still monitoring the HDD I/O, and running 4 tasks decreases the operations considerably (as well as the overall temperature of the system in about 10ºC). If nothing happens for a couple of days I'll change back to 8 tasks but change the values for the RAM that BOINC is using. |
|||
|
||||
|
MrKermit
Advanced Cruncher Joined: Jun 13, 2009 Post Count: 95 Status: Offline Project Badges:
|
Any luck? I am running with Max 8 tasks and getting occasional error'd results but no crashes under linux FWIW.
----------------------------------------Cheers! MrKermit |
||
|
|
Wolf Fivousix
Cruncher Joined: Mar 22, 2013 Post Count: 10 Status: Offline Project Badges:
|
Hi Mr Kermit,
I have been lowering the values of the RAM by 10% every couple days, right now I am at 30% when computer in use and 40% when computer is idle, but so far no crashes have happened. I plan to keep this approach until 5% and 10%, which will be way less than the usually 2~3GB that BOINC takes. |
||
|
|
|