Windows 11 22h2 insider, radeon graphics, lenovo R480 thinkpad having too many failed tasks
30 May 2023 10:27:10 UTC
Topic 229601
(moderation:
Hi there,
I am fairly new to the BOINC scene having joined a little less than a year ago. Of all projects the fail rate of E@H seems too high. Just today in less than 10 minutes 15 tasks failed right before my eye... this seems like a problem that I need to report.
Graphics radeon 540, intel
)
Graphics radeon 540, intel uhd 620,
General
URL
https://einstein.phys.uwm.edu/
User name
Snowball101
Team name
Resource share
100
Scheduler RPC deferred for
00:57:03
Disk usage
1.89 GB
Computer ID
13126535
Suspended via GUI
no
Don't request tasks
no
Host location
default
Tasks completed
153
Tasks failed
656
Credit
User
168,537 total, 2,257.79 average
Host
138,037 total, 1,941.09 average
Scheduling
Scheduling priority
-7.48
Duration correction factor
1.0021
Last scheduler reply
5/30/2023 6:09:27 PM
Please unhide your computers
)
Please unhide your computers here on the web site https://einsteinathome.org/account/prefs/privacy so others can view the failing tasks to give you suggestions what might be wrong.
Snowball101 wrote: Graphics
)
You also don't say what kind of tasks you are trying to run, cpu or gpu or both? Unhiding your pc's, mine are not hidden so you can see what we can see once you unhide yours so someone can try and figure out what's going on.
OK... I did not even know
)
OK... I did not even know that option existed lol
But the fact that it is hidden by default really surprises me since I am really used to the Microsoft and Google etc. big companies which collect data for who knows what ( at least they are better than the communist Chinese)
Also thanks for your fast reply... seriously did not expect that.
Snowball101 wrote: OK... I
)
Happy to help when I can!! As for the default being set to not display your data it's because the EU Rules say that IS the default if you operate within their borders and yes most Boinc Projects are setup that way because of it.
As for Windows pc try running just a single task at a time, your pc only has 8gb of ram in it and with the Windows overhead etc the tasks could be unable to operate like they should.
As for your ARM pc I have no clue, a friend Keith Myers who's a cruncher and frequent poster here on Einstein, helped me setup mine years ago but after updating 2 of mine they now also fail to run the tasks without errors.
OP doesn't have enough memory
)
OP doesn't have enough memory to run the Gravity Wave tasks.
OP had network or virus scanner issues that prevents the client from contacting the project.
OP has strange 32 bit FFTW 3rd party primitives.
Thanks for the feedback. I
)
Thanks for the feedback. I run the tasks at night so I do not think any other programs except MS bloatware would hog memory intended for BOINC. ( why would MS push Windows 11 on this 2018 PC? the battery life now if ridiculously short ) However, I have noticed that even if I configure the computer to not sleep in settings, the computer would hibernate somehow. Since it is hibernation instead of shutting down, I wonder if the files or data are messed up somehow. Looking in event viewer yields no results, as the information is a bit too technical for me to handle. I am now worried that this weird error may be impacting other projects as well. How does a program know that there is a computation error? I certainly hope that I have not been unknowingly feeding projects the wrong information...
Snowball101 wrote: Thanks
)
Do a single click on a running task and then click on Properties in the column to the right, this will show you two numbers you care about as far as memory goes:
Virtual memory size 15.23 MB
Working set size 19.78 MB
What this is saying about the task I'm running, a MilkyWay task, is that it's using a max of about 20MB of memory. If you look at this LHC task you can see the memory usage is different:
Virtual memory size 5.18 MB
Working set size 11.14 MB
You can see it's only using a bit over 11MB for that task.
Some of the Einstein tasks can take more than 4gb of memory for each one, you care about the larger number whichever one it is, so if an Einstein task take 4GB of memory and you are trying to run just 2 of them you can see there is no memory left for the OS to do it's thing and your tasks will have problems and probably have errors.
Computation errors don't hurt
)
Computation errors don't hurt projects. They just waste your time and energy. Only when the result from two task iterations from two different hosts agree does the result get moved into the science database.
Your errored tasks just get sent to the next host to try and achieve consensus with the science result.
mikey wrote: Snowball101
)
Yes indeed I see that E@H tasks take more memory, but even the biggest one for a GPU task took only 1.1 GB of memory, and if I understand correctly this should be virtual memory which is actually stored on my disk? Anyway, it should not give me so much errors. If a program is out of memory it would postpone itself if I remember correctly instead of giving me an error and aborting the whole task.