computational errors

Jim Iskiyan
Jim Iskiyan
Joined: 3 Mar 06
Posts: 1
Credit: 17416799
RAC: 19539
Topic 225858

Running EINSTEIN on 2 PCs both windows 10. One PC runs with no errors, the other PC constantly gets the following error:

computational error   Gravitational Wave Search O2 Multi-Directional 2.09 (GWnew)

Any thoughts??

mikey
mikey
Joined: 22 Jan 05
Posts: 12676
Credit: 1839076099
RAC: 3992

Jim Iskiyan wrote: Running

Jim Iskiyan wrote:

Running EINSTEIN on 2 PCs both windows 10. One PC runs with no errors, the other PC constantly gets the following error:

computational error   Gravitational Wave Search O2 Multi-Directional 2.09 (GWnew)

Any thoughts?? 

All the techy people are in the Crunchers Corner forum you should post this over there.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117461720620
RAC: 35508699

Jim Iskiyan wrote:Any

Jim Iskiyan wrote:
Any thoughts??

Contrary to what someone who should know better might have said, the correct forum to post about a problem would be the "Help section" - perhaps a forum labeled, "Problems ..." might be a rather obvious choice :-).

The other very important requirement is to actually give some detail about the background to the problem and what you might have already tried to resolve it.  For instance, you could give a link to the particular machine.  You could also look through the information that gets returned for a failed task to see if the reason for failure is listed there.  Have you tried any of that??  If you give no information, other volunteers who are busy with with their own stuff are more than likely to give your very brief message a complete miss.

I did find the machine showing the compute errors.  I picked this failed task and looked through the complete list of information that was returned to the project about it.  You could easily do the same.  There is quite a bit of stderr output because the task was stopped and restarted multiple times.  Towards the very end of the output, I found:-

Internal function call failed: Memory allocation error

which probably corresponds to the point where the task actually failed.

I don't normally run CPU tasks - all my machines have GPUs that do the crunching.  Recently, I was donated a couple of HP desktop machines which don't have GPUs suitable for crunching.  I've been using them to run GW CPU tasks.  I know from personal observation that these tasks require more than 2GB RAM per instance.  Your machine is listed as having 8 threads and 16GB of RAM.  From the above error message I would suspect you may be trying to run all 8 threads with not enough memory.  It shouldn't fail - it should just be horribly slow.

Maybe you have some flaky memory.  You could try testing your RAM.  You could also try setting BOINC preferences to use only 50% of the cores (ie. only run 4 tasks) to see if that solves the issue.  You will probably see an amazing speedup.  Having enough memory certainly makes a huge difference on the HP desktop machines I'm playing with.

I'm running Linux and the memory used by the OS is quite modest.  With Windows, you need a fair bit of RAM just for the OS itself.  Since you don't mention how many tasks you are running, we can't know if lack of RAM is causing your problem.

Cheers,
Gary.

Exard3k
Exard3k
Joined: 25 Jul 21
Posts: 66
Credit: 56155179
RAC: 0

This smells like overclocked

This smells like overclocked or damaged RAM. Capacity isn't really a concern for BOINC in my experience...memory gets stored in swap and you just get the performance hit for running applications in swap space, but you shouldn't get compute errors unless memory+swap is filled 100%, but then the OS would probably stall as well.

Memtest86 would be my advice too.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.