Crashes Windows 10 x64

Joined: 22 Apr 10
Posts: 1
Credit: 287,821
RAC: 94
Topic 221269

After a lot of trial and error, mostly error, I've determined that the cause for my Win 10 1909 x64 computer has been crashing is Einstein@home. Right now I've suspended it in BOINC. How can I determine why it's causing problems, and how to fix things.
Thanks - DaveH52

Joined: 6 Dec 05
Posts: 2,741
Credit: 2,892,319,459
RAC: 3,297,357

Since a legion of other

Since a legion of other Einstein users are running Einstein applications on Windows 10 x64 computers without crashing, there is something else different about your computer.  I don't know how we would help you figure out and correct that something else.

If you are getting blue screens on the way down which list a reason, sometimes looking up that reason on the Internet can yield a suggestion.  

Also you may find clues in the Windows event viewer.


Joined: 20 Feb 05
Posts: 45
Credit: 350,082,047
RAC: 299,722

DaveH52 - I hope someone

DaveH52 -

I hope someone smarter than me comments on this, but here is a place to start.

Your profile shows you have an INTEL Intel(R) HD Graphics 4000 (1400MB) graphics card.

I am going to suggest that graphics card might be marginal for running Einstein work.

You certainly have enough CPU power and memory.

Sign up for Milkyway and see if you still have problems. If you don't have problems, BOINC is not the issue.




Gary Roberts
Gary Roberts
Joined: 9 Feb 05
Posts: 5,122
Credit: 36,816,994,946
RAC: 37,612,322

DaveH52 wrote:After a lot of

DaveH52 wrote:
After a lot of trial and error, mostly error,

Since nobody is able to see what you see, it would be helpful if you could list the various things you have done and what has shown on your screen as a result.  As well as listing error messages, etc., you should include things like any preference changes you have made and which particular searches you have selected to run.

DaveH52 wrote:
I've determined that the cause for my Win 10 1909 x64 computer has been crashing is Einstein@home.

It would be helpful to know exactly what caused you to reach this determination.  I had a look at the list of tasks assigned to your computer.  There is only one task currently showing, issued on 15th Feb.  On 12th March, your computer contacted the project and at that point, since the task hadn't even been started, the scheduler immediately canceled it.

I decided to see what the scheduler recorded as a result of that 12th March contact.  You can easily do this for yourself.  There is a full set of instructions for accessing scheduler (server) logs listed in a pinned thread in the 'Getting Started' forum.  To save time, here is a direct link to that log entry.  Whilst the full log is rather daunting, here are some key points.

There are 6 examples of the message, "... setting outcome ERROR".  One of the 6 relates to the currently still showing task (see first link above) that was canceled by not being started before the deadline.  So the message here is that your computer didn't even try to start to process that task.  My guess is that all six were probably in the same condition and that 5 have subsequently been removed from the online database after being reprocessed by other computers.  This will also happen to the last one shortly, when its quorum is finally completed.  If you click on the WU ID link for that one remaining task, you will see the other computers involved and that there is (at the time of writing) still one outstanding task yet to be returned.  So this quorum should hang around for a couple more days at least before being removed.

Of greater concern are the 52 examples of the message "... can't find result".  When your computer makes contact with the project's scheduler, it sends a full list of all the tasks, data, apps, etc., that are present on your computer and the scheduler checks that list against its own records.  It would appear that, outside of the normal BOINC procedures for aborting/returning tasks, you have somehow intervened and caused the BOINC client to lose track of 52 tasks.  Perhaps you might remember taking some action that caused that to happen?  Even if tasks crash and burn, the client will return the debris and there will be information about what went wrong.  If you just delete stuff, there will be no information about the cause of the problem.

The 52 missing tasks belong to either the FGRPB1G search or the BRP4 (Arecibo) search.  Both of these are GPU searches.  Whilst the FGRPB1G search isn't really suited to Intel GPUs, the BRP4 search should work, provided that you had taken care to choose a suitable driver.  Past experience suggests that there are many driver versions that cause results not to be valid within the required accuracy.  Because no results, valid or otherwise, currently exist, it's impossible to say if selecting just the BRP4 search might work for you.

If you really wish to process GPU tasks, trying to use an Intel GPU might be quite a restriction.  If your computer has an available PCIe x16 slot, upgrading with a cheap discrete GPU might be a relatively simple exercise.  There are models that don't even require an external PCIe power connector that could work well and be very productive.  I'm sure there would be plenty of people willing to give advice on something suitable.

There was no evidence in the scheduler log of any CPU tasks.  Have you tried running CPU tasks?  Your computer should have no problem with those.  If you decide to use your CPU, you should select one only of the two available CPU searches, the FGRP5 search for Gamma-ray pulsars or the O2MD search for gravitational waves (GW), using a small work cache size until you see how things progress.  Whilst the GW search is of the highest scientific interest to the project, the data download requirements are much larger and might cause issues for volunteers with a limited internet connection.


Joined: 8 Aug 05
Posts: 4
Credit: 14,624,537
RAC: 9,157

This may help. I have an

This may help. I have an older computer running Win 10 X64. The machine has 12 GB of memory. Einstein@Home is the only project running in BOINC at the moment and there are 8 tasks currently running in the list. BOINC only runs in the idle mode when I am not using the computer. Using the monitoring software I can see that BOINC along with everything else in idle mode is using about 4 GB of memory. After some time of at least hours, the memory usage in the latest incident over a period of 3 minutes ramped up to 11.9 GB and then a process called "memory compression" started and was competing with the Einstein tasks for CPU usage, with a lot of disk usage that I am assuming was writing to the swap space. I happened to start using my computer just in time to catch and watch this. The memory compression fight with Einstein continued for about 15 minutes before I moved my mouse. It took 5 minutes before the Einstein tasks quit. It took another 5 minutes before the memory compression process quit. During that 10 minutes the computer was unresponsive to mouse movements and unusable. The memory usage dropped back to 3.5 GB.

Comment viewing options

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