Invalid tasks

Joined: 18 Mar 20
Posts: 1
Credit: 31,430
RAC: 0
Topic 221349

G'day! New to Einstein@home and BOINC in general. I recently started running BOINC on my system and found that about 40% of my completed tasks are invalid. When I looked at the tasks, they appear to be invalid from multiple users so it rules out any system issues at my end.

1. Are these bad work units?

2. How many users is the work assigned to before the unit is stopped from being assigned? For instance, 444133429 has been sent out 7 times at the time of writing this of which it has failed validation in 6 cases and is pending results in the other case.

A slightly different topic: Are these tasks that can only be done on CPUs that cannot be run on GPUs? If so, shouldn't assignment for those can be run on a GPU be preferred/ limited to GPUs since it invariably results in faster returns (and I'm guessing, a more efficient use of computing resources and time?) For instance, in 444133429, the total run times for the GPU issued work (3) totals up to 40 minutes but the other 3 assignments were to CPUs and they spent over 31 hours arriving at the same conclusion. I'd think these 31 hours of CPU time could have been used by work that could not be run on GPUs.



Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 1,304
Credit: 2,146,149,657
RAC: 11,757,232

1) Yes.2) Your answer is in

1) Yes.

2) Your answer is in the task description.

Max # of error/total/success tasks:20, 20, 20

So the task needs to be sent out 20 times and have twenty failures before it is aborted.

Depends on whether the scientists/developers create a gpu for that campaign search.  In the case of the BRP campaign, they did in fact once have a BRP4G campaign that used the gpu but it has long finished many years ago.

I actually am running the BRP4 cpu apps on my Jetson Nano gpu because a developer created the app for the Nano.  But it is not an official stock application and runs on the anonymous platform.

The project is concentrating on Gravity Wave searches and there is a gpu app for that search.



Comment viewing options

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