http://www.gpugrid.net/workunit.php?wuid=323278
http://www.gpugrid.net/workunit.php?wuid=324367

these keep popping up.
what does it mean?

is it aborting before running, or ignoring returned WUs? is it anything i should be correcting in my settings?

http://www.gpugrid.net/workunit.php?wuid=323278
http://www.gpugrid.net/workunit.php?wuid=324367

these keep popping up.
what does it mean?

is it aborting before running, or ignoring returned WUs? is it anything i should be correcting in my settings?

To answer the last question first... no ... there is no need to change settings ...

DEPENDING... the project at times is issuing new and improved tasks to us ... some are less new and improved than others. Meaning, they are very and definitely broken. So, they fail, or mostly fail ... this, when discovered by the first people to run these tasks the remainder are canceled before they waste other people's time.

This is the natural consequence of the research and trying new things. I don't know of ANY project that does not have problematical tasks from time to time ...

I was curious because the WU itself gives credit. and one of the machines that crunched it gives credit.

this appears to happen on my quad core with an 8800gt (3 wu's queued) than it does on my dual core with the gts250 ( only 1 queued wu).

from what it appears the server is canceling one of my queue'd tasks. and i was hoping for some verification that it is not dumping running or completed tasks.

unless someone else has an answer, I'll chalk this up to basing the work queue on cpu count not gpu count.

We are temporarily sending redundant results per WU.
As a consequence, when one of these results return successfully, the unsent or queued ones are canceled out, it should NEVER cancel the running ones.

This feature improves a lot the return time of the simulations we send and therefore it allows us to modify/improve/analyze the results sooner.

ignasi

Hello,

I got 2 WU cancelled too :

http://www.gpugrid.net/workunit.php?wuid=324310
http://www.gpugrid.net/workunit.php?wuid=326132

My Computer is 28909.

My 8800GT crunched these 2 WU for 8-10 hours each and
the results were cancelled after reporting.

As i understand it because 'minimum quorum = 1' and 'initial replication=2'
and because someone reported WU before me i had no credits and my computer crunched for nearly 20 hours for nothing !

minimum quorum should be set to 2, no ?

Is it normal ? Am i wrong ? It seems i'm loosing my computer time these last days.

Regards,

Lionel

Edit : Perhaps i misunderstood something. i'll try to look deeper tonight.

We are temporarily sending redundant results per WU.
As a consequence, when one of these results return successfully, the unsent or queued ones are canceled out, it should NEVER cancel the running ones.

This feature improves a lot the return time of the simulations we send and therefore it allows us to modify/improve/analyze the results sooner.

ignasi

thanks. was just looking for some clarification.

i guess the real fix for the frequency that this happens on my quad is for boinc to allow queue/scheduling based on gpu count not cpu count.

Have a good one.

Yes. And we experienced troubles trying to udse this feature.
It must be fixed from BOINC dev team.

cheers,
ignasi