Completed, marked as invalid
13 Feb 2012 14:12:00 UTC
Topic 196185
(moderation:
Hi All,
I have a slight problem that I do not understand.
My crunching has been 100% with all tasks upto: (Task ID 271449246).
Results for Arecibo and all others have been good since - so far.
Time taken on the bad task was in-line with all the other Arecibo tasks I have submitted.
Can anyone help explain, as it must be my end that has messed up?
With Regards to all crunchers
ashj
Completed, marked as invalid
)
Not necessarily. It could be just one of those things. There are a couple of possibilities to consider.
All returned results undergo a validation process where there must be a high degree of agreement between the two copies of the result for the same task. Your copy was computed on a CPU, whilst that of your wingman was computed on a GPU. There are always minor differences between these two computing devices but, in the vast majority of cases, the validator finds that the answers are 'close enough' to be acceptable. In this case, however, the validator wasn't satisfied so a third result was called for. By chance, that third result was calculated on a GPU and, as you might expect, the two GPU results would tend to give better agreement with each other and so the validator ruled against your result. So it's possible that there was noting much 'wrong' with your result - just an unfortunate case of 'not close enough' and then 'being outvoted by the other two'.
Another possibility is that some 'glitch' occurred on your machine that interfered with the output and so the discrepancy was real. It wasn't a permanent problem because subsequent results (so far) have turned out fine. It could be the first indication of an impending problem (eg machine starting to run too hot) so it would be wise to keep an eye on things. If more results start turning up 'invalid', you would need to start looking for some sort of a hardware problem.
Cheers,
Gary.
Thanks Gary for your
)
Thanks Gary for your reply.
I have now taken the box down to let it cool off - I relised, after your message, that it has been running Einstein constantly for ages.
Hopefully, my problem will not now re-occur.
Regards
ashj
Validate error on a Gamma-ray
)
Validate error on a Gamma-ray unit:
% Time spent on semicoherent stage: 1174 s
% Starting coherent stage.
% Time spent on coherent stage: 312 s
% checkpoint written: skypoint 50
% Writing output file.
FPU status flags: COND_0 PRECISION ZERO_DIVIDE DENORMALIZED INVALID
08:09:02 (4242): [normal]: done. calling boinc_finish(0).
08:09:02 (4242): called boinc_finish
Tullio
Hi Gary, It now seems it
)
Hi Gary,
It now seems it is the other way round.
With this:
Task id 269185227
It looks like my wing men are going down like the glorious WW2 Fighter pilots.
James is currently in flight, but on his past record it it does not bode well.
What is going on?
Regards
James
I have renounced to crunch
)
I have renounced to crunch gamma-ray units, too many validate errors.
Tullio
RE: Task id
)
That task belongs to this particular Workunit. I don't see anything particularly unusual with what's happening there. Your computer completed the task fine and, so far, there has been one comp error and one deadline miss. Sure, the latest host to be issued the task is also having problems, mainly with GPU tasks, but all of this is pretty much what tends to happen in a not insignificant number of cases. If you go looking through your older pendings, you will always find examples like this. There's nothing to worry about. Eventually someone will complete the quorum.
Cheers,
Gary.