Homogenous redundancy would probably help. Rounding problems are really mean, even worse when MSVC libraries are involved - different versions have different rounding policies (bad experiences from the RNAWorld development phase) :-/
A huge project like this one could easily affort enabling HR.
Seems they could also only send units to one or the other instead of both until they figure it out too! If the 1st unit goes to a gpu, then so does every additional resend of that unit, same thing if sent to a cpu. Yet the problem continues.
And Saenger I did not know any of the units had validated like you said. Hmmmmm!
You're not using the option in a cc_config.xml are you?,
you're reporting your task before the server has found your uploaded result files,
like every hour 15 minutes,
You're not using the option in a cc_config.xml are you?,
you're reporting your task before the server has found your uploaded result files,
like every hour 15 minutes,
Claggy
I'm using this option, a relict from GPU-Grid crunching until last week, and I never experienced any problem here.
You're not using the option in a cc_config.xml are you?,
you're reporting your task before the server has found your uploaded result files,
like every hour 15 minutes,
Claggy
I'm using this option, a relict from GPU-Grid crunching until last week, and I never experienced any problem here.
I went through a spate of Validate Errors in December with my 9800GTX+, in my case it was probably to do with my slow/flaky internet connection,
some of the files needing several attempts before they'd upload, then even if you left some time before you reported them, they'd still get validate error,
i never got an answer wether there was a script that rechecks the validate errors prior to purging,
You're using a different version of Boinc to [AF>Libristes] Elgrande71, that can make all the difference,
the latest Development version of Boinc won't ask for work within 5 minutes of an upload to stop these thing happening: [trac]changeset:22841[/trac] and [trac]changeset:22910[/trac]
My 9800GT had similar symptoms, and after a reboot I got valid results again. There's no guarantee that this is the problem, but it's sure worth a try.
It still produces some validation failures because of the problem discussed in this thread, but at least now the majority of tasks go thru.
You're not using the option in a cc_config.xml are you?,
you're reporting your task before the server has found your uploaded result files,
like every hour 15 minutes,
Claggy
Exactly, I'm not using the option.
I am going to try it.
Thank you.
You're not using the option in a cc_config.xml are you?,
you're reporting your task before the server has found your uploaded result files,
like every hour 15 minutes,
Claggy
Exactly, I'm not using the option.
I am going to try it.
Thank you.
The point was you DON'T want to use it, you're reporting too soon after upload anyway, you should try and delay your reporting,
RE: Homogenous redundancy
)
Seems they could also only send units to one or the other instead of both until they figure it out too! If the 1st unit goes to a gpu, then so does every additional resend of that unit, same thing if sent to a cpu. Yet the problem continues.
And Saenger I did not know any of the units had validated like you said. Hmmmmm!
RE: And Saenger I did not
)
I have a lot of tasks, where CUDA and SSE match. But also many, where there is a mismatch
RE: RE: And Saenger I
)
92011082, 91330383, 91993880, 91986863, 91892702, 91890236 only from the first page
And an interesting one: three times Cuda, one invalidated: 91840962
Grüße vom Sänger
All my CUDA workunits
)
All my CUDA workunits terminate in valide errors ( http://einsteinathome.org/host/3459688/tasks )
It's very frustrating.
RE: All my CUDA workunits
)
You're not using the option in a cc_config.xml are you?,
you're reporting your task before the server has found your uploaded result files,
like every hour 15 minutes,
Claggy
RE: RE: All my CUDA
)
I'm using this option, a relict from GPU-Grid crunching until last week, and I never experienced any problem here.
Grüße vom Sänger
RE: RE: RE: All my CUDA
)
I went through a spate of Validate Errors in December with my 9800GTX+, in my case it was probably to do with my slow/flaky internet connection,
some of the files needing several attempts before they'd upload, then even if you left some time before you reported them, they'd still get validate error,
i never got an answer wether there was a script that rechecks the validate errors prior to purging,
You're using a different version of Boinc to [AF>Libristes] Elgrande71, that can make all the difference,
the latest Development version of Boinc won't ask for work within 5 minutes of an upload to stop these thing happening: [trac]changeset:22841[/trac] and [trac]changeset:22910[/trac]
Claggy
Hi! My 9800GT had similar
)
Hi!
My 9800GT had similar symptoms, and after a reboot I got valid results again. There's no guarantee that this is the problem, but it's sure worth a try.
It still produces some validation failures because of the problem discussed in this thread, but at least now the majority of tasks go thru.
HB
RE: RE: All my CUDA
)
Exactly, I'm not using the option.
I am going to try it.
Thank you.
RE: RE: RE: All my CUDA
)
The point was you DON'T want to use it, you're reporting too soon after upload anyway, you should try and delay your reporting,
Claggy