@Ananas & areeda:
"marked as invalid" is no validation error!!!
Greetings,
Gundolf
Of course it is. It went fine through the machine, but first got no consensus, and then got invalidated.
That sort of outcome - invalid, but without the word "error" visible on the webpage - belongs in the separate thread Inconclusive results, which the appropriate developers are also monitoring
I apologize but the only thing I can conclude from this is that I don't know which thread to post in.
It seems that a conclusive result of invalid is not a validation problem but an inconclusive result problem. I understand the developers are looking into specific problems and it's helpful to have examples of those problems collected together. I just don't understand how to do that.
May I request a short explanation of what these problems are for my own curiosity and a brief tutorial on how to examine which logs (and where to find them) so that we can better classify the messages. This thread is probably not the right place. But I had the same problem with "too many exit 0 messages"
I seem to be spending a lot of processing time (haven't calculated a percentage but I think it's in single digit) resulting in errors or invalid results. Some of these processes take 100 hrs on my slower machines.
Got a Validate error on task http://einsteinathome.org/task/218036973, WU http://einsteinathome.org/workunit/92434796. This is the first task on a new (to me) card, and the first Validate error I've encountered. I'm suspicious of the new card, though I have heard that Validate error is usually a server problem. There was about 2h 42m between uploading and reporting. Guidance is appreciated.
Has anyone noticed that these validation errors that appear lately are always between CUDA and non-CUDA tasks. Whenever I get an inconclusive result, I find that it was a CPU calculation of the same workunit that I processed with my GPU. If the third calculation is a CUDA task, it validates mine. If the third calculation is a CPU task, it validates the other CPU task and invalidates my CUDA task.
I believe that the results of some of the new workunits do not agree between the CPU version and the GPU version. It is a small percentage, but of the 5 invalid tasks that I see for my computer, the results are that two CPU tasks agree and my GPU task is invalid. For the valid tasks that were originally inconclusive, the result is that two GPU tasks agree and and the CPU task is marked invalid.
I'm not including cases where one of the original tasks reported an error, or timed out, only cases where the two tasks finished without errors but did not agree.
Has anyone noticed that these validation errors that appear lately are always between CUDA and non-CUDA tasks. Whenever I get an inconclusive result...
And again: inconclusive results are no validate errors as per this thread. Only if the "Status" column in the task/workunit display says "Validate error" it'a a validate error! ;-)
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
Okay, third Validate error from the same 9400GT card. I'm taking it off line. The disagreement between CUDA and CPU leading to invalid results is enough of a headache, I don't need to worry about Validate errors too. I can better use the CPU cycles wasted on these CUDA WUs. Oh, and if you've got a problem, don't post here. Apparently no one cares except Gundolf, and only if you *aren't* getting Validate errors.
Again, Do the Validate error tasks get checked again by the Validator prior to deleting and purging, like they do at Seti?
(If it's caused by a server side problem it's the least they should do)
RE: RE: RE: @Ananas &
I apologize but the only thing I can conclude from this is that I don't know which thread to post in.
It seems that a conclusive result of invalid is not a validation problem but an inconclusive result problem. I understand the developers are looking into specific problems and it's helpful to have examples of those problems collected together. I just don't understand how to do that.
May I request a short explanation of what these problems are for my own curiosity and a brief tutorial on how to examine which logs (and where to find them) so that we can better classify the messages. This thread is probably not the right place. But I had the same problem with "too many exit 0 messages"
I seem to be spending a lot of processing time (haven't calculated a percentage but I think it's in single digit) resulting in errors or invalid results. Some of these processes take 100 hrs on my slower machines.
Joe
Got a Validate error on task
Got a Validate error on task http://einsteinathome.org/task/218036973, WU http://einsteinathome.org/workunit/92434796. This is the first task on a new (to me) card, and the first Validate error I've encountered. I'm suspicious of the new card, though I have heard that Validate error is usually a server problem. There was about 2h 42m between uploading and reporting. Guidance is appreciated.
Has anyone noticed that these
Has anyone noticed that these validation errors that appear lately are always between CUDA and non-CUDA tasks. Whenever I get an inconclusive result, I find that it was a CPU calculation of the same workunit that I processed with my GPU. If the third calculation is a CUDA task, it validates mine. If the third calculation is a CPU task, it validates the other CPU task and invalidates my CUDA task.
I believe that the results of some of the new workunits do not agree between the CPU version and the GPU version. It is a small percentage, but of the 5 invalid tasks that I see for my computer, the results are that two CPU tasks agree and my GPU task is invalid. For the valid tasks that were originally inconclusive, the result is that two GPU tasks agree and and the CPU task is marked invalid.
I'm not including cases where one of the original tasks reported an error, or timed out, only cases where the two tasks finished without errors but did not agree.
Yes, please see
Yes, please see http://einsteinathome.org/node/195567
RE: Has anyone noticed that
And again: inconclusive results are no validate errors as per this thread. Only if the "Status" column in the task/workunit display says "Validate error" it'a a validate error! ;-)
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
Again I got a Validate error
Again I got a Validate error on a WU from the same card, http://einsteinathome.org/workunit/92510303.
Okay, third Validate error
Okay, third Validate error from the same 9400GT card. I'm taking it off line. The disagreement between CUDA and CPU leading to invalid results is enough of a headache, I don't need to worry about Validate errors too. I can better use the CPU cycles wasted on these CUDA WUs. Oh, and if you've got a problem, don't post here. Apparently no one cares except Gundolf, and only if you *aren't* getting Validate errors.
Again, Do the Validate error
Again, Do the Validate error tasks get checked again by the Validator prior to deleting and purging, like they do at Seti?
(If it's caused by a server side problem it's the least they should do)
Claggy
http://einstein.phys.uwm.edu/
http://einsteinathome.org/workunit/92132552
http://einsteinathome.org/workunit/92143096
http://einsteinathome.org/workunit/92143061
http://einsteinathome.org/workunit/92132552
--------------------------
- Crunch, Crunch, Crunch -
--------------------------
http://einstein.phys.uwm.edu/
http://einsteinathome.org/workunit/92850642
http://einsteinathome.org/workunit/92841954
--------------------------
- Crunch, Crunch, Crunch -
--------------------------