Validate errors [pls post here] [CLOSED]

Because, the validator was called to examine a quorum (two successfully completed sets of answers for a given workunit) but it found that either (a) there was no agreement or (b) one of the answers could immediately be ruled out from further examination because it couldn't possibly be correct or perhaps was missing or corrupt in some way. If you take a good look at the two examples you linked, you will see an example of each of these two separate cases.

These two cases can be summarised as:-

• * These two answers are structurally OK but they don't agree closely enough with each other so both are marked inconclusive until an extra answer from an extra host can be returned and used as a decider.

* These two answers don't even need to be checked. One of them is either corrupt, missing or clearly wrong since its contents don't conform with the expected structure that it should have. The label is 'inconclusive' for the OK task and 'validate error' for the presumably missing, corrupt or otherwise unacceptable task. The 'inconclusive' result will remain as a 'pending' until such time as an extra result is returned and a proper comparison of results is made. The 'validate error' task can be rejected immediately.

This is all fine and dandy for weeding out corrupt results coming from dodgy clients (eg excessive overclocking) but is a bit tough on the participant if the 'validate error' is actually a symptom of a server-side failure - as it has been in a number of instances. The purpose of this special thread (why it was made sticky in the first place) is to provide a "low noise" mechanism for contributing examples of "potential server-side" validate errors back to the Devs so that what caused the error (like bugs in the validator itself) can be found and fixed. As an example of a bug (now fixed) take a look at the two posts immediately prior to yours.

Your two 'inconclusives' are NOT 'validate errors'. They are really no different to any of your other 'pendings'. They are just waiting for a further result to be returned so that the validator can check for agreement (or otherwise). There is nothing inherently wrong with having the 'inconclusive' tag on your result. If you aren't sure about something it's usually best to start your own thread unless you are really certain that it's related to an existing thread.

No one seems to be looking here, and posting under the non-sticky threads does not seem to work either, but here goes another try. See the message above for some details. What happens now is that one third to one half of the CUDA work units fail immediately upon uploading with an error code of Validate error (2:00000010)

Those units that pass the initial validation and classified as successful later fail validation but these post an error of Validate Error (8:00001000).

In one comment I finally got in the non-sticky threads, Bernd indicated that the error 2:00000010 indicated some sort of calculation error. No one answered with the 8:00001000 is trying to indicate.

As noted in the earlier post, this suddenly started happening after many weeks of 100 % success on CUDA tasks. The video card seems to be working okay as a video card, but is there a chance there is some failed memory or something that would affect CUDA tasks but not show up in normal display ?

I have recently upgraded to Ubuntu 11.04, but the problem still occurs in the same way. The upgrade to Ubuntu 11.04 upgrades the Nvidia driver to the most recent stable one (which is more recent than the version running under 10.4), and I do have the error that a lot of prople have where it says the driver is activated but not in use (but I have a working display, so it DOES seem to be in use ...).

Any comments would be most welcome.

A couple of workunits to study would be

Name PM0119_01811.dm_64_0
Workunit 98091109
Created 26 May 2011 2:11:15 UTC
Sent 26 May 2011 11:29:38 UTC
Received 28 May 2011 18:27:29 UTC
Server state Over
Outcome Validate error (2:00000010)

and

Name PM0119_01641.dm_348_1
Workunit 98088126
Created 26 May 2011 1:36:45 UTC
Sent 26 May 2011 9:47:49 UTC
Received 28 May 2011 4:05:10 UTC
Server state Over
Outcome Validate error (8:00001000)

Regards
Richard

two work units have stalled out
Both report "waiting to run (waiting for GPU memory)"

WU# 103876166
WU# 103845176

Not sure why the message
I don't have a compute capable GPU for Boinc programs to use
I did recently upgrade Boinc to version 6.12.33 and have noticed
since the change my results seem to be much more problematic with validation errors happening regularly
particularly with the Gamma Ray Pulsar search #1 v0.22

I had a newest BOINC version - 6.12.34; WindowsXP SP3; nVidia GTX470 video card and 270.61 version of graphic drives. Workunits S6Bucket and FGRP1 successfully finish width claimed credit but all of BRP4 workunits computed on my GTX470 (successfully compute and send to server) finally had status - Validate Error.
Why ?
My graphic card is good and no errors reported in BOINC. What's wrong ?
If the problem isn't resolve I don't want to compute with my video card. My computer work for this workunits but finally nothing points are claimed. And this work are senseless.