Validation inconclusive


log in

Advanced search

Message boards : Number crunching : Validation inconclusive

1 · 2 · 3 · Next
Author Message
Profile ccappel
Avatar
Send message
Joined: 27 Jan 00
Posts: 362
Credit: 1,516,412
RAC: 0
United States
Message 889864 - Posted: 30 Apr 2009, 15:07:02 UTC
Last modified: 30 Apr 2009, 15:15:38 UTC

I'm seeing an increase in frequency of these "validation inconclusive" messages. I hope this doesn't mean there's something going on with my computers:

http://setiathome.berkeley.edu/results.php?userid=652775&offset=0&show_names=0&state=2

edit:
http://setiathome.berkeley.edu/workunit.php?wuid=432051318
http://setiathome.berkeley.edu/workunit.php?wuid=418871556
http://setiathome.berkeley.edu/workunit.php?wuid=432050525
____________
"Life is a tragedy for those who feel, and a comedy for those who think."

"I never get into an argument that I cannot win."

Profile Labbie
Avatar
Send message
Joined: 19 Jun 06
Posts: 4083
Credit: 5,930,102
RAC: 0
United States
Message 889865 - Posted: 30 Apr 2009, 15:08:48 UTC

Your link doesn't work for the rest of us, but I know what you're talking about. I've been seeing more on mine too. I think a lot of it is the matching of CPU and CUDA as wingmen.

____________

Calm Chaos Forum...Join Calm Chaos Now

Profile ccappel
Avatar
Send message
Joined: 27 Jan 00
Posts: 362
Credit: 1,516,412
RAC: 0
United States
Message 889869 - Posted: 30 Apr 2009, 15:13:14 UTC - in response to Message 889865.

I edited the links, should work now.
I also should have mentioned theses are all AP, r112 SSE3 opt, 1 on each of 3 different computers.
____________
"Life is a tragedy for those who feel, and a comedy for those who think."

"I never get into an argument that I cannot win."

Profile Labbie
Avatar
Send message
Joined: 19 Jun 06
Posts: 4083
Credit: 5,930,102
RAC: 0
United States
Message 889872 - Posted: 30 Apr 2009, 15:20:09 UTC

Mine are all CPU MB, and usually tied to a CUDA wingman.

____________

Calm Chaos Forum...Join Calm Chaos Now

Profile elbea64
Send message
Joined: 16 Aug 99
Posts: 114
Credit: 6,352,198
RAC: 0
Germany
Message 889905 - Posted: 30 Apr 2009, 16:57:26 UTC

very interesting especially the second link as your wingman has a validation error and so there is nothing you could validate against.

i have some of them too
http://setiathome.berkeley.edu/workunit.php?wuid=428964969
http://setiathome.berkeley.edu/workunit.php?wuid=417924802
http://setiathome.berkeley.edu/workunit.php?wuid=428964969
http://setiathome.berkeley.edu/workunit.php?wuid=431231896
http://setiathome.berkeley.edu/workunit.php?wuid=418487995
http://setiathome.berkeley.edu/workunit.php?wuid=434521033
http://setiathome.berkeley.edu/workunit.php?wuid=425639304
http://setiathome.berkeley.edu/workunit.php?wuid=429069735

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3500
Credit: 47,704,570
RAC: 46,448
Russia
Message 889918 - Posted: 30 Apr 2009, 17:58:31 UTC - in response to Message 889905.

There is pretty big frequency range where shader will do work incorrectly but w/o any error reporting. GPU OCed to this freq area will produce lot of invalid results (validation incomplete first, then weakly similar or failed to validate with third wingman).
Unfortunately it's hard to notice unless user will browse his host results. System will appear stable w/o any visual signs of excessive OCing...

Profile elbea64
Send message
Joined: 16 Aug 99
Posts: 114
Credit: 6,352,198
RAC: 0
Germany
Message 889977 - Posted: 30 Apr 2009, 21:26:50 UTC - in response to Message 889918.

There is pretty big frequency range where shader will do work incorrectly but w/o any error reporting. GPU OCed to this freq area will produce lot of invalid results (validation incomplete first, then weakly similar or failed to validate with third wingman).
Unfortunately it's hard to notice unless user will browse his host results. System will appear stable w/o any visual signs of excessive OCing...


The linked WUs of ccappel and me are all AP

john deneer
Volunteer tester
Avatar
Send message
Joined: 16 Nov 06
Posts: 331
Credit: 20,996,195
RAC: 0
Netherlands
Message 889986 - Posted: 30 Apr 2009, 22:08:18 UTC - in response to Message 889864.

I'm seeing an increase in frequency of these "validation inconclusive" messages. I hope this doesn't mean there's something going on with my computers:

I've had a couple of AP units with this message over the last few days as well. No idea what's causing it.

Regards,
John.

Zydor
Send message
Joined: 4 Oct 03
Posts: 172
Credit: 491,111
RAC: 0
United Kingdom
Message 890106 - Posted: 1 May 2009, 8:28:29 UTC - in response to Message 889986.

I got one of those a couple of days ago. I had a look at the Task ID for those with similar errors, all have the same symptom. They appear to be from markedly slower machines, and all have the stderr message (or similar) several times over:

In ap_gfx_main.cpp: in ap_graphics_init(): Starting client.
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 2688
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 2816
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 2944
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3072
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3200
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3328
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3456

No idea what that actually indicates on those machines - anyone know?

Regards
Zy

____________

WinterKnight
Volunteer tester
Send message
Joined: 18 May 99
Posts: 8685
Credit: 24,995,808
RAC: 29,495
United Kingdom
Message 890108 - Posted: 1 May 2009, 8:34:14 UTC - in response to Message 890106.

I got one of those a couple of days ago. I had a look at the Task ID for those with similar errors, all have the same symptom. They appear to be from markedly slower machines, and all have the stderr message (or similar) several times over:

In ap_gfx_main.cpp: in ap_graphics_init(): Starting client.
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 2688
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 2816
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 2944
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3072
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3200
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3328
In ap_client_main.cpp: in mainloop(): at dm_chunk_large 3456

No idea what that actually indicates on those machines - anyone know?

Regards
Zy

No idea what these msg's mean but they appear in the stderr for all tasks completed with the default Berkeley app. Here is task 1187117097 that has validated, against mine using the optimised app.

Profile Byron S Goodgame
Volunteer tester
Avatar
Send message
Joined: 16 Jan 06
Posts: 1151
Credit: 3,936,993
RAC: 0
United States
Message 890109 - Posted: 1 May 2009, 8:40:20 UTC - in response to Message 890108.
Last modified: 1 May 2009, 8:45:30 UTC

No idea what these msg's mean but they appear in the stderr for all tasks completed with the default Berkeley app. Here is task 1187117097 that has validated, against mine using the optimised app.

My guess was they refer to the checkpoints made while the app is running. Guess now is good time as any to test that theory.
____________

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3500
Credit: 47,704,570
RAC: 46,448
Russia
Message 890110 - Posted: 1 May 2009, 8:46:49 UTC - in response to Message 890109.

It's debug message indicating that new iteration over large DM chunk begins.

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,932,256
RAC: 11,953
United States
Message 890588 - Posted: 2 May 2009, 16:53:00 UTC

How about this one? http://setiathome.berkeley.edu/workunit.php?wuid=433362764

Four of us have completed it. two on Cuda and two on CPU. It has gone out again today to another cruncher. Wonder what is wrong with it?
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3500
Credit: 47,704,570
RAC: 46,448
Russia
Message 890893 - Posted: 3 May 2009, 18:14:40 UTC - in response to Message 890588.

How about this one? http://setiathome.berkeley.edu/workunit.php?wuid=433362764

Four of us have completed it. two on Cuda and two on CPU. It has gone out again today to another cruncher. Wonder what is wrong with it?

well, it's MB result so my former post fully applies to it.
Weakly similar most probably due to excessive GPU OCing or overheating.
When two strongly similar results were recived all (including weakly similar) recived credit.
I consider it as very dangerous practice - to give credit for weakly similar results cause such not stable hosts actually counterproductive to project - their results needed additional verification - waste of processing time and server bandwith.

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,932,256
RAC: 11,953
United States
Message 890904 - Posted: 3 May 2009, 19:00:36 UTC - in response to Message 890893.

Hmmm, just noticed something else now. It looks like they averaged them instead of giving the lowest credit. That's strange.

I have been getting a few of the inconclusive. I'm not OCd and most of my work is coming in valid. Wonder if it is just luck of the draw or something of mine going bad? I'll keep an eye on my pendings to see if many more of them go inconclusive.
____________


PROUD MEMBER OF Team Starfire World BOINC

Alinator
Volunteer tester
Send message
Joined: 19 Apr 05
Posts: 4178
Credit: 4,647,982
RAC: 0
United States
Message 890932 - Posted: 3 May 2009, 20:41:04 UTC - in response to Message 890904.
Last modified: 3 May 2009, 20:46:17 UTC

Nothing strange there, it did exactly what it should have.

The BOINC scoring rules for WU with 3 valid tasks is to throw out the high and low claims and grant the middle one. For 4 or more it throws out the high and low and grants the average of the remaining ones.

I guess it's just been a long time since most people have seen a WU with 4 or more valid tasks in play. ;-)

<edit> BTW, with what looks to be two CUDA hosts in the mix, you ended up getting a little 'premium' paid on the WU over what it would have been if paid for straight CPU tasks. :-)

Alinator

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,932,256
RAC: 11,953
United States
Message 890965 - Posted: 3 May 2009, 21:55:34 UTC - in response to Message 890932.

You're right, it has been a long time since I've seen 4 or more valid results. Oh, and I'm one of those Cudas.
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3500
Credit: 47,704,570
RAC: 46,448
Russia
Message 891069 - Posted: 4 May 2009, 3:29:52 UTC - in response to Message 890904.

Hmmm, just noticed something else now. It looks like they averaged them instead of giving the lowest credit. That's strange.

I have been getting a few of the inconclusive. I'm not OCd and most of my work is coming in valid. Wonder if it is just luck of the draw or something of mine going bad? I'll keep an eye on my pendings to see if many more of them go inconclusive.


Cause I OCing my GPU I trying to look on results regularly - now 1 inconclusive per 2 pages of results approx.
All that I noticed in "definitive" state was failures of wingmans.
Sometimes I meet CUDA wingman that totally broken. Unfortunately, current "quota" politics allows such host continue to trash results always w/o reducing its quota (so user can just not know what going on with his GPU and only wonder why RAC is so low).

StickProject donor
Volunteer tester
Send message
Joined: 26 Feb 00
Posts: 85
Credit: 1,720,015
RAC: 674
United States
Message 891178 - Posted: 4 May 2009, 14:29:25 UTC

http://setiathome.berkeley.edu/workunit.php?wuid=435875813

The above AP workunit now has 3 "inconclusive" results (mine was the 3rd one). Anyone care to bet on whether the 4th will settle anything?
____________

Profile elbea64
Send message
Joined: 16 Aug 99
Posts: 114
Credit: 6,352,198
RAC: 0
Germany
Message 891192 - Posted: 4 May 2009, 14:59:21 UTC

I think it's very interesting as it looks like there are two stock apps that don't validate against each other.

1 · 2 · 3 · Next

Message boards : Number crunching : Validation inconclusive

Copyright © 2014 University of California