| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Invalid GPU work units |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 171
|
|
| Author |
|
|
widdershins
Veteran Cruncher Scotland Joined: Apr 30, 2007 Post Count: 673 Status: Offline Project Badges: 
|
I've had a few units where one or two others have received the same unit and they've been marked as invalid whilst mine has turned valid. Looking at their logs there's nothing jumping out as to why. Some had newer BOINC versions, some were newer and more powerful NVIDIA cards and some older or similar vintage as my Titan card.
I'm thinking something like overclocking or bad memory would show as an error, but can't think what would cause the invalid. Anyone have any suggestions what might cause these to turn invalid (as opposed to error) and then have my machine sent a copy to process? Not concerned for myself as the more invalids others have, the more GPU resends there are for me to snag.  But it'd perhaps help others to not waste GPU power if the reason was known. |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 732 Status: Offline Project Badges: 
|
As it happens, it has already been discussed in the Work Unit Availability thread; Uplinger replied about it here and if you track back a few posts you'll see some of the other reports...
There are also folks reporting errors in other threads (including the one in the GPU Support Forum) but the above is the only place I've seen an explanation so far... Cheers - Al. |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
Good afternoon,
Yes, there seems to be a group of ligands that cause the calculations to be incorrect which is actually scientifically valid. I am in discussions with the researchers to see what they would like for us to do in handling these types of results. The end goal would be to allow for these to be marked as valid but the researchers would be able to easily tell it was a valid failed test. Thanks, -Uplinger |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
Sorry, just read your post again. I missed the point that you said others were invalid but yours was valid. There are some graphic cards that produce bad results and fail validation. As to why these cards are not processing things well, I do not have a specific answer to that. I can speculate, but without access to a card causing the issues, it is an unknown to me.
Thanks, -Uplinger |
||
|
|
maeax
Advanced Cruncher Joined: May 2, 2007 Post Count: 133 Status: Offline Project Badges:
|
Three invalid one successful and one Server aborted (for me)
----------------------------------------https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619758428
AMD Ryzen 2700 / AMD Radeon (TM) Pro WX7100. OS Win10pro
----------------------------------------[Edit 1 times, last edit by maeax at Apr 14, 2021 5:42:01 PM] |
||
|
|
Andrew80431
Cruncher Joined: Nov 25, 2005 Post Count: 36 Status: Offline Project Badges: 
|
Sorry, just read your post again. I missed the point that you said others were invalid but yours was valid. There are some graphic cards that produce bad results and fail validation. As to why these cards are not processing things well, I do not have a specific answer to that. I can speculate, but without access to a card causing the issues, it is an unknown to me. Thanks, -Uplinger I had 4 invalid WUs on my card today. (I noticed for the first time ever since the beta had started.) The first two came out wrong on all machines who did the WU. OPNG_0002425_00177 OPNG_0002337_00077 OPNG_0002337_00282 OPNG_0002337_00291  [Edit 1 times, last edit by Andrew80431 at Apr 21, 2021 6:43:52 AM] |
||
|
|
widdershins
Veteran Cruncher Scotland Joined: Apr 30, 2007 Post Count: 673 Status: Offline Project Badges:
|
Sorry, just read your post again. I missed the point that you said others were invalid but yours was valid. There are some graphic cards that produce bad results and fail validation. As to why these cards are not processing things well, I do not have a specific answer to that. I can speculate, but without access to a card causing the issues, it is an unknown to me. Thanks, -Uplinger I was all set to accept that it may be individual cards with issues, except... I recently got a WU on the same card where my result was the invalid and two others then turned valid. the WU is OPNG_0002357_00237 That would suggest that it's not individual bad cards but some instability in the application which is varying the results. Which then begs the question, if a few results are inaccurate enough to be classed as invalid are there others that are passing validation that are less accurate than they could/should be? [Edit 1 times, last edit by widdershins at Apr 14, 2021 6:40:32 PM] |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
The researchers have identified an issue with some of the work units that were generated. I am working to prevent these from sending out additional copies to the volunteers. This causes an issue on both OPNG and OPN1.
I will post more in a bit, working on cleanup at the moment. Thanks, -Uplinger |
||
|
|
Grumpy Swede
Master Cruncher Svíþjóð Joined: Apr 10, 2020 Post Count: 1868 Status: Recently Active Project Badges: 
|
Well, there are examples obviously of not all results of a WU turning invalid. Some becoming valid, and som invalid. Strange creatures these problematic WU's
----------------------------------------Here's one that I just crunched. My result is valid, so is another one, but one so far is invalid: https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=620029462 Edit: It became 3 valid, and one invalid. Extremely short runtime, and therefore very low credit.   [Edit 4 times, last edit by Grumpy Swede at Apr 14, 2021 9:45:27 PM] |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
Keith I have all sorts of tasks marked invalid on all 4 machines. If you need to look at ones that have 4 or 5 copies with none valid or 1 valid I have them and I can post the log if needed.
----------------------------------------
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
|