Advanced search

Message boards : Number crunching : WU disappeared

Author Message
Profile JohnMD
Avatar
Send message
Joined: 4 Dec 10
Posts: 5
Credit: 3,672,356
RAC: 0
Level
Ala
Scientific publications
watwatwat
Message 57649 - Posted: 25 Oct 2021 | 22:52:06 UTC

Today a WU which had run 11 days on my i5-8250U/MX150 disappeared. I can find NO record - neither on the host nor on the user (mig). Unfortunately I have no record of WU or Task number.
How does GPUGrid remove history from a contributor ?

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2335
Credit: 16,178,080,749
RAC: 0
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 57650 - Posted: 25 Oct 2021 | 23:03:47 UTC - in response to Message 57649.

The deadline for GPUGrid workunits is 5 days. You won't receive any credit if you return the result 6 days after the deadline. Do not crunch (long) GPUGrid tasks on an MX150, as you burn electricity (and your GPU in your laptop) in vain. Crunching on laptops are not recommended anyway.
The batch which this workunit belongs to has been completed, hence the records belong to its workunits purged to shrink the database. The real problem here is that there are workunits from 2014 which are still present in the database somehow.

Ian&Steve C.
Avatar
Send message
Joined: 21 Feb 20
Posts: 744
Credit: 4,943,693,494
RAC: 515,366
Level
Arg
Scientific publications
wat
Message 57651 - Posted: 26 Oct 2021 | 0:00:46 UTC - in response to Message 57650.

The deadline for GPUGrid workunits is 5 days. You won't receive any credit if you return the result 6 days after the deadline.


this isn't necessarily true all the time. it depends totally on what happens with resends. if he returns a result first, he will still get credit, no matter how long it has been since deadline. though the chances of a resend not being returned yet are increasingly slim the longer past deadline you go. but very possible if the resends land in the hands of someone else that:

-also has a slow GPU
-turns their system off intermittently
-combination of the previous two
-turns their system off indefinitely
-experiences some system issue preventing computation
-etc

have this scenario happen 2-3x and you're now in the range of 15+ days since the first task was issued with no result returned. I've seen this exact scenario not only at other projects, but also here at GPUGRID.
____________

jjch
Send message
Joined: 10 Nov 13
Posts: 91
Credit: 15,040,000,871
RAC: 1,015,809
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 57653 - Posted: 26 Oct 2021 | 6:02:25 UTC

There could be a number of things wrong here.

JohnMD appears to have one PC currently connected to GPUGrid. http://www.gpugrid.net/show_host_detail.php?hostid=518959

I don't see any tasks listed here or on the app details page
http://www.gpugrid.net/host_app_versions.php?hostid=518959

The all tasks also doesn't show anything.
http://www.gpugrid.net/results.php?hostid=518959

This system is reporting in to GPUGrid and the Last contact info is current but I don't think it is fully connected properly.

First thing I would note is the BOINC version 7.16.11 is old and it should be updated to version 7.16.20.

Unless he manually updated the certificate file earlier that could be causing the client not to correctly communicate with GPUGrid.

Another oddity is that the total credit shows 1,168,395. Somewhere along the way he completed at least a couple of tasks based on the current credit range.

If you go back to the Application details page you can see a number of tasks that were completed with older ACEMD versions.

I'm wondering if this BOINC client may have been copied from another earlier host or there may be a duplicate host that needs to be merged.

Another possibility is there could be a split or duplicate CPID but I can't remember how to check that.

A few other thoughts. The NVIDIA driver version 457.30 is somewhat old and I would recommend updating that to version 471.11 or newer.

Unfortunately I would also agree that the MX150 GPU really doesn't have enough horsepower to compute GPUGrid tasks within the deadline timeframe anymore.

If the problem is resolved the GPU could potentially continue to contribute as much as possible but I would hate for it to get cooked and fail.

I have a number of old Nvidia Grid K2 cards that I would not use for GPUGrid but they are happy plugging along and computing for Milkyway@home.

If JohnMD is still interested in computing for medical science I would suggest World Community Grid. They also have some limited GPU work at times.

Another option is Rosetta@home but I haven't been running that one lately. Just decided to run WCG for now.



Post to thread

Message boards : Number crunching : WU disappeared

//