Advanced search

Message boards : Frequently Asked Questions (FAQ) : Why do I have so many failed tasks?

Author Message
jmtc3man
Send message
Joined: 15 Apr 19
Posts: 3
Credit: 8,467,225
RAC: 0
Level
Ser
Scientific publications
wat
Message 51712 - Posted: 22 Apr 2019 | 4:11:19 UTC

https://imgur.com/a/31iI42D

I've been running BOINC for several days on a stable overclock. Why do so many of these tasks fail? What information do you all need so that we can figure this out?

Thanks!

Zalster
Avatar
Send message
Joined: 26 Feb 14
Posts: 211
Credit: 4,496,324,562
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwat
Message 51713 - Posted: 22 Apr 2019 | 4:48:33 UTC - in response to Message 51712.

How much of an overclocking are you doing? I would say start with the card at baseline first and see if the errors go away. If they do, then your overclocking was too severe for the project.
____________

mmonnin
Send message
Joined: 2 Jul 16
Posts: 337
Credit: 7,765,428,051
RAC: 1,425,398
Level
Tyr
Scientific publications
watwatwatwatwat
Message 51715 - Posted: 22 Apr 2019 | 11:45:12 UTC

Are others failing the same tasks?

Zalster
Avatar
Send message
Joined: 26 Feb 14
Posts: 211
Credit: 4,496,324,562
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwat
Message 51716 - Posted: 22 Apr 2019 | 14:39:44 UTC - in response to Message 51715.
Last modified: 22 Apr 2019 | 14:39:56 UTC

If you go to his machine and look at the tasks that errored, you will see they were resent to others and have been completed and validated. So it's his machine that has the issue.
____________

jmtc3man
Send message
Joined: 15 Apr 19
Posts: 3
Credit: 8,467,225
RAC: 0
Level
Ser
Scientific publications
wat
Message 51718 - Posted: 23 Apr 2019 | 20:30:12 UTC - in response to Message 51713.

On my EVGA 1070 FTW I can go up to
+125 clock speed
+425 memory clock speed
(power target 112%. I use EVGA Precision X, and I use KBOOST.)
which gives me 2088mhz clock 4233mhz memory clock

Without changing voltage. This is stable as far as graphics are concerned, but after having a long run finish without issue on the default settings, I guess calculations are a different case. I'll see if it continues to complete valid tasks on default settings, but how can I test for a stable overclock in this environment?
Thanks!

P.S. The card does not throttle. It hasn't gone above 68 C.

Zalster
Avatar
Send message
Joined: 26 Feb 14
Posts: 211
Credit: 4,496,324,562
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwat
Message 51719 - Posted: 23 Apr 2019 | 21:24:43 UTC - in response to Message 51718.

The only way to know is by testing at different values. I can give you a head start and say you could probably get away with +300 memory and +40 speed. I believe that would be a stable starting point. I'd run those for a day and see if they validate. If they do then you can try increasing the memory first by 50 by itself each day until you hit a point where you start to get errors. Then back down to the last stable. Then start increasing the speed by 20 until it either crashes or you get errors. This is going to take about a week before you find the sweet spot. There is no one configuration that works on all cards as we are dealing with the silicone lottery in the cards. Good luck.
____________

kksplace
Send message
Joined: 4 Mar 18
Posts: 53
Credit: 2,721,403,396
RAC: 3,083,124
Level
Phe
Scientific publications
wat
Message 51720 - Posted: 24 Apr 2019 | 0:05:24 UTC - in response to Message 51718.

For another input, I have an EVGA 1080 FTW. I have found a graphics speed more than 2062 MHz is not stable for GPUGrid, and at 2062 about 1 in 40 long tasks will error. 2050 is my sweet spot for that card (+65 for me). For some odd reason, I cannot go very far with memory speed before actually seeing a decrease in overall computation speeds; I only do +12.

jmtc3man
Send message
Joined: 15 Apr 19
Posts: 3
Credit: 8,467,225
RAC: 0
Level
Ser
Scientific publications
wat
Message 51721 - Posted: 24 Apr 2019 | 2:34:06 UTC - in response to Message 51720.
Last modified: 24 Apr 2019 | 2:34:19 UTC

Good to know that about the memory speed, I'll factor that into finding my sweet spot for this. Either way, the 1070 and 1080 cards are going to be really good at computing so the clock isn't super necessary, especially since I run BOINC 24/7, but that's the fun part!

Post to thread

Message boards : Frequently Asked Questions (FAQ) : Why do I have so many failed tasks?

//