Welcome to MilkyWay@home

What am I missing in order to get more GPU tasks?

Message boards : Number crunching : What am I missing in order to get more GPU tasks?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Ulrich Metzner
Avatar

Send message
Joined: 11 Apr 15
Posts: 58
Credit: 63,291,127
RAC: 0
Message 64650 - Posted: 14 Jun 2016, 22:27:26 UTC

Hmm, now i only have ~40 for both GPUs and nothing, null, niente, nada for the CPUs.
In a few moments my cruncher runs completely dry... :?
Aloha, Uli

ID: 64650 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Thunder
Avatar

Send message
Joined: 9 Jul 08
Posts: 85
Credit: 44,842,651
RAC: 0
Message 64651 - Posted: 14 Jun 2016, 23:51:53 UTC - in response to Message 64650.  

And I'm absolutely drowning in CPU tasks (47, which will take a bit over a full day to complete) and only got 24 GPU tasks (which were finished in a bit under 4 minutes).

It's a shame because I specifically built this machine to do MW@H work and now it spends ~80% of it's time working other projects (that actually have work available). :-/
ID: 64651 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Cliff
Avatar

Send message
Joined: 28 Nov 14
Posts: 51
Credit: 86,696,721
RAC: 0
Message 64652 - Posted: 15 Jun 2016, 2:09:33 UTC
Last modified: 15 Jun 2016, 2:10:08 UTC

This is totally daft!!
Until today I could complete 28 GPU tasks, upload the results and get a dozen or more WU in return.

Now I can get sweet FA, because the server is telling boinc to backoff for 55 mins plus. And My rig is therefor processing zilch from MW@H a
instead of processing an almost constant stream of WU.
Manual update is impossible due to the backoff. So far I'm loosing several hundred WU due to this shambles.
There was NO limit on the number of MW100 tasks allowed, now its max 13!!
ans the server stats page reports all ok. HaH, ok mu foot!

Maybe someone should just put the setting back as they were? And fiddle during working hours so any misconfiguration gets sorted out promptly.
Regards,
Cliff.
--
Been there Done That, still no Damn T-Shirt
ID: 64652 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Keith Myers
Avatar

Send message
Joined: 24 Jan 11
Posts: 696
Credit: 539,281,364
RAC: 95,190
Message 64653 - Posted: 15 Jun 2016, 5:57:16 UTC - in response to Message 64646.  

It appears to be deferring communication for 1 hour after an update, not 30 minutes.

(Admittedly this is just an anecdotal observation from one client machine)


I guess I was configured for the most beneficial outcome before server changes. I have run MW at a low share compared to my other projects so I must have downloaded the initial 150 tasks after the change. Plenty of work to get through with my low utilization. However, the other comments are correct. Server is only being contacted every hour, NOT every 30 minutes as Jake stated. Also, when I have contacted the server at the new update interval, I am getting no tasks. I have whittled down my initial 150 tasks to about 100 or so. So plenty of work to be done still though can see that after an extended outage and low replenishment schedule from the servers, I can see a situation where I run out of work.
ID: 64653 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ulrich Metzner
Avatar

Send message
Joined: 11 Apr 15
Posts: 58
Credit: 63,291,127
RAC: 0
Message 64656 - Posted: 15 Jun 2016, 8:53:24 UTC

I still only get GPU WUs on my main cruncher (Core2Quad), not a single CPU WU.
On the other hand my work laptop gets CPU work without problem - what's going on?! :?
Aloha, Uli

ID: 64656 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Vortac

Send message
Joined: 22 Apr 09
Posts: 95
Credit: 4,808,181,963
RAC: 0
Message 64657 - Posted: 15 Jun 2016, 9:07:46 UTC

Same problems here. Getting only about 10 workunits, finished in five minutes, then 55 minutes wait for the next batch.
ID: 64657 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
wb8ili

Send message
Joined: 18 Jul 10
Posts: 76
Credit: 635,998,708
RAC: 0
Message 64658 - Posted: 15 Jun 2016, 9:18:03 UTC

I am not sure what "problem" was being addressed by changing the server contact interval and number of work units that can be downloaded per server contact.

Add my vote to those who now get 5-10 minutes of GPU work every hour.

By the way, after I contact the server, the next contact time is 56 minutes in the future not 30 minutes. I think it was 5 minutes before. BACK TO 5 MINUTES!
ID: 64658 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Sebastian*

Send message
Joined: 8 Apr 09
Posts: 70
Credit: 11,027,167,827
RAC: 0
Message 64659 - Posted: 15 Jun 2016, 10:40:47 UTC
Last modified: 15 Jun 2016, 10:56:44 UTC

Hi, just got this message:

15/06/2016 12:13:18 | Milkyway@Home | Not sending work - last request too recent: 3296 sec

i would like to go back to the old 1 minute checkup time.

Can the project increase WU-generation? So there will be work for everyone?
ID: 64659 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Herge

Send message
Joined: 11 Oct 09
Posts: 19
Credit: 202,475,569
RAC: 0
Message 64661 - Posted: 15 Jun 2016, 11:18:17 UTC

The scheduler was fine with me. My issue is with the quota. At 60 units with 2 GPUs, I run out in about 30 minutes at each little hiccup. A day worth of units no matter what the configuration (CPU, GPU) would be nice.
ID: 64661 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ulrich Metzner
Avatar

Send message
Joined: 11 Apr 15
Posts: 58
Credit: 63,291,127
RAC: 0
Message 64663 - Posted: 15 Jun 2016, 11:58:27 UTC

Finally even my work laptop has come to a complete halt, because i have to wait (waste) another 35 minutes to get new WUs... :/
Aloha, Uli

ID: 64663 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jake Weiss
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 25 Feb 13
Posts: 580
Credit: 94,200,158
RAC: 0
Message 64665 - Posted: 15 Jun 2016, 13:00:53 UTC

Hey everyone,

It's all fixed now. Sorry when I changed the config file, I made a mistake that accidentally made the minimum polling time longer. All fixed to 30 minutes now for the polling frequency for work units. Let me know if this is still too long.

Jake
ID: 64665 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Keith Myers
Avatar

Send message
Joined: 24 Jan 11
Posts: 696
Credit: 539,281,364
RAC: 95,190
Message 64669 - Posted: 15 Jun 2016, 19:01:33 UTC - in response to Message 64665.  

Jake, I think you changed the config file to go back to maximum 40 GPU tasks per card too along with the correct 30 minute polling interval. I'm back at 80 GPU tasks per machine (two cards in each machine) just like it was before the attempted changes.
ID: 64669 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jake Weiss
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 25 Feb 13
Posts: 580
Credit: 94,200,158
RAC: 0
Message 64670 - Posted: 15 Jun 2016, 19:11:33 UTC

Okay. I went through the boinc wiki to figure out what is actually going on in this config file. I reverted it back to the way it was originally configured. The polling interval was 60 seconds not minutes. Therefore, you should be able to get work units much more frequently that once per hour. (I only left the total work units you can request slightly increased.) If you are having issues with only being able to get work units once per hour I would guess it is a client configuration problem.

Sorry for the confusion.

Jake W.
ID: 64670 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ulrich Metzner
Avatar

Send message
Joined: 11 Apr 15
Posts: 58
Credit: 63,291,127
RAC: 0
Message 64683 - Posted: 18 Jun 2016, 11:45:34 UTC

Just to inform:

Now everything is working smooth again - thank you!

40 per GPU and 8 per core: 40 * 2 + 8 * 4 = 112 WUs :)
Since i run 2 instances per GPU there are always 8 units in parallel.
Aloha, Uli

ID: 64683 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Alan K

Send message
Joined: 2 Nov 12
Posts: 2
Credit: 53,331,581
RAC: 18
Message 64840 - Posted: 7 Jul 2016, 22:37:33 UTC

I'm not getting any tasks for computer 566347. Is there a reason for this?
ID: 64840 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,769,034
RAC: 28,019
Message 64843 - Posted: 8 Jul 2016, 11:16:11 UTC - in response to Message 64840.  

I'm not getting any tasks for computer 566347. Is there a reason for this?


Are you crunching another gpu project on that pc?
ID: 64843 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jake Weiss
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 25 Feb 13
Posts: 580
Credit: 94,200,158
RAC: 0
Message 64844 - Posted: 8 Jul 2016, 13:00:30 UTC

Hi Alan K,

Looks like a 32-bit computer. We currently are unable to support 32-bit systems and as such you are not getting any work units.

Jake
ID: 64844 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Vortac

Send message
Joined: 22 Apr 09
Posts: 95
Credit: 4,808,181,963
RAC: 0
Message 64847 - Posted: 8 Jul 2016, 17:57:49 UTC

I have 4x7970 ready to deliver some serious TFLOPS but I am unable to get enough MW workunits to keep them crunching continuously. When requesting new tasks, I often get only 1-2 workunits, sometimes even 0 with message "Project has no tasks available". With 4x7970, I need at least 25 new workunits every minute for non-stop crunching. Is it possible to get that much during every update?
ID: 64847 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Alan K

Send message
Joined: 2 Nov 12
Posts: 2
Credit: 53,331,581
RAC: 18
Message 64848 - Posted: 8 Jul 2016, 18:24:06 UTC - in response to Message 64844.  

Thanks Jake. Might upgrade to 64bit - will probably need to if I want to go to Win10.
ID: 64848 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jake Weiss
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 25 Feb 13
Posts: 580
Credit: 94,200,158
RAC: 0
Message 64849 - Posted: 8 Jul 2016, 18:47:37 UTC

Vortac,

I am currently working on increasing the availability of workunits. It's not the easiest thing to do, but I think I am close to a solution.

Jake
ID: 64849 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : What am I missing in order to get more GPU tasks?

©2024 Astroinformatics Group