Questions and Answers :
GPU applications :
9 Cuda tasks waiting to run, why is that?
Message board moderation
Author | Message |
---|---|
AspirationTower Send message Joined: 4 Jan 06 Posts: 97 Credit: 946,069 RAC: 0 |
I have 9 cuda tasks waiting to run whilst 2 others are running on the cards I have. Firstly why is this happening and secondly will they ever complete? Resistence is futile! {evil cackle} |
Chuck Gorish Send message Joined: 19 Jun 00 Posts: 156 Credit: 29,589,106 RAC: 0 |
I have 9 cuda tasks waiting to run whilst 2 others are running on the cards I have. Firstly why is this happening and secondly will they ever complete? unfortunately i don't have an answer. i get the same thing but mine do eventually get to run so i assume yours will too. still trying to understand boinc's logic when it pauses a workunit to do another one that has a much later due date. at least they all get done. i don't know the version but i heard there is a version of boinc that abandons workunits and doesn't get back to them. if you are running a version much earlier than 6.6.36 maybe try a newer version to be sure you don't have that one with problems. |
AspirationTower Send message Joined: 4 Jan 06 Posts: 97 Credit: 946,069 RAC: 0 |
I'm running the latest version of Boinc 6.6.36 but at least it's good to know that they will eventually be completed. I have also never understood the logic of Boinc and the way it does the tasks. Date order would have made more sense to me and if something didn't get completed by for example a cruncher that left the project then that task should then be reissued with the old date on so that it would then be done first by anyone else that got that particular task next. Still programmers have never made much sense to me especially the way programs grow and grow rather than becoming more streamlined and quicker. Best example of all is Windows itself. The programmers seem to like adding things to it just for the sake of saying there is something new about it. Anyway thanks Chuck for the reasurance that the jobs will be done I guess I'll have to keep my eye on it and I'll mark this thread as dealt with. Resistence is futile! {evil cackle} |
Nick Send message Joined: 17 May 99 Posts: 96 Credit: 17,356,094 RAC: 0 |
Don't blame the programers for the bloat, blame marketing. BTW, I have the same problem with lots of suspended tasks. Makes no sense to me either. |
Nick Send message Joined: 17 May 99 Posts: 96 Credit: 17,356,094 RAC: 0 |
Vampire - Are you running an optimized version of CUDA? |
The Weasel Send message Joined: 6 Jun 99 Posts: 127 Credit: 53,205,208 RAC: 0 |
I had the same probaly, try changing your settings from, switch tasks every 60min(default) to a higher value and most of mine went away. I went to 90min then 120 and have not had a problem since with a suspended units. |
Nick Send message Joined: 17 May 99 Posts: 96 Credit: 17,356,094 RAC: 0 |
Is there a document somewhere that explains what all these switches actually do? |
Tom Send message Joined: 29 Oct 01 Posts: 53 Credit: 4,615,009 RAC: 0 |
I'm a bit of a noob when it comes down to CUDA. But my best guess is, since crunching with CUDA, is so fast, SETI make sure to deliever enough of packages so you don't run dry of packages. Hope anyone have more knowledge about that, and a more qualified answer. |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
I'm a bit of a noob when it comes down to CUDA. But my best guess is, since crunching with CUDA, is so fast, SETI make sure to deliever enough of packages so you don't run dry of packages. Hope anyone have more knowledge about that, and a more qualified answer. You resurrected a quite old task :-) The problem was not the tasks being "ready to start" but "waiting to run" since they had been started after downloading, just to be preempted again due to a bug in that BOINC version. The newest BOINC versions don't have this problem any longer. Gruß, Gundolf Computer sind nicht alles im Leben. (Kleiner Scherz) SETI@home classic workunits 3,758 SETI@home classic CPU time 66,520 hours |
Tom Send message Joined: 29 Oct 01 Posts: 53 Credit: 4,615,009 RAC: 0 |
ok... |
Fred J. Verster Send message Joined: 21 Apr 04 Posts: 3252 Credit: 31,903,643 RAC: 0 |
Hi, 2 WU's running on your CUDA cards and none on your CPU, did you set it that way in your accounts preferences? I had the same problem, try changing your settings from, switch tasks every 60min(default) to a higher value and most of mine went away. I went to 90min then 120 and have not had a problem since with a suspended units. When you have a lot of WU's that take f.i. 85 minutes, it could be worth to change this setting to 90 minutes. With CUDA, it's a different story, since times to complete an MB WU, has dropped, sometimes by 10x, which, ofcoarse depends on the CUDA card used. And this makes it almost impossible to say, what the best setting is. Only way to find out, is to try this. And avoid, IMHO, too short settings, cause the switching, also takes time and CPU-resources. And can cause errors. [color=darkblue]Does the ERROR occur, during the switching? Could be memory - problems. Hard to tell, since taskview is still disabled. When you pause your task's, by clicking right-click the BOINC icon in the taskbar, you'll see, all(SETI-CUDA-MB) WU's are 'branded' 0.0x CPU-CUDA. I see this with BOINC 6.6.36, 32&64BIT and 6.6.38. This 'way' you can REbrand MB WU's to CPU-handling (6.03) |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.