Message boards : Graphics cards (GPUs) : Any changes to project since March30 ?
Author | Message |
---|---|
I may only have a 9600gso, but up until ~ March 30, that was good enough to complete ~ 1 wu /day. | |
ID: 8096 | Rating: 0 | rate: / Reply Quote | |
NOt sure,, I updated the onee system to 6.6.20 and now have elapsed time showing and some of my tasks on the GTX 295 are showing 16 to 21 hours and they have 2-3 hours to go ... | |
ID: 8103 | Rating: 0 | rate: / Reply Quote | |
I've got a couple on my GTX260 cards that have been running all day and they haven't finished yet. BOINC is saying its elapsed time is 18 hours 47 mins. They normally take 6 hours. | |
ID: 8108 | Rating: 0 | rate: / Reply Quote | |
There is nothing new on the workunits. | |
ID: 8109 | Rating: 0 | rate: / Reply Quote | |
This was my first ... now the internal number says 6 hours, the elapsed time said over 20 ... sigh ... now I gotta try to figure out if it is an accounting bug in BOINC 6.2.20 or a serious issue. I have two tasks now listed as taking 25 plus hours ... and a third at 18:47 with 5:48 to go ... | |
ID: 8112 | Rating: 0 | rate: / Reply Quote | |
This was my first ... now the internal number says 6 hours, the elapsed time said over 20 ... sigh ... now I gotta try to figure out if it is an accounting bug in BOINC 6.2.20 or a serious issue. I have two tasks now listed as taking 25 plus hours ... and a third at 18:47 with 5:48 to go ... Well I know mine have been going all day because network times prevent internet access after 6:55am. It hasn't completed any other GPUgrid tasks, normally it would have 4 or 6 done by now. I haven't updated my video card drivers (182.06) since I installed it 2 weeks ago and I haven't upgraded BOINC for a week (its 6.6.17). I'll try shutting BOINC down in the first instance and see if the "to completion" time starts to move. If not will try a reboot and see if that helps. ____________ BOINC blog | |
ID: 8115 | Rating: 0 | rate: / Reply Quote | |
I'll try shutting BOINC down in the first instance and see if the "to completion" time starts to move. If not will try a reboot and see if that helps. Restarting BOINC seems to have done the trick. Not sure if this is an app issue or a BOINC issue. Either way its moving fairly quickly now. ____________ BOINC blog | |
ID: 8116 | Rating: 0 | rate: / Reply Quote | |
I've got a couple on my GTX260 cards that have been running all day and they haven't finished yet. BOINC is saying its elapsed time is 18 hours 47 mins. They normally take 6 hours. Result uploaded and reported now. Interestingly it thinks its elapsed time is approx 23000 seconds (6.5 hours), when in fact it took over 20 hours. ____________ BOINC blog | |
ID: 8117 | Rating: 0 | rate: / Reply Quote | |
I usually complete one wu per 24 hours. | |
ID: 8123 | Rating: 0 | rate: / Reply Quote | |
Its an infection I tell you ... | |
ID: 8126 | Rating: 0 | rate: / Reply Quote | |
Oh dear.. too bad I can't watch my card closely any more, as it has been replaced by an irresistible offer from the red team and has moved to a *remote* machine now. | |
ID: 8145 | Rating: 0 | rate: / Reply Quote | |
Well, that is too bad, but something is going on. I have 4 tasks in flight (as usual) two of which are running at 6 hours each and I have two listed at 12 (roughly) and they are only half way done. | |
ID: 8156 | Rating: 0 | rate: / Reply Quote | |
Is the time per step showing anything unusual? | |
ID: 8159 | Rating: 0 | rate: / Reply Quote | |
I'm gonna be switching to F@H for the time being. | |
ID: 8169 | Rating: 0 | rate: / Reply Quote | |
I'm gonna be switching to F@H for the time being. I must admit I have been seeing a few "cancelled by server" work units and they are usually ones downloaded within the last 24 hours, so its not as if they'd reached the deadline. Here is one. Another And a 3rd They are on the machine with dual GTX260's so they normally get returned within 24 hours. ____________ BOINC blog | |
ID: 8170 | Rating: 0 | rate: / Reply Quote | |
The WUs are cancelled by the server because they're redundant (see here). They only get canceled if the result is not needed any more any if you didn't already start them. That's much better than wasting computation time, isn't it? | |
ID: 8176 | Rating: 0 | rate: / Reply Quote | |
Is the time per step showing anything unusual? No... 35 MS per step ... I have the two latest coming up for turn in ... I will check those also ... Since BOINC does not tell me which GPU it is running on, it is hard to know, but one of the tasks that is long running has the end tag "Ignasi", and history shows that many of those are "expirimental" and may cause side effects. So, what I was wondering if the one odd ball is slowing down the others. {edit} The current odd ball is the third of a triplet issued... the other two have returned the result already so I wonder if I am going to get cheated out on this one ... | |
ID: 8178 | Rating: 0 | rate: / Reply Quote | |
Ok, my "bad" task just completed and on THIS one the time step sucked at 159 ms, and the time reflects the long run time at 79,645 seconds (21 hours) ... | |
ID: 8182 | Rating: 0 | rate: / Reply Quote | |
Could you downgrade BOINC/driver to check what's the problem? | |
ID: 8183 | Rating: 0 | rate: / Reply Quote | |
Could you downgrade BOINC/driver to check what's the problem? Random in the sense that I can't find a common factor EXCEPT that while I am running slow tasks, one or more of those tasks end in "ignasi" which is why I brought that up. Usually when iggy makes new tasks he has the decency to put his name on them. What is puzzling is that not only the iggy task is slow, but a companion task is slow ... if two iggys are running at the same time, I had 4 slow tasks ... Now I have NO iggy tasks on that machine and the 4 inflight seem to be on track for 6 hour run times. The oddest thing is that of all the tasks that seemed "troubled" only ONE of them seems to have recorded a bad time (150 MS plus step time) ... If you look back at the last couple days production you can see I went from 15-16 per day to about 8 ... but when I llok at the tasks I can't see anything that out of place. Which is why I also asked the question about trusting the internal run time number ... I think I listed tasks below where I *KNOW* the run time listed on the machine as 20+ hours and yet the run time logged was no where near that ... Well, i will keep watching ... but I think you do have a problem ... and a task issued 3 times? Gotta be something going on there ... Oh, at least the tasks complete successfully and validate, for what that is worth ... | |
ID: 8188 | Rating: 0 | rate: / Reply Quote | |
MarkJ reported a hanging task similar to what you see / have seen and it also happened on his box with 2 GPUs.. could mean something. | |
ID: 8192 | Rating: 0 | rate: / Reply Quote | |
@ Paul: Did you restart BOINC? | |
ID: 8194 | Rating: 0 | rate: / Reply Quote | |
@ Paul: Did you restart BOINC? I did a restart, though not a cold start, between "sets" of these odd tasks. But the second pair just completed this afternoon. With NO change to the system, and no restart, I have 4 in flight with all of them on track to complete in the normal 6 hours, for example the one at 25% done is 1:33 hours with 7 hours to run ... Oh, now that is interesting ... one GPU task is in "Waiting" ... now why is that ... Even more interesting is the fact that the task suspended has a deadline sooner than the task that is running ... Hmmm, two tasks where the time to complete is INCREASING 3 seconds for each runtime second ... Ok, I just shot two of those tasks and down leveled to 6.5.0 to see what will happen (also did a re-boot) ... {edit, add} Anyone else having this issue, would you please note which version of BOINC are you using? This *MAY* be an issue with 6.6.x, assuming we were all running 6.6.15+ when we saw this issue. If we were all running 6.6.x, the next question is if you are running into this problem in a system with multiple GPU "cores" (more than one GPU task in progress at the same time. I have a suspicion as to what might be happening and need this information to make a more coherent report to the Alpha mailing list ... I have been watching since I down leveled to 6.5.0 and I don't seem to have the odd behavior of slowly increasing time to completion ... I will also note that this increased time may NOT be correctly reflected in the data shown on the task page of GPU Grid after you report the task ... | |
ID: 8201 | Rating: 0 | rate: / Reply Quote | |
I see the following from the Home page news: A new batch of workunits out Maybe this has something to do with it? ____________ BOINC blog | |
ID: 8202 | Rating: 0 | rate: / Reply Quote | |
I see the following from the Home page news: I don't think so, but it could... The reason I don't think so is that I changed my version BACK down to 6.5.0 and the issue seems to have gone away. I have this suspicion as to what else may have happened. It could be one, or the other, or both ... but, when I first asked this question, GDF seemed to indicate that this was not the issue ... At any rate ... if no one else has the issue, then it was just me, and in that case I would have to ascribe it to 6.6.20 ... | |
ID: 8216 | Rating: 0 | rate: / Reply Quote | |
I picked up a GIANNI wu which locked up at 13%. Was sitting there when I got home tonight. No movement in %, just the times counting up. I restarted boinc which got it going again (now past 20%), but seems to be rather unusual. Might be a feature in boinc 6.6.20 which that machine is now running. | |
ID: 8242 | Rating: 0 | rate: / Reply Quote | |
Occasionally lock-ups happen. Might be nice if the app could detect that it's not progressing any more and could initiate a restart of itself. But I guess this is not trivial to implement.. as a task which is stuck doesn't execute its code any more. | |
ID: 8252 | Rating: 0 | rate: / Reply Quote | |
I picked up a GIANNI wu which locked up at 13%. Was sitting there when I got home tonight. No movement in %, just the times counting up. I restarted boinc which got it going again (now past 20%), but seems to be rather unusual. Might be a feature in boinc 6.6.20 which that machine is now running. The new version is supposed to be detecting stalled tasks that increase in time but not percentage... but like all new features, I don't know if it works yet or not ... | |
ID: 8260 | Rating: 0 | rate: / Reply Quote | |
The new version of BOINC, not GPU-Grid itself? | |
ID: 8262 | Rating: 0 | rate: / Reply Quote | |
The new version of BOINC, not GPU-Grid itself? Supposedly in BOINC. But its only meant to work for cpu-intensive tasks, so it doesn't work with GPUgrid. That GIANNI task locked up again at 87& this time. I have posted about it on the boinc alpha mailing list. Restarted BOINC and its progressing again now. ____________ BOINC blog | |
ID: 8263 | Rating: 0 | rate: / Reply Quote | |
The new version of BOINC, not GPU-Grid itself? According to the glossy advertising it is supposed to work with both ... But, I have serious doubts about 6.6.20 for other reasons ... | |
ID: 8264 | Rating: 0 | rate: / Reply Quote | |
Away for almost three weeks at F@H. | |
ID: 8814 | Rating: 0 | rate: / Reply Quote | |
Message boards : Graphics cards (GPUs) : Any changes to project since March30 ?