log msg direction to server log
12 Oct 2014 3:21:50 UTC
Topic 197752
(moderation:
Hi Folks,
12/10/2014 04:17:59 | Einstein@Home | Sending scheduler request: To fetch work.
12/10/2014 04:17:59 | Einstein@Home | Requesting new tasks for CPU
12/10/2014 04:18:01 | Einstein@Home | Scheduler request completed: got 0 new tasks
12/10/2014 04:18:01 | Einstein@Home | No work sent
12/10/2014 04:18:01 | Einstein@Home | see scheduler log messages on http://einstein5.aei.uni-hannover.de/EinsteinAtHome/host_sched_logs/11672/11672980
went there but could not understand 'what' I was supposed to be seeing.
Anyway WU for CPU have dried up, just some WU for GPU which I'm crunching 2 at a time..
Can anyone explain 'what' I was supposed to pick up from that server log?
Regards,
Cliff,
Been there, Done that, Still no damm T Shirt.
log msg direction to server log
)
Nothing visible at the moment. The log just shows you the most recent interaction between your computer and the server, and by bad luck they'd just had a little chat before I started typing this answer. Your computer asked for some GPU work, and received it. You perhaps need to wait until your computer next asks for CPU work, and have another look then.
You would be looking for the reason why no work was sent. No work available, perhaps? No disk space left? (unlikely in your case, you have 'available disk 6.15 GB'). Just for any line which says 'no work sent', and post that (plus any lines that appear to be part of the same message) here.
Hi Richard, Well
)
Hi Richard,
Well looking at the status page it seems a few servers are down and not generating WU.
I've since got some WU and have added another NV card to my rig.. so its crunching away right now.
The CPU tasks are Gamma Ray 4 ones and are long duration, so it will be some time getting them done so I've removed them from options until I'm properly sorted out.
Regards,
Cliff
Cliff,
Been there, Done that, Still no damm T Shirt.
RE: Hi Richard,
)
That's perfectly normal as the work unit generators work with high and low limits so turns on when reaching the low limit and shuts down when reaching the high limit.
Well you won't get any more CPU work then as Gamma Ray #4 is the only CPU work available at this time. We're all waiting for the next Gravity wave search to be launched but no news on that yet.
RE: ... have added another
)
When you had one card only (presumably before the errors kicked in), the BRP4G tasks had run times around the hour mark with the CPU component being close to 20 mins. Those figures seem to indicate that the CPU is struggling to properly service the GPU. Now that you have 2 cards, and after the spate of errors, the BRP4G task run times have blown out further to around 4800 secs with CPU times around 1500 secs. Are you still running CPU tasks of any description on all CPU cores?
If all CPU cores are busy with CPU tasks, it would seem that you might get better overall performance if you freed up a couple of cores, using preferences, to allow better servicing of the GPUs. You should experiment and see what works best. It also could be that with two cards, each PCIe slot is only running at x8 and you have lost performance because of that. You should check your motherboard specs to see what happens with two cards inserted.
I suspect that an FX-8350 might have difficulty keeping up with the demands of two GTX970s when running the EAH GPU tasks.
Cheers,
Gary.
Hi Gary, The Errors
)
Hi Gary,
The Errors wer'nt due to lack of CPU availability, they were due to a system crash for other reasons [finger trouble:-(]
I've run a checking routine and my rig will handle both cards and do both types of WU..
In fact it has done so for a while now, got my GPU fans running a bit higher speed with afterburner to cover the increased workload, and my CPU is water cooled.
Regards,
Cliff,
Been there, Done that, Still no damm T Shirt.
RE: .. Errors wer'nt due to
)
I wasn't suggesting they were. It seemed likely that they occurred around the time of adding the second card so I was comparing crunch times before any errors to those after the errors had finished, just as a crude method for identifying the two situations. I was simply using that as a guide when looking at tasks in your tasks list.
Of course it will but are you getting the best 'return' from it? I asked the question about the number of cores running CPU tasks simply to alert you to the possibility of improving your overall performance. I'm not at all criticising the way you run your machine and I wasn't suggesting that anything was getting too hot. I was just trying to suggest things that many others have found to help in improving performance.
Cheers,
Gary.
Hi Gary, Relax I
)
Hi Gary,
Relax I wasn't having a go at you, just saying my rig can handle it:-)
As for CPU, its 8 core with 4 core free for the GPU's [in theory at any rate]
I could pull back the core count for CPU work a bit more, but not until I get rid of some long runners that I mistakenly asked for and got...
Just proves the old maxim of being careful what you ask for:-)
[edit] Just how does one get rid of 6 FGRP4-SSE2 WU without aborting them?
Just done 2 of them at 15hours plus:-/ And the all time out on the 26th of this
month...
With 2 projects going and hourly swap overs I doubt I can process 6 of the things.
Regards,
Cliff
Cliff,
Been there, Done that, Still no damm T Shirt.
RE: Just how does one get
)
Let 'em run? =)
Make sure you have "Leave tasks in memory while suspended?" set to yes as the FGRP4 checkpoints are far apart. Setting it to yes enables the app to resume right where it was suspended and eliminates the need to reprocess from the last checkpoint.
If Boinc decides they are in risk of missing the deadline it will switch to "High priority" and run the "in danger" tasks first, the imbalance in resource share will later be compensated by running more tasks from other projects.
Hi Holmis, Ta for
)
Hi Holmis,
Ta for the advice.. I'll just let em run on then:-)
I'd better change the suspend setting tho..
Regards,
Cliff
Cliff,
Been there, Done that, Still no damm T Shirt.
