Gravitational Wave S6 CasA) having Issues
6 Aug 2014 11:53:50 UTC
Topic 197668
(moderation:
These work units are giving my machine FITS... They are doing several things. They will run up to 18-20 percent then instantly revert back to 8.333% and freeze there for hours...I allowed it to run 4 hours after and nothing happens.
I have also watched them freezs at different stages
90.909, 45.454, 60.606... always repeating digits, Then just sit there for hours...
I had this happen in both ver 7.2.42 and in the Beta Version.
I have 2 machines...
AMD 9370 8 core running at 4.4ghz win 8.1
AMD 8350 8 core running at 4.0ghz win 7
All other work units appear to run fine.
Any Suggestions.
Gravitational Wave S6 CasA) having Issues
)
That's normal operating pattern for the S6 CasA tasks, the app checkpoints 12 times during a full run and only update the percentage done when it checkpoints. The next version of the app should update the percentage done more often, but no news on when that app might be released.
As to Boinc showing up to 18-20 percent done and then goes back to 8,333 that's because newer versions of Boin does not wait for the app to report the percentage done but increase it itself to converge but not reach 100% done. So depending on the speed of the machine Boinc might have time to increase the number quite a bit before the app reports the true value.
Conclusion: Just let them run and all will be fine. But if you don't like the way this app works then go to your Einstein@Home preferences and under "Run only the selected applications" opt out of this search.
If you haven't done so, check
)
If you haven't done so, check your event log in boinc manager. It may tell you of any problems.
Also, try rebooting. You may have a process not working properly.
Make sure your machine is deleting old files properly. A machine slow to delete files may get disc full errors.
Phil
Edit: And what Holmis said. I'm running the same on this MiniMac and all seems to be ok. The above that I wrote assumes there really is a problem. How these apps work can take some getting used to.
I thought I was wrong once, but I was mistaken.
My thoughts are this.. and I
)
My thoughts are this.. and I may be totally wrong. If BOINC only updates 12 times a work unit, then when it gets to the 2nd check point, 16.666 it should update again, no matter if it took 45 minutes or 4 hours. If they're crunching 18 percent in 45 minutes, then they should be almost done in 6 hours. Boinc should show a better number than 8.333% after 4 hours because it passes several checkpoints.
So, I'm going to try it them again and just let it run. I think 24 hours should be enough time to know. These machines are FAST and are built just to crunch for this project. I just don't want to waste processor time needlessly. and thanks for all the replys
As for Disk Space, I have 500gig allotted for BOINC. Should I Increase it?
RE: As for Disk Space, I
)
Just make sure old work is being deleted. Sometimes old works stops getting deleted and fills up the disk. It doesn't sound like that's your problem tho.
Try letting them just run and see what happens.
Phil
I thought I was wrong once, but I was mistaken.
RE: My thoughts are this..
)
Boinc is only showing you what the science application is reporting and the current application is rather bad at giving updates. Anything shown by Boinc before the 8.33% done is just a bad guess from Boinc and can be and should be disregarded. But I agree that an updated number should be given every 8.33% done no matter how long it took. You should also now that the last stage of computations might take some time and Boinc will show either 99 or 100% done for quite some time before the task is actually done.
I have little to no experience with AMD CPUs but my Intel i7 3770K @4.2GHz with HT on takes about 9 hours to complete one S6CasA task.
Sounds like a plan and as long as the status shown by Boinc is "Running" and windows tasks manager shows CPU usage then they should eventually complete.
RE: I have little to no
)
I usually have an Intel machine running, but decided to switch to AMD just to see the difference. Im actually happy with the AMD so far. BOINC is a good bench mark for processor speeds. Your I7 3770 should be comparable to my AMD FX9370. SO ill look at the 10 hours mark and see whereits at. I might have got my panties in a bunch for nothing.
now Im working on getting 2 gpu's to work in the machine. any help would be appreciated
Ok, well I learned a few
)
Ok, well I learned a few things here...
1) G Wave S6(CasA) units are easy to process with a GPU. My machine crushes one out in about 38 minutes. The CPU takes 8.5 Hours or so, by itself.
2) Setting the machine to use 100% of the processors only slows down the net output. I backed off to 95% of the processors running 100% of the time yields more total work done in a faster time.
3) BOINC estimates how fast the machine should do the work then assigns a "Remaining" time. If the time is faster than what the machine can physically do, BOINC will show 20% done in 45 minutes and then at the first checkpoint, adjust to what really was processed. In my case it was 8.333% in about 45 minutes with the CPU.
4) After your machine processes a few work units it figures out how long it really took, then adjust the "Remaining" time according to the true "Elapsed" time for future work units. This will make the next work unit you start to process appear to have a "More" correct progress percentage through out the work units processing.
5) Be patient. Its not rocket science, its just plain science. so let it do what it is going to do.
Thanks for all your help.
RE: Ok, well I learned a
)
One 'monkey wrench' is if a unit takes longer then the 'average' time, all the cached workunit times will then adjust to that 'new' time frame, even if it is only a one time thing. Your pc will then have to go thru the 'averaging out process' again.
Also be aware that the percentage completed is only a rough guess, not an exact representation of the actual status. Lots of units will get to 100% and then just keep going for a bit, it's okay and normal. Other units may be crunching right along and not show an increase for a long time, this is also normal. It all depends on the programmer making the units.
Mikey, your absolutely
)
Mikey, your absolutely right,
It Seems to be stabilizing... The FX9370 CPU is crunching 7 units at an average of under 9 hours. The GPU take about +-30 minutes for a work unit.
so in 9 hours the FX9370 is getting 23+ work units done.
My FX8350 is crunching 7 units in Just over 9 hours and its GPU is also in the +- 30 minute range.
So again I'm getting roughly 23+ work units done in 9ish hours.
Total production between both machine is approx. 123 units in 24 hours.
now, Why is the servers dishing me so many of these units. I have all the other units active to download. Every once in a while I'll see something else, But 95% are the GW S6 CasA units.
RE: Mikey, your absolutely
)
Because you are finishing them so fast is why you get so many. You can always go into your settings and uncheck them and you won't get any more, or uncheck all the others and only get those if you prefer. Most people have some kind of problems with this or that kind of unit at some point and just deselect them and let someone else run them. In the big world it all works out as what doesn't work for you end up working very well for someone else.