Extremely large files????
24 May 2007 11:33:21 UTC
Topic 192779
(moderation:
Recently I've been getting very large files (e.g. 250+ hours until completetion) with deadlines of 2 weeks or less. Not only is it impossible to meet these deadlines but it means that my other BOINC projects (e.g. SETI) never run at all because BOINC switches to earliest-deadline-first scheduling because the computer is "over-committed." What gives with the file sizes?? How can I go back to the smaller work units??
Mike S.
Extremely large files????
)
The task sizes will probably remain the same, but the speed will increase. S5R2 is a middle run where they are testing some new codes. Then they will start to optimize these codes so they go faster. Once testing is done, it will most likely go to S5R3.
This will take some time, but it sounds like some of the first optimizations may be released pretty soon.
Watch the Number Crunching forum for more information.
RE: Recently I've been
)
Hi!
I think your BOINC statistics for computing the estimated completion time were messed up by one particular result that claimed to have spend over 1 million seconds for 200+ credits.
http://einsteinathome.org/task/83764210
In fact, quite a lot of workunits crashed on your Pentium 4 recently, so I guess there's something seriously wrong with that box.
CU
BRM
RE: Okay, So what can I do
)
RE: Okay, So what can I do
)
With that many crashes, I would consider running a hardware test suite, there are several available for free , e.g. a lite version of "SiSoft Sandra".
CU
BRM
RE: RE: Okay, So what
)
My last Einstein download was over 200 hours also. I'm now nearing only 10% DONE with one day left till deadline. My old 800 mHz machine just isn't that fast! I have not had much of any workunit crashes. The few I have had were with Rosetta.
What's happening?
John R
Okay, I downloaded and ran
)
Okay, I downloaded and ran SiSoft Sandra. It found nothing wrong. Also I'm running Seti on Boinc and it is running just fine - no crashes and no extremely large files. Any other ideas??
Hello y'all. Seems that
)
Hello y'all.
Seems that I'm not alone in recently getting large Einstein Workunits (in my case 125hrs) that cannot reasonably be processed in a two-week's deadline (here on an iBook G4 1.33GHz).
Unfortunately there appears to be no way to limit the size of Workunits with respect to max. time.
Hope that PoohBear's answer of 2007-05-24 12:01 in this thread is correct that things might improve with optimizations on the Einstein engine.
CU
Frank.
RE: Hello y'all. Seems
)
Either that or (and) the deadline is increased, maybe. I can understand the frustration of casual crunchers, and maybe part of the loss in overall computing power is even caused by hosts not getting work from E@H because BOINC believes that they won't make the deadline (even worse when BOINC is somewhat conservative in the estimate). So it's in the best interest of the project to correct this problem, and I'm sure this will happen rather sooner than later.
EDIT: The WU you are crunching now shouldn't be THAT long after all, maybe BOINC manager is too pessimistic in it's estimate.
CU
BRM
RE: Hello y'all. Seems
)
Do you run other programs that need a lot of CPU time? 125 hours sounds too much for your processor.