Message boards :
Number crunching :
Expected processing times not accurate
Message board moderation
Author | Message |
---|---|
Send message Joined: 17 Aug 04 Posts: 2 Credit: 7,114 RAC: 0 |
Machine: Windows XP SP2, INtel P4 2.4 (533) 512Mb Ram BOINC Manager 4.21 When my computer first downloaded units from LHC it was showing 1h21 as the expected completion time, but a number of these units finished really quickly. Now I am downloading units which are taking nearer the time shon in the "To Completion" is this because you are still calibrating how long it takes, or a symptom of another problem? |
Send message Joined: 3 Sep 04 Posts: 212 Credit: 4,545 RAC: 0 |
I'm not sure but I think that this is expected behaviour. Some work units just take longer to compute than others because if initial parameters are in unstable range, the protons will collide to walls and simulation ends prematurely. But it might be something else, too. Anyway, we haven't done anything to the computation time estimates. Markku Degerholm LHC@home admin |
Send message Joined: 2 Sep 04 Posts: 309 Credit: 715,258 RAC: 0 |
I could have sword that when I recently downloaded some wu's all the estimated completion times were 1hr13min on my 3.2GHz HT running XP and using BOINC Manager V4.21. I have my preferences set to contact the server every 3 days. It then downloaded 23wu's, which looked about right for the resource share I have set up for LHC. In looking at the system a small time after I noticed that 16 out of the 23 were in fact estimating completion times of 12hrs12min! All I can say is I'm glad the wu deadlines are 1 March! There is no chance of completing these within my set preferences and at 50% share of one HT cpu it will doubtful they will complete before the wu deadline. Is there a problem with handing out wu's? Have you seen the note on the front page of einstein? ---------------- Feb 13, 2005 We have enabled the BOINC configuration option 'enforce_delay_bound'. This prevents work from being sent to your machine if BOINC determines that your machine could not finish the work before the deadline. This may affect users with very slow machines, or users attached to multiple BOINC projects, whose Einstein@Home resource share is small. ---------------- Live long and crunch. Paul (S@H1 8888) BOINC/SAH BETA |
Send message Joined: 1 Sep 04 Posts: 275 Credit: 2,652,452 RAC: 0 |
I am also seeing some bad time estimates. About 1 hour initial estimate, about 8 hours total at 40% complete. BOINC WIKI BOINCing since 2002/12/8 |
Send message Joined: 18 Sep 04 Posts: 27 Credit: 2,559 RAC: 0 |
i got two wus one at approxamately an hour and the other at 12 somthing. times were off but thay has somthing to do with my computer. they alway do double what the prediction is for every project. but the 12 hour one was a million rotations. im guessing that has somthing to do with the length. |
Send message Joined: 2 Sep 04 Posts: 309 Credit: 715,258 RAC: 0 |
My point is that BOINC downloaded work that was way more than what my preferences were set for (3 days). It sounds like LHC need to activate the same option that Einstein recently activated - 'enforce_delay_bound'. I also thought that million rotation wu's were not required as 'nearly' all useful information could be gained from 100,000 rotation wu's (but that's not one for our discussion - I just want to ensure I do not loose 80 or 90 credits if someone returns a dud wu that is granted - that would just suck!). Live long and crunch! Paul (S@H1 8888) BOINC/SAH BETA |
Send message Joined: 1 Sep 04 Posts: 139 Credit: 2,579 RAC: 0 |
The first load of jobs for the run were 100,000 turns but they're mostly all done. The rest are million turn jobs and will take 10-30 hours depending on your CPU power. They ARE useful! Ben Segal / LHC@home |
Send message Joined: 18 Sep 04 Posts: 27 Credit: 2,559 RAC: 0 |
learn to predict how much work you will need to download. i like the 1 M better than the 10 K ones anyway. the 10ks finish before i can dl another one |
©2024 CERN