Message boards :
Number crunching :
Never ending WU's ?
Message board moderation
Author | Message |
---|---|
Send message Joined: 24 Oct 20 Posts: 1 Credit: 4,526,419 RAC: 0 |
Hello have 8 WU's looking like they will never end... ? (running since 10 up to 14 hours) 11346342 6932272 11346025 6932114 11339666 6928940 11338798 6928506 11323960 6921102 11317557 6917909 11317569 6917915 11317615 6917938 |
Send message Joined: 26 Dec 20 Posts: 11 Credit: 1,118,021 RAC: 2,030 |
What was the runtime before? Since yesterday we are cruncing on the fifth target: https://www.sidock.si/sidock/forum_thread.php?id=97 Please note that these new workunits can run five times as long as the old WUs, so it seems pretty normal to me. These are just way bigger WUs. |
Send message Joined: 6 Feb 21 Posts: 8 Credit: 356,002 RAC: 2 |
My tasks have gone from 2000-3000 sec to 17000-20000 sec. Credit increased too so it's not all bad! |
Send message Joined: 12 Jan 21 Posts: 13 Credit: 2,513,888 RAC: 0 |
Maybe it is a good time to think about breakpoints? If it is not already implemented,that is. Sorry, i ment checkpoints. |
Send message Joined: 16 Jan 21 Posts: 4 Credit: 2,403,376 RAC: 1,052 |
most of my results are too round 22000 seconds my highest run time so far was 38000 seconds (more than 10 hours) https://www.sidock.si/sidock/workunit.php?wuid=6927206 so a deadline with at least 3 days would be very helpful and for this run time longer than 2 hours I would too like to have check pointing on the results Matthias |
Send message Joined: 26 Oct 20 Posts: 53 Credit: 2,520,836 RAC: 0 |
My longest CPU time: 14 hours 30 min 28 sec |
Send message Joined: 23 Dec 20 Posts: 20 Credit: 1,360,768 RAC: 0 |
I had 20 hours on a slow PC, much shown as a few minutes time remaining. Paul. |
Send message Joined: 18 Dec 20 Posts: 3 Credit: 421,604 RAC: 0 |
My 4GHz i7-6700K Windows 10 system limits BOINC to using 7 CPUs. It downloaded 14 of the corona_Eprot tasks yesterday, with each of them having a 48-hour deadline and a 70-minute runtime estimate. The first task took 9 hours, the second 9.5 hours. The deadline is now less than 16 hours away, but Boinc doesn't recognise there's a problem because the estimated runtime (calculated using the workunit's rsc_fpops_est value) is out by a factor of at least 8. I'm expecting some of the 12 pending tasks to miss their deadline and others to be cancelled by the server because they haven't been started before their deadline. "The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer |
Send message Joined: 11 Oct 20 Posts: 332 Credit: 25,405,843 RAC: 9,158 |
Next bunches of tasks will be created with increased value rsc_fpops_est. |
Send message Joined: 30 Dec 20 Posts: 7 Credit: 7,850,014 RAC: 0 |
nothing changed. processing time is set to 40 minutes, and the job is running for 2 hours or more |
Send message Joined: 23 Dec 20 Posts: 20 Credit: 1,360,768 RAC: 0 |
Were they re-sends? I had several re-sends from time-outs/no replies/aborts that had old short estimate. Now got some with longer estimates. Paul. |
Send message Joined: 30 Dec 20 Posts: 7 Credit: 7,850,014 RAC: 0 |
now it’s much better, of course, but not as good as rosetta@home. If 8 hours are assigned to the task, this will be the case (+- some minutes) sidock is not so accurate - they plan two hours on a job, but it actually works four doesn’t that interfere with the other projects boinc scheduler’s work ? |
Send message Joined: 30 Oct 20 Posts: 57 Credit: 9,112,528 RAC: 0 |
sidock is not so accurate - they plan two hours on a job, but it actually works four It goes the other way with me on two Ryzen 3950s. They run between 5 and 6 hours, but the estimates are 9 hours on one machine, and 12 hours on the other. They should be converging to the correct value by now, but are correcting more slowly than usual. |
Send message Joined: 11 Oct 20 Posts: 332 Credit: 25,405,843 RAC: 9,158 |
Hi folks! Estimate execution time, displayed by BOINC client, computed by it in last several days, may be two weeks and number of estimated flops for task. After increasing rsc_fpops_est in 2.5 times estimated time on my Ryzen up top ~09:45:00, but real execution time ~ 05:45. But, I think, after several days estimate time will be close to actual. |
Send message Joined: 30 Dec 20 Posts: 7 Credit: 7,850,014 RAC: 0 |
i have one more question, perhaps off topic. why the tasks are reset after stopping, and do not start from the moment where they were interrupted ? rosetta do so |
Send message Joined: 9 Feb 21 Posts: 7 Credit: 7,364,462 RAC: 1,982 |
because the WUs have no checkpoints see also this thread: https://www.sidock.si/sidock/forum_thread.php?id=17 |
Send message Joined: 30 Oct 20 Posts: 57 Credit: 9,112,528 RAC: 0 |
But, I think, after several days estimate time will be close to actual. Yes, they have converged to 7 hours on one Ryzen 3950X, and 8 hours on the other, which are much closer to the actual time of about 5 1/2 hours. |
Send message Joined: 30 Dec 20 Posts: 7 Credit: 7,850,014 RAC: 0 |
five with the planned eight is better than eight with the planned five ) |
Send message Joined: 11 Oct 20 Posts: 332 Credit: 25,405,843 RAC: 9,158 |
And, some tasks, as usual can be computing longer than others. But credit is also larger. |
Send message Joined: 30 Jan 21 Posts: 2 Credit: 5,557,183 RAC: 1,536 |
Just aborted 62895949 after nearly 12h CPU runtime and showing 45% progress. WU's come in showing a runtime of 5.5h. I would like to see the expected runtime somewhat more accurate if WU's run normal. There seems to be quite a big spread in runtimes. Other projects are more predictable. |
©2024 SiDock@home Team