Im curious if there's a page or something that has information about the average time for work units. I've got two WU one tells me 56 days and the other is 65 days to complete

thanks

In reply to Petrctale's message of 8 Dec 2024:

Im curious if there's a page or something that has information about the average time for work units. I've got two WU one tells me 56 days and the other is 65 days to complete

thanks

It looks like BOINC benchmarks haven't been run on your PC, so those numbers are too high. To run the benchmarks, open BOINC manager, go to the Tools menu and select Run CPU benchmarks, they'll take a minute or so to complete. I don't think it'll change the estimates for the current tasks but it will be much more accurate for any future tasks you get.

Based on the progress of one of your tasks, I'd guess it'll take 3-4 weeks of computation time to finish the tasks. Depending on how much you let BOINC run and how heavily loaded your PC tends to be, actual time may be longer. You get exactly 100 days from the time you get the tasks to finish them, before they time out. You got those 2 tasks because they timed out for the original users.

To see recent averages, go to Computing -> Server Status, from the project website.

That is rather excessive for an i7. If you look at percentage completed and time elapsed, does that suggest they will take rather less time? (I have often seen excessive times predicted if I haven't run toos>run cpu benchmarks on a new installation before running any tasks.) I know I should do it if I want fairly realistic predictions but nearly always forget!

@Petrctale, if I go to your Tasks page https://main.cpdn.org/results.php?hostid=1554226 and select the task with credit, it shows the model takes 8 cpu secs per model timestep. There are 276779 timesteps in these workunits. Multiply by 8 cpu secs per timestep gives about 25 days of cpu time. How long that will take in real clock time depends on how much your computer is on and what percentage of cpu boinc tasks get.

If you had your computer on just 12hrs a day and boinc got 100% cpu all those 12 hrs you'd get to about 56 real days to finish the task. So, boincmgr is probably correct.

Mobile/laptop cpus & memory have lower power limits than their desktop equivalents and won't run the tasks as fast. Or you may have limited the amount of cpu time boinc gets.

---

---
CPDN Visiting Scientist

After much research I found this is CPDN FAQ's

CPDN runs experiments using 2 types of climate models – Global Climate Models and Regional Climate Models. When the project started, we mostly ran global models looking at long timescales of decades, but more recently we have been running regional models looking at much shorter time periods of months to a few years as part of our Weather@Home series of experiments.

The long, global model experiments take up to several months to run. The shorter, regional model experiments are much smaller and take between a few days and a month to run.

SEVERAL MONTHS YIKES 😬
I guess my machine will be on overdrive for a while.

In reply to Petrctale's message of 13 Dec 2024:

The long, global model experiments take up to several months to run. The shorter, regional model experiments are much smaller and take between a few days and a month to run.

SEVERAL MONTHS YIKES 😬
I guess my machine will be on overdrive for a while.

That text is out of date. CPDN does not run the very long, multi-year global model experiments like it used to. You are running the regional experiments.

See my previous message regarding how long they will take on your laptop.

---

---
CPDN Visiting Scientist

Update the 56-day Work unit errored out; the 65-day work unit only took 20 days. Overall, BOINC seems to overestimate the processing time required to complete all my projects. I have rerun benchmarks several times but it changes nothing.

All comments welcome

HAPPY NEW YEAR

Thanks

---

It takes BOINC a number of correctly completed work units to improve the estimate so don't worry (yet).
Running benchmarks won't have a significant effect on the time estimate.

Running benchmarks won't have a significant effect on the time estimate.

I have found that on my Ryzen, if I don't run the benchmarks before running any tasks the estimate is out by a factor of over 10. Ruining them first brings it down to three or less. However, when I forget, I tend to just wait till I have run a sufficiency of tasks for accuracy to get there on its own.