Hi there,

I have just signed up to CP and been sent a work unit that will take 1836days to complete. Does this strike anyone as a little silly?

The deadline date is well before the excpected end date. I wonder what kind of supperdupper laptop they think I have.

Tim

---

That is not too unusual a time estimate. The workunits here are BIG. You should update your BOINC client to at least 5.2.8, earlier versions had some problems with the estimates on CPDN work. However don\'t expect this to radically change the estimate, it will still be over 1000 hours.

---

BOINC WIKI

BOINCing since 2002/12/8

You\'ve hidden your computers and we cannot see what it is, so it is difficult to say how long it will take. What are the specs of the computer you are talking about?

Tim did say \"will take 1836days to complete\".

If he really means \"days\", I would agree that it seems silly. I don\'t know if the \"to completion\" can display days or not (I would guess not).

Tim... the format for the \"to completion\" field is \"hours:minutes:seconds\".

1800 hours is still quite a long time, but you do get \"credit\" each time you \"trickle\" (every couple of days or so), so you don\'t wait months for credit (if you care about such things). I have a Pentium D 3.2 GHz that is estimating about 1500 hours, a PIV at 3.0 GHz estimating about 2000, and one at 1.6 GHz estimating 2500 (all running Windows).

Tim did say \"will take 1836days to complete\".

I know, but that is impossible unless he is using a 486 or something. If you meant 1800 hours, then that is not an extreme amount for a computer. But BOINC since 4.7x has been way overestimating time to completion for climateprediction.net work units so it would be useful to know the specs of the computer you are running this on.

Ok it was 1836 hours. Sorry if I wasted anyones times.

---

Ok it was 1836 hours. Sorry if I wasted anyones times.

hehe no problem. I answered the question without even noticing.

---

BOINC WIKI

BOINCing since 2002/12/8

Ok it was 1836 hours. Sorry if I wasted anyones times.

hehe no problem. I answered the question without even noticing.

And all I did was notice that you didn\'t notice...

That is not too unusual a time estimate. The workunits here are BIG. You should update your BOINC client to at least 5.2.8, earlier versions had some problems with the estimates on CPDN work. However don\'t expect this to radically change the estimate, it will still be over 1000 hours.

I have been watching the \"To Completion\" and \"CPU Time\" columns and have noticed the Completion time goes down 45 seconds and then goes back up 30 seconds or more. My computer starts with 1416 hours to Completion and after 21% progress still shows 1345 hours remaining. I was just wondering why the time to completion is not even close to the actual time to finish the work unit. If it is something I have set up incorrectly on my computer I would like to change it, otherwise I will just continue to watch the percent progress.

Mark

---

I was just wondering why the time to completion is not even close to the actual time to finish the work unit.

The simple explanation is that BOINC is designed to run many different projects, as I\'m sure you know, and these projects have very different computing requirements. BOINC measures the speed of your computer (that is what is doing when it says it is \'benchmarking\') and then tries to calculate how long the new work unit from a project will take. Unfortunately, because the computing processes are so different, a calculation which is reasonably accurate for one type of program (eg SETI) may not be accurate for another. BOINC also learns from experience how long work units take to complete on each machine, but again, that is a problem for climateprediction because the work units take much longer and so the experience takes months to accumulate. BOINC cannot rely on the percentage done in order to rebase the calculation, because with some projects the speed varies widely as it processes through the work unit - that is less true with climateprediction.

It is irritating for climateprediction participants, but because the speed calculations are more important for other projects, and it is a lot of work to change things to get BOINC to estimate accurately for climateprediction, it is something that we are having to live with for now. As people tinker, things sometimes get better, sometimes worse.

No doubt somebody will say if I have done too much violence to the technical explanation.

---

Andrew explained it quite well. Everything worked fine in estimating CPU time to completion for cpdn up until version 4.7x of BOINC. Everything above that does horrible at estimating time to completion for this project, confusing people and resulting in numerous questions to the boards. The best way to estimate time to completion right now is to see how far you have gone (% done in work tab) and see how long it has taken (hours CPU time), and figure it out from that.

Andrew explained it quite well. Everything worked fine in estimating CPU time to completion for cpdn up until version 4.7x of BOINC. Everything above that does horrible at estimating time to completion for this project, confusing people and resulting in numerous questions to the boards. The best way to estimate time to completion right now is to see how far you have gone (% done in work tab) and see how long it has taken (hours CPU time), and figure it out from that.

Actually, I think the OP has a point, although he mistook the units. My current Sulphur run has completed 3.76% in 136 hours(host id 244387), which makes a completion run time estimate of about 3600 hours, while there is a report deadline of 12/11/2006. When I was a project manager, we used a figure of 1650 - 1700 man hours of work per year, so if computers running CP are on during work hours, I\'m looking at a completion at end 2007! Since we\'re being urged to save energy by turning off equipment when it\'s not being used, should the CP people be hoping that contributors will keep their PCs on 16 hours a day?
Anyway, how significant is the report deadline? Will my contribution be ignored if the deadline is not met?
Regards to all.

---

There are a lot of points here, touched on in this thread. Unfortunately, the only work on offer at the moment is the large sulphur models (if you ignore the spinups being done elsewhere, taking four months on very fast machines running 24/7). That\'s the nature of climate research, which until this project was really restricted to the world of supercomputers.

The inaccurate time calculation offered in BOINC makes it very difficult for people to make an informed judgement whether to participate. All we can say is that there is no formal cutoff point for results to be returned. The deadline is for guidance only and whether the results will be useful when they do come in depends on what research is being done. But undoubtedly the project is best suited to the more capable machines.

---