Apologies if this has been raied before, I have had a look through the first few pages of threads and cannot find anything related.

I am currently running BOINC and Classic SETI on 2 PCs. I have noticed that on my main PC, a Pentium 4HT, the time counter on one of the work units does not appear to be counting correctly.

When running the work units I assume that one of them will be sharingone of the 'virtual' processors and should take longer to complete. However, while the analysis is actualy be carried out, the time counter does not accurately reflect the amount of real time that has passed.

One work unit's time passed increases correctly, and updates at 1 second intervals. The second WU's time seems to count only 1 second for every 2 seconds of realtime. Therefore, if BOINC calculates a WU's completion time to be 3.5 hours, it has actually taken 7. Shouldn't the time taken be shown as 7 hours?

Also, towards the end of a work unit, this phenomena is exagerated. For example, when the work unit I was checking last night reached 98%, for every 20 seconds (or so) of real time passed, BOINC only counted 1 second.

Is this how it has been designed? Stopping classic SETI may resolve this but I do not see this happening on my 2nd PC (AMD Athlon) where the time appears to be counted correctly.

Any ideas?

The time counters are counting CPU time, not wall time. If some other process is using the CPU, then the counter will slow down (or stop for a while in extreme cases). In your case I believe that you have a hyper threaded CPU, and the second process only gets in occasionally. Hence the slow clock.

I notice that most WUs complete in about the same time (3.5-4.5 hours), although I did have one which completed in less than 1 minute.

The percentage completed seems erratic, though: it can run up to 50% in around 90 seconds, then take 4 hours to complete the rest.

The most worthless feature seems to be the time remaining display. It appears to bear no relationship to the actual time remaining. Having climbed steadily for two hours or more, the figure then descends painfully slowly. Even the last 5 minutes of predicted remaining time can take 10-15 minutes of real time to complete.

I am not too familiar with hyper threading but it sounds like boinc is using 2 processors and then you are also trying to run classic... boinc uses 2 processors unless you change it. that would be your problem there

ttyl
Jeff (Nightowl)
All your answers in one spot:
http://setiweb.ssl.berkeley.edu/transition.php

> I notice that most WUs complete in about the same time (3.5-4.5 hours),
> although I did have one which completed in less than 1 minute.
>
> The percentage completed seems erratic, though: it can run up to 50% in around
> 90 seconds, then take 4 hours to complete the rest.
>
> The most worthless feature seems to be the time remaining display. It appears
> to bear no relationship to the actual time remaining. Having climbed steadily
> for two hours or more, the figure then descends painfully slowly. Even the
> last 5 minutes of predicted remaining time can take 10-15 minutes of real time
> to complete.

The "Progress" and "To Completion" columns are completely worthless until the unit is REALLY about 80% complete, then it's at least usable as a rough guess.