I have a question about the operation of the v4.7x completion time adjustment factor. Using optimized applications on the main SETI BOINC project, I have reported a dozen or so WU's right around 90 minutes in actual completion time. The core client, however, seems to have met a bottom level of resistance in adjusting estimated crunch times at around 165 minutes. Why won't it drop down any lower, even when I always complete WU's with times much lower than this?

This is one of the best features to be added to BOINC in a long time, in my opinion. Great job so far!

---

Probably because you are using one of the optimized clients. I'd think it would go down if you used the Seti 4.18 standard app.

I see the same thing, btw. Using one of the 4.11 apps. Don't ask me which, I can't remember. ;)

---

Jord, I don't think that the use of a custom application has much to do with it. Initially, my estimates started out at around five to six hours and quickly dropped down to where they are now (90 minutes). They just seem to have plateaued out there, despite the fact that I'm always doing work much faster than that. It’s working, but not quite the way I thought it would. I want to check and see if my understanding of how this works is incorrect, or if the process still needed to be refined. (Probably both!) ;)

---

Yes, probably both. Let's wait for JM7 to pass along, to see if he can clear it up. Isn't the guestimater still a work in progress?

---

Initially, my estimates started out at around five to six hours and quickly dropped down to where they are now (90 minutes).

Mis-spoke there. (I've got to stop trying to type messages while on the phone at the same time!) Estimates started out around five to six hours and quickly dropped down to <b>165</b> minutes. Actual crunch times are always right around 90 minutes.

---

Initially, my estimates started out at around five to six hours and quickly dropped down to where they are now (90 minutes).

Mis-spoke there. (I've got to stop trying to type messages while on the phone at the same time!) Estimates started out around five to six hours and quickly dropped down to <b>165</b> minutes. Actual crunch times are always right around 90 minutes.

The estimate should drop 10% of the difference between actual and estimated crunch time. The closer that it gets, the more slowly it moves. Wo with a crunch time of 90 minutes and an estimated crunch time, it should drop around 4 minutes for each WU.

---

BOINC WIKI

The estimate should drop 10% of the difference between actual and estimated crunch time. The closer that it gets, the more slowly it moves. Wo with a crunch time of 90 minutes and an estimated crunch time, it should drop around 4 minutes for each WU.

Something is amiss. I was watching while BOINC Manager finished up a SETI WU. It knocked 9 seconds off my completion estimates.

Estimate to complete was 2:44:45
Complete WU in 1:31:10
Estimate to complete changes to 2:44:36

http://setiathome.berkeley.edu/result.php?resultid=98330823

---

John, overnight I completed several more WU's and my estimated time has now <i>increased</i> to 2:45:01.

---

The estimate should drop 10% of the difference between actual and estimated crunch time. The closer that it gets, the more slowly it moves. Wo with a crunch time of 90 minutes and an estimated crunch time, it should drop around 4 minutes for each WU.

Funny thing running it with enhanced 4.04: It starts off at 10% less than what the end result of the last unit was, but goes up (up to 62 hours in my occasion) until it reaches 65%, then the estimate rockets down again to something alike the last unit.

At the same time though, the percentages grow very slowly, up until that 65% (it took 37 hours to get there!), then it skyrockets down. It's done 30% in the last hour. I'm at 95.36% with an estimated run time of 4 hours to go.

Those 4 hours were about 40 minutes when it resumed. Meaning that the last 5% take a lot of time again. What is it with 4.04? ;)

---

John, overnight I completed several more WU's and my estimated time has now <i>increased</i> to 2:45:01.

I found the bug, and have submitted the fix. Sometimes the simplest ones are the hardest to see. That code looks right. Fiftieth or sixtieth time staring at the code you have a DOH moment.

---

BOINC WIKI

I found the bug, and have submitted the fix. Sometimes the simplest ones are the hardest to see.

:) I know what you mean. I have moments like that at work too. Thanks for the great effort, it is very much appreciated!

---