Jump in CPU Time per Task

Your computers are hidden so it's not possible for anyone else to check your results to see what you are talking about exactly.

What I can guess is that it might be a bit unusual but not unexpected.  Quite some time ago, I seem to remember that the FGRP5 search enjoyed quite a reduction (perhaps a factor of more than 2.5x) in the crunch times.  So it's probably just a return to data that is more like what was being processed at an earlier time.

A lot probably depends on the content of the large data files that the individual tasks you receive actually analyse.  You can tell which data file is being used as its name is part of the task name.

There have been quite a few different data files being used for that search recently.  I'm more interested in the ones for GPU tasks - I collect them as they are released.  It just so happens that the CPU ones get collected as well so here is a recent list of those.

-rw-r--r-- 7 gary gary  360829 Mar 21 16:07 LATeah1007F.dat
-rw-r--r-- 7 gary gary  153980 Mar 25 06:14 LATeah1008F.dat
-rw-r--r-- 7 gary gary 1191820 Apr  6 15:11 LATeah1009F.dat
-rw-r--r-- 7 gary gary  386512 Apr  7 18:11 LATeah1011F.dat
-rw-r--r-- 7 gary gary  647595 Apr  9 18:18 LATeah1012F.dat
-rw-r--r-- 7 gary gary 4308992 Apr 10 15:16 LATeah1013F.dat
-rw-r--r-- 7 gary gary  299773 Apr 12 00:36 LATeah1014F.dat
-rw-r--r-- 7 gary gary  359140 Apr 14 06:13 LATeah1016F.dat
-rw-r--r-- 7 gary gary 5181740 Apr 14 17:11 LATeah1017F.dat

After my name, you can see the size in bytes, the date downloaded and the full name of the data file.  I only even looked at this as a result of your query but the thing that strikes me is that many of these files used to be the same size and now that has changed.  The other unusual thing is the frequency with which new ones are being issued - almost on a daily basis.  It used to be a *lot* longer than that.  Maybe this has something to do with the Seti influx.

Maybe if you check your tasks list on the website (others can't) you could tell us if there is a correlation between crunch time and the actual data file being analysed.  Do particular files give tasks that are 'long' whilst others give tasks that are 'short'?  I suspect you'll find something like that.

Also, as a separate exercise, you could see if there is any correlation between the size of the file in bytes and the crunch time for tasks based on that file.  There are big changes in size.  Do those contribute to big changes in crunch time?  I suspect there may not be much effect there but it would be nice to know.