It has been advised that the HPF2 are very demanding, such that a single segment or chunk will take a very long time.....hence the restart (check)points lying far apart in time. Imagine a HPF2 with the number of segments of a HPF1, then going bust in the 11th hour......big waste. The startup problems proof the foresight was right, not to cram too many in one.

FAAH are entirely differently and smaller segmented ..... 1 or 2%. Progress should be seen steadily with time. In the UD Agent ** progress can be seen under the 'i' button in steps of 0.1% (the graph screen).

the 48, 50, 48, 50 etc suggests its attempting to complete a segment of 2% or thereabout, but not succeeding. Never seen this phenomena before and certainly not heard of anyone showing this on 3 machines of 1 user at the same time.....FAAH is generally very steady.

If it does this for extended time, i'd abort by killing the UD_7xxxxxx.exe process in Taskmanager and fetch a new WU (Work Unit), which it should do automatically. But, lets see if a WCG technician picks up on this first.

** Guessing you have UD Agent, not BOINC. Power cycling runs risk of damaging WU....best is to properly exit UD first.

EDIT 12.55pm 7.11.06: I'M SEEING THE SAME NOW BY COINCIDENCE ON THE 'i' SCREEN OF A FAAH ON UD. progresses slowly from 47.7 to 50.0%, than skips back to 47.7. Now cycled thru that 4 times in the last 2 hours......unless getting a quick reply, i'm sending this one to eternal hunting grounds, in the prescribed manner.DEVICE IS 160465, FAAH v 4.0.3.4, WU download approx. 01:21am UTC

I am running FAAH under UD Agent on two dedicated machines 24/7. I usually connect to the Internet 2 or 3 times a day to send completed WUs. Yesterday, one work unit took about 10 hours with only one checkpoint after about 5 hours. If someone is running a slightly slower processor and/or is doing other work on the machine, it is entirely possible that the program would never reach the first checkpoint in an 8-10 hour work day. If these extended checkpoints are really necessary, then the project may lose a very significant amout of processing time.

Didactylos:

Thanks for the explanation. I understand the reason, but it's the magnitude of the change that is difficult to fathom. I was seeing segment sizes of 1-2% which would correspond to checkpoints of about 12 minutes on my slow machine. Now I'm seeing a segment size of 50% or a checkpoint of about 7 hours on average. BTW, there is a post on another thread of a user who completes 40+% of a work unit and starts over at 0% after a proper exit and restart. Could be the same problem.

Ok, thank you all for the additional info.
2 out of 3 devices seem to be past the 50% barrier after 4 hours of crunching.
So, it seems I was a bit impatient.....