Now at 736 views and still no comment from WCG staff.. Oh well, I'll keep hoping!

MM, aside from the FAHV variability, are you see this just with MCM1? What we don't seem to know is the algorithm BOINC uses in the server 700 code. Wouldn't it be nice if we could get to that code like we can for the BOINC client. I haveno doubt we have someone who could figure out how BOINC is calculating points. In the mean time, it's kind of like trying to split an atom with a meat cleaver. I'm no expert but can't help but suspect that the apparently inherent variability in MCM1 WUs at least contributes to this issue. We have a long history of projects that have at least seemed quite stable in that each given WU ran in about the reasonably same amount of CPU time on a given machine. We haven't had to deal with a more variable project. When two WU with the same estimated amount of GFLOPs can take times to complete that are orders of magnitude different, I can only suspect that that plays havoc on a good day with BOINC's points algorithm. I honestly believe that Kevin and company are well aware of the issue and greater transparency on this would merely reveal that there is little to nothing that they can do. Knowing that fact would be both good and bad. I expect that they are at the mercy of the BOINC developers on this and, however good or bad a relationship WCG has with them, they have to convince them that it's a problem that needs to be higher on the priority list, that has a reasonable solution, and that that soluntion would not break something for other projects. I have been involved in crunching here for far longer than I have been a cancer survivor though that only served to increase my commitment. For me, the points are nice. It makes for the occasionally interesting observation and mainly provides a yardstick that I can use to say here's what I've done. Still, irrespective of any actual inequity, the simple perception of inequity can be a serious, if not mortal, threat to a project. Witness others that have fallen by the wayside over the years due to such perception.

An example from a snapshot of my last 15 results earlier today, same rig on FAHV would be lowest pph = 9.36 highest = 63.2 pph

My current average is in the order of 18pph. more usually it is about 23-24pph per thread.

I wonder is there some constant that relates to the difficulty of work as judged by some other obscure machine that is part of the calculation?

There has been no change to the way points are credited. We have not done anything to change that code in about a year.

My guess about the fluctuating points is due to the "flex" parameter as well as different targets being run on WCG.

Thanks,
-Uplinger

Ok, I can buy that but OMG from 8 PPH per thread to 80?
That's a heck of a swing..