Why is PPD so inaccurate?

[RMouse]

Using the latest client for Windows, I find my PPD is all over the place. One moment it is 1000 PPD, 5 mins later it is up past 3,000 PPD and then everywhere in between. This is after 65% of the WU is done.

Why so inaccurate?

[bollix47]

There are numerous reports and tickets for inaccurate TPFs. See Ticket # 888. Hopefully it will be fixed in the near future but in the meantime HFM.net does report accurate figures.

[codysluder]

To calculate PPD, you must predict the future. How quickly will your system complete the current projects?

Computing an AVERAGE frame time will allow FAH to ASSUME that it has measured FAH's progress over a long enough time period to project that progress will continue to be made at the current rate. It cannot know if you're running something else which is slowing FAH down briefly or if you're going to start something soon which will. Changes like that are unpredictable. Then, too, how long does it have to measure progress to get a good average? One frame of each WU? Three frames of each WU? When did the WU start and has it progressed at exactly the same rate since then or did you just make a change of some kind? How much of that time has been spend in a PAUSED state?

If you can predict what the average PPD is going to be better than FAH is doing, please tell us how to do it.

[billford]

If you can predict what the average PPD is going to be better than FAH is doing, please tell us how to do it.

Or, as bollix47 says, use HFM… it's a lot closer than FAHControl !

The worst WUs for me seem to be the 900x projects- they vary by a factor of about 2 between alternate frames (with nothing else running!) with FAH, pretty steady (and close) with HFM.