Fluctuating Estimated PPD

[billford]

This is mainly idle curiosity, but may have relevance for any who, for one reason or another, may find themselves crowding either of the completion deadlines at all closely.

Having nothing better to do for a few minutes, I watched the progress of a WU (8084,29,4,44 if it matters) via FAHControl, and noticed that between successive frames the estimated PPD could jump between >20,000 and <10,000 and back again without any intervention on my part and without (as far as I could tell) anything else using significant chunks of CPU.

Is this just different "phases" that the computation goes through or is there something on my machine I should be looking at?

[Zagen30]

It's an issue with the way FAHControl calculates PPD. It's just a cosmetic issue, though; if you look at the log files you should see that the actual frame times remain basically the same. The fluctuations are even more drastic in an absolute sense on bigadv; on my 4P E5-4650 it'll swing between say 750k and 1.2M PPD.

[billford]

I'd noticed that it didn't seem to affect the end result, but it's nice to have oddities explained- thanks.

[Joe_H]

The fluctuating estimate is also known to happen with WU's from specific projects. There are a number where you will see a relatively stable estimate, in my experience most of the projects are not effected by this issue. The ones for projects that are effected will tend to oscillate between a value shorter than reported in the log file to one that is longer.

[7im]

Also, it does a median value of the last 3 frames instead of an average like some of the common 3rd party monitoring tools. However, that isn't the cause of the shifting values, just an FYI.