Forget what you see. Your RICE jobs will run 8 hours and a few seconds , not more. The client is confused because it got these mispredicted FA@H jobs that took 5x longer than normal.

I KNOW I am right and would freak if they would not. They have an in-build routine to finish the seed/structure that goes over the 8 hours CPU time exactly. That takes depending on machine speed and complexity of seed 1 to several minutes.

Yes it's enough. The Duration Correction Factor (DCF) is pushed up full size immediately when the duration of a WU was underestimated. This is probably done for avoiding to overload your work queue if the new duration is not an accident, i.e. if all new WUs to come are so much bigger.

When the duration of a WU has been overestimated by more than 10 % the DCF will be lowered by 10 %. If it was overestimated by less than 10 % the DCF will decrease by only 1 %.

If your machine is crunching correctly sized WUs from now on (e.g. Rice or HCC, currently) the DCF should be back to a more reasonable value after about a dozen of these WUs have completed. In your case when your 10 Rice WUs will be finished the DCF should be close to 2 instead of the current 5.75. And it will go on decreasing with each new WU which had been correctly estimated, although the adjustment will be slower as the DCF comes closer to what it was normally.

The above description is meant to give you the principle of this mechanism. Since I am not a Boinc developer I do not guarantee its accuracy but you get the spirit.

Cheers. Jean.

I have had a whole batch of these jobs and all of them finished right about 8 hours no matter what the original estimate was.

Hello faldor,
Sure. Dangerous if you don't know what you are doing. So exit BOINC, apply change, restart BOINC. And be prepared to change back if you mess up.

Lawrence