The title describes pretty well what is happening now. In addition, the model overestimates the time remaining. Thus, the time remaining is inflated not only by the s/TS being 7-8 instead of ~3.5 (extrapolated from uncorrupted s/TS value for other computer), but also this additional overestimation. When I had the crash, BOINC had run for about 120 hours out of a standard ~910 hour unit. When it restarted, it suddenly decided that the time remaining was ~3000 hrs, iirc (instead of 800). Now, with 304 model hours, it's listing the TR as 1227 hours. Right after the crash, it should have yielded a TR of ~1600h even in its screwed up state. The TR is still declining faster than the completed hours is increasing, so perhaps it will end up finishing in ~910 hours after all. In any case, I figured y'all would want to know that BOINC could get messed up in this way after such a crash. BTW, my Athlon XP 2000+, 1667MHz / 768 MB DDR 2700 RAM, BOINC 4.19, wu apparently 4.12, is still yielding the same benchmarks: ~1550 Whetstone, 3740 Dhrystone.

---

It took you 6 days to get through the first 10 trickles of your model. It then took 6 days to get your 11th trickle.

What happened with the crash was that the progress was set back to 0 in the same work unit, but the CPU time was kept. So it reran the first [edit] 100000 odd timesteps of the model. Since it calculates sec/TS by dividing the total CPU time by the number of timesteps, it is giving you too large of a value, and too large of an estimate to completion.

This pretty much sucks as you received no credits for those 6 days you reran the same model years. It's happened to me twice where progress was reset to 0 for the same work unit. One was halfway through the third phase.