Cool. Glad things are working (or at least starting to work) now.

The few i have are running and doing well fingers crossed!
i restarted a coupe that had check pointed but i have one that has gone 5 hours and no checkpoint- not restarting that.
in the middle of the stats update so i'm going to leave it alone..

My units have an estimated run time of 24 hours while the percentage complete is still working relative to the 18 hours hard limit.

Original runtime is based on project current average flops stored in the task header. Your device then computes based on it's benchmarked performance how much time it needs. Projecting 24 hours implies it's not fast enough to do it all, but given quite a few tasks exit long before job #6, not much can be said of the true end time, hence the 18 hours percent progress base. Seen too many that went from 40-50 percent to 100 in a blink of an eye.

Also wcg has decided to block the function of the duration correction factor, the agent will never learn then adapt. Probably it never would as the runtimes on a per job bases are so hard to predict.

As there's only just been a resume, the servers may have been intentionally set for a long runtime, high flops. As more results return, the average flops will be adapted in new work task headers.

SAY WHAT? Does it matter? As long as the programs run and the time to complete it is 10 days start to end the only event that should matter is if it gets in a loop, and that is what a upper limit CPU time is there to prevent.

It really shouldn't matter if the CPU clock runs at 800MHz as long as the unit is completed before the 10 day dead line or if the CPU time is 2 full days as long as its error free.

You shouldn't confuse the quality of science with technology at hand Forest, science is as science does. The grid is like a box of chocolates, you never know what putter will find the best result. just saying....

edit for appearance sake...