Posts by bozz4science

1) Message boards : News : Running again (Message 725)
Posted 22 Jun 2021 by bozz4science
Post:
The idea IMO is rather that you don't waste "more valuable" compute resources with this project. Would you rather have your AMD Athlon 64 X2 3800+ run into the "runtime exceeded" issues for failed tasks or your latest gen Intel/Ryzen CPUs? My stance is that my Ryzen is better off somewhere else.

Why it bothers me: The high lost compute time ratio! Even if 1 out of every 10 would fail, the lost time for the runtime exceeded task is roughly 1.5 hrs while the 9 valid tasks are usually just 2-3 min each. That is too much ... (90 min vs. 20 min compute time failed vs. valid if one were to assume a failure rate of 10%)
2) Message boards : News : Running again (Message 723)
Posted 20 Jun 2021 by bozz4science
Post:
Honestly, this lack of any communication whatsoever definitely speaks for itself. The few tasks that ran into the runtime exceeded issue account for more compute time (time that is actually just lost) than all the other valid tasks combined. All tasks that errored out did run for 4365 ± 5 sec. The incredible write speeds needed can be combatted by setting up a RAM disk but are proof of suboptimal code. Will only finish the tasks that my client pulled at least until the high invalid rate and write speeds have been sorted out.

Jim, you might be right that it's rather slower computers that might be suited for this project. Can only repeat that Ryzen 3000 series CPUs are definitely more helpful on other projects!




©2022 COPYRIGHT 2017-2018 NCN