zombie67
03-06-23, 10:58 PM
https://www.primegrid.com/forum_thread.php?id=10168
This is different from TdP, which was hunting primes. For the challenge series, it is work that counts not finding primes. So it does not matter is you are first vs your wingman.
The challenge is one day in duration, Mar 8 15:00:00 to Mar 9 14:59:59 (UTC). There is a countdown clock on the the PG home page, to make UTC translation easier.
Tasks count only if they were sent to your machine after the challenge start and returned before the end. So no bunkering.
The sub-project is SGS, which is CPU only. They have apps for everything except ARM. It will also run on macOS/ARM, via rosetta2 just fine. AVX helps, and AVX-512 helps even more. E cores on newer intel chips do not help, and just burn more electricity.
Best configuration is with HT/SMT off (or leaving HT/SMT on, and telling BOINC to use 50% of CPU), and running one task per core. Cache requirement is 1mb per task, so almost all CPUs can handle 1 task per core. You CAN use MT to speed up task completion, but that is less efficient overall. And the goal here is to return the most tasks in 24 hours, not to be first.
It is not necessary for your wingman to report in order for your tasks to be counted. But if the wingman reports later, and it turns out that your task is invalid, the score for that task will be removed after the challenge, during the cleanup.
Scores will be kept for individuals and teams.
FWIW, I have all my machines set up and running SGS now, and I will be doing a rolling start. The tasks complete so fast, I am willing to take the hit, in order to just not worry about timing it just right.
This is different from TdP, which was hunting primes. For the challenge series, it is work that counts not finding primes. So it does not matter is you are first vs your wingman.
The challenge is one day in duration, Mar 8 15:00:00 to Mar 9 14:59:59 (UTC). There is a countdown clock on the the PG home page, to make UTC translation easier.
Tasks count only if they were sent to your machine after the challenge start and returned before the end. So no bunkering.
The sub-project is SGS, which is CPU only. They have apps for everything except ARM. It will also run on macOS/ARM, via rosetta2 just fine. AVX helps, and AVX-512 helps even more. E cores on newer intel chips do not help, and just burn more electricity.
Best configuration is with HT/SMT off (or leaving HT/SMT on, and telling BOINC to use 50% of CPU), and running one task per core. Cache requirement is 1mb per task, so almost all CPUs can handle 1 task per core. You CAN use MT to speed up task completion, but that is less efficient overall. And the goal here is to return the most tasks in 24 hours, not to be first.
It is not necessary for your wingman to report in order for your tasks to be counted. But if the wingman reports later, and it turns out that your task is invalid, the score for that task will be removed after the challenge, during the cleanup.
Scores will be kept for individuals and teams.
FWIW, I have all my machines set up and running SGS now, and I will be doing a rolling start. The tasks complete so fast, I am willing to take the hit, in order to just not worry about timing it just right.