This challenge happens at PrimeGrid's "FPS" project. From the sound of it, I expect The FPS Team to do well in this challenge, otherwise I would be thoroughly confused.
Estimated cache footprint of recent tasks is 10 MBytes. Therefore, multithreading is recommended, based on the ratio of core count to last level cache size of your particular CPUs. E.g. those AMD CPUs which have 32 MB sized core complexes should probably run no more than two tasks at once on each core complex, and dedicate half of the physical cores of a CCX to each task.
(Squeezing three tasks at once onto such a CCX might work too, perhaps with some SMT usage even, but this should better be tested individually for a given host and given socket power settings. Besides, while throughput is the main goal of optimization for such a challenge, task durations should also be kept in check for better chances of getting some more tasks finished right before the end of a challenge.)
This is the average CPU time of recently returned results. Task duration is correspondingly less than this if multithreading is used. E.g. roughly half of this if 2-threaded, and so on.
Estimated cache footprint of recent tasks is 10 MBytes. Therefore, multithreading is recommended, based on the ratio of core count to last level cache size of your particular CPUs. E.g. those AMD CPUs which have 32 MB sized core complexes should probably run no more than two tasks at once on each core complex, and dedicate half of the physical cores of a CCX to each task.
(Squeezing three tasks at once onto such a CCX might work too, perhaps with some SMT usage even, but this should better be tested individually for a given host and given socket power settings. Besides, while throughput is the main goal of optimization for such a challenge, task durations should also be kept in check for better chances of getting some more tasks finished right before the end of a challenge.)