Hi, coming from Primegrid, I know that to get optimal performances from LLR workload, the goal is to maximize the number of task and thus reducing the core count for each task while not exceeding the CPU's cache memory. This can be calculated using the FFT size.

I know SRBase isn't using LLR anymore, it has been replaced by PRST, but does it still works this way? The information available on the FAQ isn't the most complete, and if it matters, I can't find the FFT size or the cache amount required per task to run it optimally.

Also, it says that hyper-threading is not recommended, is it still the case?

Right now my main CPUs are Ryzen 7 6800H (8c/16t 16MbL3), Ryzen 7 4800H (8c/16t 8MbL3), i7 6700 (4c/8t 8MbL3) and i3 7100t(2c/4t 3MbL3) I've set up Boinc to use only 50% of the cores/threads because I can't disable HT on all motherboards and other usages benefits from HT. So right now, every CPU runs an amount of simultaneous tasks corresponding to their amount of cores (1 task per physical core).

Is it the best way, or is it better setting Boinc to use 99% of the cores/threads (reserving one to feed the GPU)?

Thanks for the quick answers, so I'll just let it be that way.