Until now, for the LLR CPU sub-projects, it has always been a race situation. Two duplicate tasks for a WU created, and the user who returns theirs first gets credit for the prime found (if any). That lead to keeping zero cache and using multi-threading to try to win the race. Also, if you had a slow machine, there was not much point in crunching at PG.

LLR2 fixes that by removing the need for a wingman. LLR2 was rolled out for TRP a while ago, and is now rolled out for most of the CPU sub-projects (but not all). This means you don't have to use MT if you don't want to, and you can keep a cache of tasks to cover any down time. You will always be first, regardless of how long it takes to return a completed task.

Note: The more tasks you run at once on a CPU, the larger the L3 cache will need to be. So you still may want to run MT.

Details here: https://www.primegrid.com/forum_thread.php?id=9303