My previous experience with Boinc projects is you need to run at less than 100% CPU threads to maximize production.
This was always on machines that mixed CPU and GPU crunching.
You needed at least 1 thread clear up to 4 threads clear on the larger multi-core count cpus.
That can translate into 75% (4 cores), 90% or in one of my cases 87.5%.

Does this "rule of thumb" change on systems that do only CPU-based crunching?

I don't think it does. But I could add up to 4 more threads crunching on one Rig if CPU-based crunching doesn't have the same experience as CPU/GPU crunching.

Thank you.
Tom M

---

Help, my tagline is missing..... Help, my tagline is......... Help, m........ Hel.....

My previous experience with Boinc projects is you need to run at less than 100% CPU threads to maximize production.
This was always on machines that mixed CPU and GPU crunching.
You needed at least 1 thread clear up to 4 threads clear on the larger multi-core count cpus.

I never found that to be the case.
I always ran with all cores & threads, but i did reserve a thread to support each GPU Task that was running- that was because the only reason for losing output was if a thread was trying to process a CPU Task and support a GPU at the same time.
Leaving a spare thread was just another way of doing the same thing, just not taking advantage of all the system's resources.

Unless you are making heavy use of the system while processing Rosetta work, there is no reason not to use all cores & threads (or unless you don't have enough RAM to support all cores & threads being in use).

---

Grant
Darwin NT

For my x64 machines I keep one thread free for GPU work. I also keep one thread free to give the OS and BOINC something if needed. I originally did that prior to the 4.20 Rosetta app. I haven’t tried setting it back to 100% to see if it increases the output.

With my Raspberry Pi’s I keep a core free due to memory concerns. Even the 4GB models are using 92% of available memory when using 3 out of 4 cores.

---

BOINC blog

I believe the primary reason, with a GPU project, to leave a CPU available is to handle servicing the GPU tasks. R@h is presently a CPU-only project, and so that requirement isn't there. However, as others have pointed out, there may be other reasons to run with less than all of the CPUs: Temps, fan noise, memory contention, active users... etc.

---

Rosetta Moderator: Mod.Sense

Some example when you should.
Not from this project, from E@h (those devcies can't do this project at all):

I have 2 identical Digma Hit phones (cheap smartphone, but 4 ARM cores in it).

See their results:
https://einsteinathome.org/ru/host/12826438/tasks/0/0
and
https://einsteinathome.org/ru/host/12826433/tasks/0/0

One running only 1 core of 4, second - 2 cores of 4.
Results self-explaining. If some other resource limiting (storage space in that particular case) no matter how many cores device has, matters how many it can really use.