I use "4".(High)
In mt, setting it to 4 also used the P core, but in st, it seems that only the E core was used.

Hmm... is the CPU hitting some power limits perhaps? Also check in Task Manager if they indeed run at that priority.

---

The priority is maintained even in Task Manager.
Core power limit exceed and Package/ring power limit exceed become yes when the P-core rate increases, so this may be due to these restrictions.

Yes, your CPU has just 55W as PL2-Limit, it's likely impossible to run Einstein on the iGPU and Milkyway on all cores with that, at least as long as you don't limit the boost frequencies in BIOS/UEFI (might also be possible in Windows energy settings). I guess this was also limiting the MT-tasks. If you can't or don't want to limit the boost clocks, the best performance (in particular on the iGPU) you will likely get with fewer tasks, now you are running 12 tasks on 8 cores, that doesn't help to speed up anything.

---

OK, its` like difficult, but I try it.

My laptop was designed so that power limits couldn't be set from the BIOS.
Instead, I was able to change PL1 and PL2 to unlimited using Intel's Extreme Tuning Utility.

However, since this software cannot change the EDP limit, it is currently capped at this restriction.
Even so, HWiNFO Portable can increase the output to a total of 140%. The speed plateaus at around 2.5GHz.
When GPU calculations are paused, the CPU reaches 3.5GHz, but the CPU temperature quickly approaches 100 C, causing the PC to go into sleep mode. Therefore, the reduction in output due to the load the GPU places on the CPU acts as an effective brake.

Sounds good. Did you get the P-cores to crunch now?

---

Yes, both the p-core and e-core are running at nearly 110-130%. When changing settings using only XTU without going through the BIOS, the PC seems to revert the settings on its own after a certain period of time. As a result, we're still operating with the default power limits. This might be the limitation of a typical laptop PC that lacks advanced configuration capabilities.

The clock speed has stabilized between 1.7 and 2.0 GHz, which is lower than the 2.4 GHz shown in the screenshot Task Manager when I first posted about this topic.
At that time, the P-core was barely functioning, while the E-core was at about 120%.

I don't fully understand the relationship between CPU usage, effective clock speed, and computational performance. However, since all cores (especially the P-core) show significantly higher usage rates despite the slightly lower clock speed, is it correct to assume current performance is better than before?

However, since all cores (especially the P-core) show significantly higher usage rates despite the slightly lower clock speed, is it correct to assume current performance is better than before?

Yes. At least there's currently no better method for Milkyway, on other projects with nearly identical runtimes for every WU it's easier, there you just need to check how long time it needs to complete a WU, so your Einstein WUs you can check that way.

---

OK, Thank you very much!

I have a similar problem. I have a two CPU Dell server, each Xeon CPU has 6 cores, so 12 threads., and should be 24 threads in total. I can only get Milkyway work units to run at a max of 16 threads. I am not splitting time with any other projects or foreground activities. So my config is 100% of CPUS, but 90% of CPU time ( because these CPUS are passively cooled. and they are OLD!) I have run Boinc under Windows 11 and Linux, no difference.

I have a two CPU Dell server, each Xeon CPU has 6 cores, so 12 threads., and should be 24 threads in total. I can only get Milkyway work units to run at a max of 16 threads.

Set "Max # of threads for each MilkyWay@home task" to 12 or 6 (better) in your MilkyWay@home preferences. Or simply to 4 as that will work for all your computers regardless of if they have 8, 12, 16 or 24 threads.

So my config is 100% of CPUS, but 90% of CPU time ( because these CPUS are passively cooled. and they are OLD!)

Add small fans to the heat sinks, even 40mm will help more than running 90% of CPU time and you are not risking WUs getting stuck. Likely they are running very hot anyway, so check temperatures.

Running not at constant load and in particular alternating between 0% and 100% all the time increases the stress on the voltage regulators on the motherboard and the CPUs get a small voltage spike each time the load pauses untill the voltage regulators adapt, so that's not a very healthy way of running BOINC anyway.

EDIT: this video is about LLC and overclocking, but the 2nd chapter starting at 5:05 (or actually a bit before that) explains pretty well, why you want constant load on your system and not fluctuations between 0 and 100%, in particular considering BOINC not running 100% of CPU time means really 0 or 100%, pretty much no other application will do that, that's like stress testing the system 24/7.

---