We are running a 12CPU job, and the job is generating credits normally, but when we look at the task manager, only 8/12 of the CPUs are working at nearly 100%. Therefore, the total CPU usage never exceeds 70%.



I have set the preferences to always run at 100% for both the web and locally. I have set the power settings to high performance from the PC's control panel, and I should have set the CPU to run at maximum capacity here as well. I also changed the value of intelppm to 4 from the registry editor.
However, the situation remains unchanged.

I can't think of any other solutions. Is there a better way?

You may get more setting max threads to 4 or 6 to run 3 * 4 thread or 2 * 6 thread jobs.

---

Paul.

How do I set the core count? There is no item in the preferences.
If I need to create a config file, what should I put in it?

In prefs the field is
Max # of threads for each MilkyWay@home task

---

Paul.

sorry, where is there the field?
at least, https://milkyway.cs.rpi.edu/milkyway/prefs.php?subset=global is not include.
(Or is the Japanese page mistranslated?)

https://milkyway.cs.rpi.edu/milkyway/prefs.php?subset=project
For 100% CPU utilization set "Max # of threads for each MilkyWay@home task" to 1.

EDIT: no idea however, how well it is going to work on a CPU with P- and E-cores. Your CPU has 2 P-cores with 2 threads each and 8 E-cores. On the screenshot you posted it seems, that Windows is running the Milkyway task just on the 8 E-cores, if we assume that the first four are the 4 threads of the P-cores. Or it runs them on 4 E-cores and the 4 threads on the two P-cores, which might be even worse.

If you won't get 100% utilization with 1-thread tasks, you might want to use something like HWiNFO64 to check what's actually going on there and which cores are in use. CPUs with P- and E-cores are great for office, but might become a major PITA when trying to run something computationally intensive on them.

---

Thank you for your advice. According to HWiNFO64, the two threads of the P core, totaling four, were less than 10%, while the E core was constantly above 70%. In this case, what preferences should I set for Max # of threads and other items?

Well, first try 1 thread per task, 100% of CPU cores and 100% of CPU time.

---

I tried changing the settings, but the situation did not improve.
By the way, einstein@home's 0.5cpu + 1 Intel GPU jobs are currently running in parallel. Could this be related? (This also has the problem of only using 50% of the GPU.)

By changing the priority of the process from the Task Manager, I was able to increase the usage rate of the P core.

By default, the priority of milkyway_nbody_orbit_fitting_1.93 was set to the lowest level. All I had to do was raise it to High.
At High, there was still about 20% capacity remaining, and setting it to the highest level, “Real-time,” would allow it to operate at 100% capacity, but since this is a computer I use on a daily basis, I decided against it. (Apparently, the process becomes too monopolized by BOINC, causing other operations to slow down.)

I tried changing the settings, but the situation did not improve.

This won't affect old tasks, only new ones. Until now you have only MT-tasks on your computer, if you changed max. threads to 1, you should get new tasks without the "(mt)" in the application name, like me. Preferably crunch ready every MT-task before asking for new ones, BOINC is sometimes doing incredibly stupid things when it has a mix of ST and MT tasks, even if going from MT to ST usually works a lot better the other way around.

By the way, einstein@home's 0.5cpu + 1 Intel GPU jobs are currently running in parallel. Could this be related? (This also has the problem of only using 50% of the GPU.)

Currently probably not, but check in HWiNFO64 if the CPU is hitting some limits like power or temperature.

---

By changing the priority of the process from the Task Manager, I was able to increase the usage rate of the P core.

By default, the priority of milkyway_nbody_orbit_fitting_1.93 was set to the lowest level. All I had to do was raise it to High.
At High, there was still about 20% capacity remaining, and setting it to the highest level, “Real-time,” would allow it to operate at 100% capacity, but since this is a computer I use on a daily basis, I decided against it. (Apparently, the process becomes too monopolized by BOINC, causing other operations to slow down.)

You should be able to get at least very close to 100% with single thread tasks. If it's necessary to raise the priority to get Windows to use the P-cores, this can be done permanently in cc_config.xml:

<cc_config>
 <options>
  <process_priority>N</process_priority>
 </options>
</cc_config>

Values are 0 (lowest priority, the default), 1 (below normal), 2 (normal), 3 (high) and 4 (highest).

However keep in mind, that when crunching on the iGPU (or any GPU), you likely need to keep 1-2 threads free to feed the GPU. For example I crunch CPU tasks only on 14 of the 16 threads of my Ryzen 5700G (and I have <process_priority_special>4</process_priority_special> in cc_config.xml, so the GPU application is running with "highest" priority). You have a different CPU, so YMMV, but it's likely, that running 12 CPU tasks (or threads, doesn't really matter) will significantly slow down your GPU work.

Simple and fast method of finding the right amount of CPU threads is watching "GPU Clock (Effective)" and GPU Load in HWiNFO64, simply disable CPU computing completely, check where those values are and than start adding CPU treads until there a significant change. Than go back one step and likely you have found the sweet spot.

---

This is useful, as there is no longer any need to manage the priority of BOINC with free software. I would like to wait and see how things go until the st task begins.

Even if I canceled and discarded the mt job, no new jobs were downloaded. Should I have waited for mt to complete without discarding it?

Should I have waited for mt to complete without discarding it?

Well, it's not nice to abort work, but you should still get new when you ask for it. What's written in the log?

---

log says "This computer has reached a limit on tasks in progress"

Have you perhaps set "Max # of simultaneous MilkyWay@home tasks" instead of "Max # of threads for each MilkyWay@home task" to 1? The first one should be at "no limit".

---

Oh, you are right!

It switched to st, and the CPU usage rate approached 100% (90.9%).
However, upon closer inspection, the P-core usage rate dropped to about the same level as when I first consulted this forum.
However, unlike that first time, the E-core has risen to 120%. Does this mean that it is not possible to allocate any more to the P-core?
Does this mean that both the P-core and E-core are actually operating at nearly 90%?

At what priority do the tasks run now? This looks like all tasks are running on the E-cores, which are boosting above their standard frequency.

---