I've noticed that Sonar is just using cores 2, 3, and 4 and not using core 1 at all. I have Use multiprocessor engine and plugin load balancing checked but for whatever reason, Sonar seems to be using just three cores with core 1 showing activity just ever so slightly every once in a while. I have a utility program that checks the cpu and all cores check out fine.

Anyone else getting this behavior?

Thanks,

Dave

What does taskmanager show in usage overall and each core?
First core is often heavily used by Windows, so maybe Sonar prefer to use it less - or leave headroom for audio engine or something.
 
Default setting in bios on my computer was also to park cores, which I changed to not do that. But if people run default setting, it means that it will not start extra cores unless total usage in computer require that. It will even overclock first core before starting another core. So for daws in general that can be turned off with dynamic overclocking as well as parking cores.
 
I'm not current on Sonar, so I know what is in gui as settings, but there used to be a config to set multicore various settings. If it was aud.ini or cakewalk.ini I don't remember.
 
To get more details I use SysInternals ProcessExplorer where you can see many interesting things about each process running. See what each thread in a process is using from cpu etc.
 
What is done in a process is distributed among threads - which each can be placed on a certain core or let Windows select which core it ends up on. Computer usually has a 1000 or so threads running, TaskManager show you how many - even if most of them are suspended waiting for an event to actually consume cpu.
 
A process running has minimum one thread, usually for gui if any.

By default we try and measure the audio processing load. It turns out its quite tricky to measure that on a per core basis without including more than the audio processing load. So the meters are not particularly accurate in this mode.
You can display the OS reported values by switching the metering mode to "System performance". To do this, right click on the performance module and change the CPU Meter Options.

On the subject of the Performance meters. Please can we choose the colour?

Thanks everyone! Turns out it was an "Eye Dee Ten Tee" error (ID10T) ;-)

Somewhere along the line I apparently (inadvertently) changed my CPU metering to audio rather than system. When I changed it back to system I got what I was expecting to get, a graph showing good core usage spread out nice and even. That of course matches what I saw when using Process Explorer.

The only bummer about the system metering is that it shows the global CPU use rather than just SONAR's. Unfortunately there are no metrics in Windows to show the per PROCESS load for each core so we have to try and compute this ourselves which is tricky. System metering will show the identical values that Process explorer shows since its using the same metrics.

ProcessExplorer can show cpu on each thread, and threads enumerated on a process - so doable.
 
But if OpenMP api has call which core a thread is on, I don't know. I only activated and selected cores I wanted to use and let windows spread threads.
 
Maybe useful for internal use if to balance load overall and tell which core is to run a thread - or a system resource thingy like asked for here.
But Windows is pretty good doing balancing - I think it moves threads as needed between cores, unless chosen by programmers.
 
But if worth the trouble is another matter.

Interesting I couldn't find a public API which gives per core per process CPU use so I'm not sure what PE is using. I'll have to ask Microsoft.
Windows default load balancing is good up to a certain point. Its up to the application to properly subdivide tasks among different threads. If you load up one thread too much Windows can't do much to help finish the task quicker.

No, please, is my english so bad....
Just threads per process and PE can do cpu usage each thread so it should be somewhere.
I guess cpu within process there is some relative within the process or something.
 
I would look in omp.h or something for prototypes for openmp api calls to do things each core - if a thread is run on a certain core. Then it should be possible to do the full calculation how much cpu is on each core.
I only have an old VS2005 standard, and it does not include this omp-stuff.
I had to do LoadLibrary and stuff to do that, rather than link in importlibraries, so cannot help with details.

Does one have to worry about core parking in Windows 10?  If so, what is the fix?