TS Bench not activating all CPU cores

[NÄQ]

Hello,

Should TS Bench test load on to all CPU cores or only some? I have i9- 12900H CPU and when running the test it looks like only half of the CPU cores are active during the test.

So I am wondering if this is a problem with the TS Bench or is there something not right with my CPU...

Hoping for advice... Thanks for reading!

 

[unclewebb]

In a perfect world, if you ran a TS Bench - 20 Thread test, one task would be assigned to each of the 129000H, 20 logical processors. Windows 11 and Intel Thread Director are not that smart. They will sometimes schedule multiple TS Bench tasks on the same logical processor and will leave other logical processors unused. If you run the TS Bench 101 times, sooner or later it might be scheduled properly. No guarantee though.

There is nothing wrong with your CPU. A future version of ThrottleStop will be more P and E core aware and will try to force TS Bench tasks to be scheduled more appropriately. I am not sure what is going to be possible.

Try running Cinebench R23. That is a good test that should load all cores and threads.

Cinebench R23

Cinebench is the perfect tool to compare CPU and graphics performance across various systems and platforms

www.techspot.com

 

[NÄQ]

In a perfect world, if you ran a TS Bench - 20 Thread test, one task would be assigned to each of the 129000H, 20 logical cores. Windows 11 and Intel Thread Director are not that smart. They will sometimes schedule multiple TS Bench tasks on the same logical processor and will leave other logical processors unused. If you run the TS Bench 101 times, sooner or later it might be scheduled properly. No guarantee though.

There is nothing wrong with your CPU. A future version of ThrottleStop will be more P and E core aware and will try to force TS Bench tasks to be scheduled more appropriately. I am not sure what is going to be possible.

Try running Cinebench R23. That is a good test that should load all cores and threads.

Cinebench R23

Cinebench is the perfect tool to compare CPU and graphics performance across various systems and platforms

www.techspot.com

Thank you for your answer! I should have noted I am currently running not Win 11 but 10.

Maybe you can advise me on the related matter... This is a gaming laptop, purchased for only one purpose - to run a real time audio VST plugin as fast and reliably as possible. When I was originally running W11 the TS Bench test, the short version of the test finished on this system in about 7-8 seconds.

Then I installed Win10 Tiny version on this machine in hopes to get it running lighter and better for the original task, but now, running TS Bench on W10 Tiny the same short test takes much longer - about 14 seconds.

I did not notice how many cores were active back when I ran the test on W11, might it have been, that there were more cores in use under W11 and that is why the test was faster or might there be something essential missing in the Win 10 Tiny install making the CPU actually perform worse?

Thanks again for your time!

 

[unclewebb]

Windows 10 is not P and E core aware. You need to use Windows 11 if you have a 12th Gen or newer Intel CPU.

When using Windows 10, tasks can randomly get scheduled on P or E cores. A task scheduled on a P core should perform better.

 

[NÄQ]

Windows 10 is not P and E core aware. You need to use Windows 11 if you have a 12th Gen or newer Intel CPU.

When using Windows 10, tasks can randomly get scheduled on P or E cores. A task scheduled on a P core should perform better.

Oh, this is great info, thank you very much!

 

[Ashlander]

In a perfect world, if you ran a TS Bench - 20 Thread test, one task would be assigned to each of the 129000H, 20 logical processors. Windows 11 and Intel Thread Director are not that smart. They will sometimes schedule multiple TS Bench tasks on the same logical processor and will leave other logical processors unused. If you run the TS Bench 101 times, sooner or later it might be scheduled properly. No guarantee though.

There is nothing wrong with your CPU. A future version of ThrottleStop will be more P and E core aware and will try to force TS Bench tasks to be scheduled more appropriately. I am not sure what is going to be possible.

Try running Cinebench R23. That is a good test that should load all cores and threads.

Cinebench R23

Cinebench is the perfect tool to compare CPU and graphics performance across various systems and platforms

www.techspot.com

Would this be why I typically get two different result ranges when using that? I noticed with my 14900HX, with all stock settings the 7680M test was consistently scoring in both the low-470s and mid-440s at a rate of about 60/40. Was doing the same but scoring 420s/380s with an early undervolt setup. (now im getting it into 290s sometimes lol)

 

[unclewebb]

The TS Bench test was designed to be a convenient way to put a load on the CPU and to create some heat. The time it takes is not that important.

When running the TS Bench on any modern Intel CPU with P and E cores, the results might vary significantly from one run to the the next or from one day to the next. Whether TS Bench tasks get scheduled on a P core or on an E core can make a big difference to the results that are reported.

I think Windows 11 does go through some sort of learning process. You might see results improve over time if Windows decides that the TS Bench deserves to run on a P core and is not just a useless background task that is assigned to an E core.