12th generation cpu max turbo frequency

[sjkpublic]

The specs for the Intel 12th generation CPU show the max turbo frequency lower than a single core.   As an example a single core can run at 4.9 but the max for all cores is 4.0.     So if all the cores are running does that mean the max that they all run is 4.0?   Or do some still run at 4.9?

[curthard89 · Accepted solution]

If all cores are running it will drop down to 4.0, if a single core is running that one core will run at 4.9.

It's basically impossible to just have one single core running for most tasks.

For instance, on a 12900k the table looks like:

P-cores:

1 active core: 5.2

2 active cores: 5.1

3 active cores: 5.0

4 active cores: 5.0

5 or more active cores: 4.9

E-cores:

1 active core: 3.9

2 active cores: 3.9

3 active cores: 3.9

4 active cores: 3.9

5 or more active cores: 3.7

[sjkpublic]

Thank you.   Tried 'kudos' but broken.    Very strange how CPU speed working these days.

[FliesEyes]

This is very interesting indeed. Do I understand this correctly, that to allow the Max Turbo Boost 3.0 Technology to operate at 5.2Ghz on one core you need to disable all the other 7 Pcores and 8 Ecores?    Curious where you found this information as I have looking for more information on this topic and coming up empty-handed. 

[FliesEyes]

Great Question, I am super keen to get to the bottom of this topic.

[sjkpublic]

It is getting rather odd how the new generation of cores and threads work.   It is even more odd as to whether the particular OS/App supports it.    At the end of the day I go with this:

https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i9-12900KF&id=4611

[FliesEyes]

I asked the question recently whether going back to using Processor Affinity functions to push some applications to dedicated cores would improve my workload productivity.  I was told the new Tread Director would do this for me so not to worry.  I did a test last night to isolate one application to 4 cores and I had a 28% improvement in my script run time.  

There seems to be a huge focus on Multi Threading benchmarking and stress testing, which is cool and I get the satisfaction in running those tests.  Personally, running real-world tests for my daily applications and improving those times is far more satisfying.  Trying to understand all the different steps and tasks that are occurring behind the scenes whether in the application itself or with hardware components I find quite challenging.  Separating the software sales and marking chaff from the actual facts is frustrating. 

The whole DDR5 DIMM slot and speed fiasco is a primary example.  Another expense lesson learnt....