C-States on Auto.

[Kol12]

Is there any reason why HWiNFO only shows the C0 C-State when both enhanced C-States and Package C-States are set to auto in the Bios? When enhanced C-States is set to enabled (from auto) up to C10 HWiNFO only shows the C3 and C7 C-States. Can HWiNFO detect C-States lower than C7?

[Martin]

This depends on particular CPU model and ability to monitor C-State residency. HWiNFO reports status of all states it can.

[Kol12]

Sorry should have mentioned CPU. I can't find anything official on what C-States the 8700K support but the Asus Maximus X Hero have options to enable up to C10. I'm not sure whether the CPU being used makes those options available in the Bios or whether that is the C-States the Bios supports. Do you have any experience with that CPU?

[Martin]

C10 residency counter should be present on that CPU and HWiNFO should be reading it.
But if it doesn't show any data (i.e. residency=0), it's not shown by HWiNFO. It will be shown only once it starts to provide a non-zero value.

[Kol12]

So that would mean the CPU is not entering a C-State lower than C7?

[Martin]

Yes, if those counters are working properly. If you attach the HWiNFO Debug File I can check the details and if HWiNFO is reading them all.

[Kol12]

I have a debug file here you can have a look at. 

Another relevant question I have is - is HWiNFO able to accurately monitor the new Speedshift technology that has been moved from the OS level and integrated into the CPU itself? Did you have to make changes in the software or is this something you are still working on? My understanding is that the CPU can perform this task of dynamically changing the clock speeds and voltages much faster than the OS is able to and is now doing this at micro second intervals. Can HWiNFO poll the information that quickly? One thing that seems apparent with Coffelake and the 8700K is that the clocks idle much higher than previous generations. My basic understanding of Speedshift is that this may be more efficient. Do you have any understanding of the idle behavior with Kabylake/Coffeelake CPU's and Speedshift?

Oh, I just realised that I was running a download in the background during that debug file so it's likely the C-Sates did not go very low at all. I will set up another one.

Trouble with upload speed so uploaded here instead: http://www.mediafire.com/file/b6gqia95kd...iNFO64.DBG

[Martin]

HWiNFO reads C8-C10 counters, but they are always 0 in your debug file, which means the CPU has not entered those states (or counters are not working properly).

No software is able to monitor such fast state changes, even logging at such a fast rate is not possible, not to mention showing the results. HWiNFO displays the values sampled per refresh interval.

[Kol12]

HWiNFO reads C8-C10 counters, but they are always 0 in your debug file, which means the CPU has not entered those states (or counters are not working properly)

Is it likely the CPU will never enter such a deep C-State such as C8-C10 due to activity? I don't really understand, should desktop processors even go there? Are the C8-C10 C-States core or package C-States?

No software is able to monitor such fast state changes, even logging at such a fast rate is not possible, not to mention showing the results. HWiNFO displays the values sampled per refresh interval.

This is only an issue since the move to Speedshift though is it not? Was Speedstep more easily polled? How accurate are the CPU values (core clock and voltage) per refresh interval we see in HWiNFO with a Speedshift enabled processor? Appreciate your time.