Question About VIDs and Vcore

[parsec]

Hi Martin, hopefully some simple questions for you.

The system is an i5-3570K Ivy Bridge processor in an ASRock Z77 Extreme 4 board.

I'm using the current Beta, hw_431_2075.

As you know, HWiNFO64 displays each of the processor core's VIDs (four for this CPU) and a single Vcore value. I understand that the Vcore value may not match any of the VID's displayed.

Questions:

Is the Vcore value truly the Vcore value (given the data available to HWiNFO64) or is it a VID value, given this system?

If it is really a VID, how does HWiNFO64 choose a single value from the four VIDs? Is it the maximum, minimum, or what?

Although we have four VIDs, one for each core, the processor has and uses only one Vcore value, true or false?

Given your answer(s) to those questions, does that also apply to an i7-2600K processor in the same mother board?

Thanks, HWiNFO is the best!

[Martin]

The fundamental difference is that:
- The VID is read directly from the CPU's particular core. It's the value the CPU requests depending on its current state and the value it 'thinks' it gets. However in reality, it might get a different voltage - the exact value depends on mainboard's voltage regulator circuitry and particular settings.

- The Vcore on the other hand is a MEASURED value by a dedicated sensor. So this should be the true voltage supplied to the CPU. Reporting correct Vcore voltage might require special adjustment in HWiNFO for each mainboard model (HWiNFO has this for most of them including yours).

This situation is same for almost all CPUs prior to Haswell (4th generation Core family). On Haswell and later, the CPU features a Fully Integrated Voltage Regulator (FIVR), which is supplied a constant voltage of ~1.8V only and then the FIVR manages the requested voltages to all rails inside the CPU. On these CPUs, the monitoring of Vcore can be a bit tricky (because it's integrated inside the CPU), but some mainboards are capable of this (though I'm not sure how they really implement this and how accurate the particular values are).

[parsec]

Thanks Martin, just wanted to verify what I thought was correct.

Interesting and curious how Vcore works for the Intel pre-Haswell processors. Each core's VID can be very different, and then the Vcore is again a different value lower than any of the VIDs.

Of course, the polling rate of any hardware monitoring program is far slower compared to the changes within a CPU that does not use a static Vcore. So the Vcore we actually see is long gone in the time frame of human perception, given a varying load on a processor.

I'm well aware of the differences and problems determining the true Vcore in Haswell processors. I asked you about that in the middle of last year, and learned my ASRock Z87 board does not provide any Vcore data, only VIDs. We determined the correct way to display the CPU input voltage on those boards if you recall.

Also interesting you are not sure about the Vcore data provided by some Haswell boards, but that is due to all the information you cannot get. It's almost a miracle HWiNFO works as well as it does given that challenge.

Thanks!