CPU VID - invariant of a particular CPU? or a setting on Mobo?

[milindsmart]

I currently use an old P45 chipset + Core 2 Quad Q6600 PC, and am currently trying to really understand OCing.

I'm getting confused about VID.... There seem to be two largely-incompatible interpretations of it.

■ Anandtech seems to indicate that it is a setting for the maximum CPU voltage.
■ This overclock.net page seems to indicate that it is an indicator for overclocking quality of the chip.

I wanted to clarify this once and for all (after all this is an 8-year old platform and high time I maxed out this machine).

• 
  ■ Is a VID an immutable and inherent property of the particular CPU chip? of a batch of CPUs? of a model of CPU?
  ■ If not a property of the CPU, is it a signal by the CPU chip, asking for that value of voltage?
  ■ If so, does that value change with load and frequency?
  ■ Or is it a value that the motherboard can set?

Hoping for some clarity... Thanks in advance.

 

[techgeek]

The VID is a table unique to each CPU. It's determined during the first qualifying tests that Intel runs on the die. The VID can be read by the motherboard when it's set to AUTO to see what the CPU needs at a certain load range. The newest Intel CPU's have more than just an idle and fully loaded voltages. So two CPU exact in model number and stepping could have different VID tables.

Generally speaking, lower voltages mean "better" silicon and may overclock better. The lower voltage means that the CPU will run cooler at stock which should mean more thermal headroom for overclocking.

I've already mentioned it, but the VID is different for loading and works in conjunction with Intel Speedstep.

Running the voltage control on Manual or Static in the BIOS causes the VID to be ignored or overridden if you will.

As mentioned before the motherboard will read the VID from a register in the CPU, however one thing that isn't well documented is the fact the most motherboard manufacturers add a small voltage offset to the VID to guarantee stability. Of course this comes at the expense of wasted power and excess heat. It is this reason that the practice of undervolting is so popular. For instance my 4770K was running a Vcore of over 1.25V at load (on a Maximus Hero VI) which kept it quite warm under load. I was able to lower it 1.14V which dropped my load temps quite drastically while still remaining stable.

 

[milindsmart]

Many thanks Techgeek, that was most helpful!

So if VID is a table of voltages as best determined by Intel, then motherboards should be using the approach of adding an offset in case of overclocking... then ignoring the VID and directly setting a static voltage would not be as good, right? The voltage should adapt to the load requirement, as I understand.

That's what my motherboard allows though... only CPU voltage, no offset option I can see.

 

[techgeek]

You have an old platform. The whole Adaptive and Offset modes didn't appear until later.

Sometimes when overclocking it's actually preferable to use static voltage. Especially on these older processors. I use static voltage control to find my stable overclock voltage so I have a starting point to set up Adaptive.

The downside to Static is that the CPU will be forced to use that voltage even when idle. This causes a bit of power wastage and some additional heat. If Intel SpeedStep is enabled it should still drop the multiplier, it just won't lower Vcore. Another thing to look for in your BIOS is something called Load Line Calibration (LLC). This counteracts Vdroop which is actually part of Intel's VRM specifications. Essentially Vdroop is the amount Vcore is allowed to "droop" when under load. When a CPU is overclocked, this "drooping" can cause stability issues. Adjusting the LLC can minimize or eliminate Vdroop. With modest overclocks, it may not be necessary to play with LLC, but since Q6600s were so overclockable (mine spent most of it's life at 3GHZ or 3.2GHz), LLC adjustment was pretty much mandatory.

Hope that helps.

 

[milindsmart]

Sorry for the delay, and thanks again for the great explanation. Meanwhile I have determined through experiments that the CPU Voltage setting on my motherboard determines the MAXIMUM voltage that is supplied (at max load). The adaptive thing does seem to be working, but the wording was a bit confusing.

However I do not see any setting for LLC on my board, I guess that would limit my overclocking potential?

Coretemp shows a CPU VID of 1.3V at my current setting of FSB = 400 MHz and Multiplier x8 , Intel EIST disabled. What would be the verdict on it? Is it a good chip?

 

[techgeek]

While many Q6600s hit upwards of 3.6GHz (some even higher), 3.2GHz is pretty decent. I was able to get to 3.2GHz on mine, but it wasn't always 100% stable.