Ok, so I have been playing with my bios lately, and along with dennisburke have learned a lot about the terminology and function of several bios settings, along with their effects on actual CPU performance. One setting that seems to be shrouded by quite a bit of mystery and confusion is vDroop, so I decided to set up my own little test to measure the effects of vDroop on a modest (yet very common) overclock of 3.2 on my i5 750. What I found was very surprising.

First, I am a firm believer that to make something work better effectively (and intelligently), you must first understand how it works, and how the variables it will encounter affect its function (as an auto technician I know that an engine needs air and fuel to produce power, and to increase performance you need to get more air, and subsequently, fuel, into and out of the cylinders). I know with a computer increasing the operating frequency makes it work faster, improving performance. As the focus of my thread is vDroop, I won't go into the other details, there are plenty of other threads on the other subjects, but not too many good ones on vDroop.

Now, I am still an overclocking rookie here, trying to understand a somewhat mysterious bios setting, and pass that knowledge along to others, be it through what I am writing here or the inevitable corrections that will follow my writing, so don't get mad if my data is not perfect, if you want quality control go pay somebody to write your forum posts, if you can live with human error, read on.

Now what vdroop does in a nutshell, is to regulate CPU vcore to what you set it to, say, 1.2v, under load conditions. It obviously will drop down when idle if you have power saving features enabled, but we are talking load here. Any electronic load, be it a motor, solenoid, light bulb, or in this case, a CPU, will pull more current under a greater load, and thus the voltage will spike under a load with vdroop off. vDroop attempts to prevent those spikes, to provide more consistent voltage to the CPU under load, which would lower temperatures, and theoretically, increase component life. I once had another member explain to me that the reason overclockers turn off vdroop is that the voltage increase actually works as an advantage under heavy load, because you can set the vcore lower and it will "ramp-up" under load, providing the extra juice needed for stability at high overclocks, while keeping the set vcore lower.
If you want more on vDroop, here it is, straight from the horses mouth.

As I have never pushed a CPU to the limit, so I wouldn't know the firsthand the advantages of turning vDroop off on a high overclock, but what I can tell you is how having vDroop on affects my i5 750 at 3.2, specifically temperature under load.

What I did is try to set up the most scientific test I could, while not wasting my whole night, so what I did was boot the computer cold, let it idle for 20 mins to let everything get a bit of heat soaked into it, then reboot, let it idle for 5 min to be sure it is completely booted, then run LinX 5x with vDroop off.
I then rebooted, let it idle, and ran 3Dmark 06.

I then turned vdroop on, and repeated the entire cycle, and was a bit surprised at my findings. I also took an ambient temp reading right before clicking "start" on LinX, so I could compensate for any changes in ambient temp. Please note that it was roughly 78f in the room when I was doing all of this, which I did on purpose, as any automotive air conditioning system will blow ice cold when its snowing, but high ambient heat will separate the good from the bad, I figured that if it was cool in the room the temps would be lower as it would be easier to keep cool, but hotter ambient would mean that the temperature gap (should) be more exaggerated.

Anyways, to my findings.

vDroop off:
    idle temps - 35-35-35-34
    max LinX temps - 77-72-72-75
    max vcore 1.29v
    3dmark score - 21,605
    ambient temp at start 24.9c

vDroop on:
    idle temps - 33-29-31-29
    max LinX temps - 72-69-67-69
    max vcore 1.29v
    3dmark score - 21,758
    ambient temp at start 24.6c
    

So what I found was that not only did having vDroop on lower temps by 3-5c in my case, but it also increased my 3Dmark scores (by very little, I know) but it also stabilized my linX scores, which makes me think it may actually be performing better.
Now I haven't had it set like this to really test stability, but early indications are that I haven't sacrificed stability by turning vdroop on, if this proves wrong, I will update. I intend on running prime blend all day tomorrow while I'm at work, that combined with the 20x LinX pass I'm about to attempt is all the stability testing I usually do, and it already passed with vdroop off, so we'll see how it does with it on.

That's it for now, I hope some of you find this helpful, or at least mildly interesting, and most important, factually correct.
Thanks for reading!

I will agree that the increase in 3Dmark is well within the margin of error, but it didn't go down.
And I'm not saying that the ramping up effect is bad, just that at my low overclock I found no change in performance or stability, and saved a few degrees by turning it on. I am aware of the fact that it can be beneficial at higher overclocks, but I was trying to make the point that you can leave it on and actually benefit, at least at a lower clock. People talk about it like vdroop is the devil, and should be turned off the second you raise your bclock, I just don't think that is entirely true, but then again, I am only at 3.2.

Actually, this is something that varies from vendor to vendor and board to board.

A level of Vdroop is a necessity on every buck controller on the market today (yes even the Volterra).

Where turning Vdroop off can cause long-term implications is boards where Vdroop has been dialled out solely to appease certain review sites and users. IE, the level of Voffset and droop is set too close to VID for the controller to safely handle. Trouble is that most reviewers/users lack the knowledge to understand overshoot and ringing artefacts and the nature of feeedback, loop gain and phase margins. So, they'll ,measure VCC with a DMM and look for VCore to be close as possible to the applied VID. While this seems to be a good ideal, you need a hf scope to see what is happening in the KHz region - things may look like they're rosy on a DMM because you 'get what you set', however, overshoot could be nasty in reality (and it gets worse at higher current draw).

Where EVGA spend a lot of time on their performance products, is making sure that their 'no VDroop' function suits the controller and FETs used right up to the hilt of user current draw in the worst case scenario. I know this because I've had a few conversations with Shamino about it. So, you should not get any issues with it turned on provided they've put the work in behind the scenes to make sure the controller and FETs are comfortable handling the transient loads.

From a user perspective - Vdroop disabled on most motherboards follows Intel specifications (the white papers show the min/max loadlines for a given level of current). Vdroop disabled and how well it is tuned varies form vendor to vendor; DON'T be alarmed if you don't quite get what you set under full -load, most of the time that small variance is required to ensure that VCC does not overshoot VID for excessive periods of time and to ensure that the margin of overshoot is not excessive (also bear in mind that any ringing effects due to an overly aggressive vdroop spec will affect overclock stability due to impedance at the ringing frequency).

later
Raja

I run my system (with)vdroop, C1E, Turbo, SpeedStep enabled and have no issues at all. Who needs to surf the web and email at max overclock anyways.

I kinda look at it this way. Intel designed the chips to throttle the speed and voltages on the fly. Overclocking them and letting the chip\mb do the same should have no ill effects on the chip....Unless your pumping massive voltages into it.

The other day I decided to run my chip at stock speeds with the 24x turbo multiplier enabled. End results the chip by default would take more voltage with a single core load than a stable 4 core 4ghz overclock.