Static voltage VS adaptive voltage when overvolting CPU?

[steffeeh]

You'll see my build in my sig

At the moment I run my processor @4.2 GHz, and I've only notched the ratio up that little extra ontop of BIOS own overclock setting when running the Asus Optimized-profile inside BIOS - which clocks the processor cores to 4.1/4.1/4.0/4.0/4.0/4.0 GHz.
What's interesting is that BIOS don't use any static voltage setting, but has an adaptive curve that changes the voltage by the needs of the CPU, so that it's lower when not needing much, but higher when the CPU has to work harder, to a maximum of 1175 mV.

 - Am I right in suggesting that this may reduce the wear and tear on the CPU, especially if I were to overvolt the CPU - as te higher voltage is only used during heavy load, and is backed down towards stock voltage while light work (surfing, office work etc). So in my case I would be at stock voltages maybe 70% of the time while working, but still be able to get that extra juice when gaming or running heavy applications when needed.
Of course it's suggested on many places that it's much more safe to overvolt these days' high end consumer CPU's as long as you have the necessary cooling, and that you shouldn't be too anxious about it as long as you've done your research and takes overvolting phase in a calm tempo. But perhaps using this adaptive voltage setting reduces the wear and tear even further?
Elsewhere there was a dude suggesting that this adaptive setting might be more harmful than setting a static overvoltage as the differencies in voltage and temperature caused by the curve, may wear and tear the material even more - however luckily I'm the son of a man with a PhD in physical chemistry active in the material branch for 25 years, and of course I asked him, and he stated that it takes much higher and much more frequent shifts in voltage and heat to become a problem - if for instance the tempearute would go from 40C -> 80-100C several times within an hour all day we would have a problem, but go between 40C -> 60-70 only a few times towards the evening there wouldn't be any problem. And isn't this what a CPU is built to handle? All those billions per second of voltage (and surely temperature also) shifts in the transistors as well as in the other parts in the CPU.

 - Is there any risk that the overclock setting gets unstable because of the curve - for instance that I've found the right voltage at 100% usage during a stresstest, but when the usage goes down to 90 (lighter stresstest) or 75% (heavy usage) the voltage according to the graph isn't optimal anymore and I get bluescreen? Or can I expect that if the voltage works at 100% I can expect the graph to still have a good voltage setting at the lower usages? Note that this curve is not linear, but seem to have been more custom made to match the different needs at different level - I've screenshoted the graph... note that this may be screenshoted inside AI Suite 3 but it only shows what BIOS already have setup.

 - In conclusion, would you recommend that I use a static voltage setting that is run all the time, or that I instead adjust the graph already active so it reaches higher as I increase the clock ratio? And why?

Here are the pros and cons of adaptive vs static as far as I'm thinking right now (some will probably be wrong):
PROS
 - Adaptive: You're at the stock voltage most of the times despite overvolting, meaning the overvolting has its impact reduced as it's not used all the time
 - Adaptive: Heat is decreased when CPU isn't used heavily as the voltage goes down, giving quieter performance from the PC as a whole
 - Static: More used and there are more references to different overclock/overvolt settings, so the results are easier to expect

CONS
 - Adaptive: The settings get less fixed and more dynamic, giving more room for some error (at least if I edit the settings poorly)
 - Adaptive: IF I were wrong - may cause more stress to the material because of shifts in voltage and temperature (unlikely imo)
 - Static: Unnecessary voltage settings at times (for instance 1.3 V during light tasks, when it really doesn't need it)

THE IMAGE: https://www.dropbox.com/s/sbbhhwu44mejxjg/Voltgraf.jpg?dl=0

 

[ProficientInMath]

I use adaptive voltage for my 5820k. I'm at 4.3 GHz and my voltage ramps up to 1.19v and then falls back to 0.792v at idle. I have not noticed any instability from intermediate voltages being applied when say at 70% load. Also the voltage never goes beyond the adaptive voltage that I set up since I have applied zero offset.

Also I don't think it will reduce or increase wear and tear since the main stressor of a cpu or any chip is the warming and cooling during load and idle cycles. This causes thermal expansion and contraction, respectively, which over time the cyclic loading fatigues the materials in the chip and will eventually cause a failure. The life of a chip as you mentioned is more dependent on the number of cycles between high and low voltages and thus high and low temperatures.

I would recommend adaptive voltage because, for example, with my 5820k chip I get instability at idle when using a static voltage. So in my case I need the voltage to ramp down at idle or else my computer will freeze, weird right but that's just how it is. Adaptive voltage will also save you on your electricity bill since your overall power draw over time will be less since you only load your cpu 30% of the time (taken from your 70% at idle number).

Your con about the unnecessary voltage at times I think is negligible if you adjust the settings to not go over a certain voltage limit with a negative or zero offset voltage. I would recommend you set your adaptive voltage to 1.175v and keep the offset at zero to keep the cpu from getting a higher voltage than it actually needs.

 

[steffeeh]

Great to hear that someone does what I'm planning of doing, with the similar components!

Right now I really don't know how to proceed on this, as people elsewhere have warned me that one alarming downside with adaptive voltage control is that it may cause voltage spikes as the processor tries to control the voltage it needs to operate properly (on some cards this is very mildly like 0.001-0.01 V, but on some cards it has been reported to cause spikes at 0.1 V above the otherwise maximum set allowed voltage on the curve, which is obviously dangerous). Though it has been reported that this only occur during very heavy stress testing.
Here's the thread if you're curious on the issue (scroll down a little):
https://www.reddit.com/r/overclocking/comments/4eq4fo/static_voltage_vs_adaptive_voltage_when/

More than this, some people seem to stick to the thought that volt and heat that goes up and down more frequently due to adaptive voltage settings is pure evil, while others have gone all in and think it's an obvious setting with pros only.
On one side I have my own father saying that practically the temperature differencies shouldn't be any issue as long as they're mild enough since the volt is so low compared to the cases he's more used to work with, while the other side is the others who saimply says that dynamic heat values cause significant wear and tear.
I'm currently trying to wrap my head around all of this - perhaps the wear and tear is equal between adaptive and static settings, but that adaptive comes with other advantages such as lower heat during light work etc?

It's like I'm being pulled from opposite directions haha

 

[ProficientInMath]

I understand your dilemma. For those spikes you can generally control them through some tweaking, testing, and verification exercises. For example, with my chip I set the adaptive voltage to 1.185v, but under load it feeds the chip a max of 1.191v. There will be some error simply because no voltage sensor is perfect so the standard error of the sensor can lead to inaccuracies in the applied voltage to the chip. You can reduce this effect by playing with the offset voltage though I would recommend using small values <= 0 and set the adaptive voltage to your stable voltage or slightly lower. For example, if I wanted to keep my chip at the adaptive voltage of 1.185v and make sure it stays there I could use a -0.005v offset to keep it from hitting the 1.191v that I mentioned earlier.

As far as people who shun adaptive voltage I think they do that becasue they don't actually understand it or have used bad settings which caused an extreme over-voltage of their chip. Every modern cpu uses adaptive voltage in the stock setup. The design engineer just decides on a safe voltage range and curve for the greatest stability and reliability so that the chip works and can deliver good performance for everyone that buys it. The folks at Intel and AMD allow you to adjust this range for more performance at your own risk, which is nice. Adaptive voltage just takes a little more care and testing to verify the chip is getting the right voltage but once you dial it in it has more advantages from my perspective than static voltage.

Static voltage is the old way of overclocking back when adaptive voltage wasn't even an adjustable setting unless you had some serious electrical engineering expertise. Some guys swear by it because it works in most cases and has little drawbacks besides power consumption. Everybody has their preference, but from a sheer pros and cons comparison I think adaptive voltage takes the cake.

 

[steffeeh]

Thank you for your answer!

I get the feeling now that we're pretty much not dealing with any dangerous spikes or such (at least not with this CPU model or series), but that you simply need to maintain some voltage headroom when using adaptive settings as the CPU may ask some extra juice time to time.

I suppose it's a good guideline for me to keep in mind that this model may ask for 0.005-0.007 V extra, and set the settings so have the headroom for that?
Or well that depends on what you're referring to with "under load" - if it's during stress tests that's good, but if it's during your daily heavy usage, then that means it could demand more during stresstests.

Also, what roles does the OC voltage and offset voltage play in addition to eachother, and what's generally the ideal relationship between those two when overvolting?

 

[ProficientInMath]

The OC voltage is the voltage that is applied at peak load and the offset is any additional voltage (or reduced voltage for negative values) that the controller thinks the chip might need. I generally like to set the offset to <= 0 because I have had times when the voltage to my cpu is greater than the OC voltage + the offset by 100 - 300 mV. Also every mobo is different in how it controls the additional voltage so that can lead to those spikes or over-voltage during load. I don't really trust how the voltage controller determines the applied offset so that's why I recommend setting it to 0 or a negative value. For my case I can dial in the max adaptive voltage applied to the cpu much easier with a zero offset and only adjusting the OC voltage then testing and tweaking to make sure the voltage controller isn't doing anything funky and my cpu is getting the voltage that I set + or - 0.005v.

 

[steffeeh]

Right now my overvolt settings by stock is 1.1 OC voltage, with an offset of +0.075 = 1.175.
I suppose there will be no practical difference if I reset this to OC voltage = 1.175 and the offset to 0?

Anyway, did the spikes you mentioned before occur during stresstests or regular heavy usage?

 

[ProficientInMath]

The spikes happened under load situations i.e. stress tests, benchmarks, loading large programs, etc. Basically if any core hit its max clock speed the voltage would jump really high unless I set the offset to zero.