Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

The Background:

I recently built a new computer, and I'm working with the ASUS P8Z68-V Pro motherboard and Intel i7 2600k processor. While this question does not pertain specifically to my hardware, I mention what I have to explain the voltages/temperatures I get. Do note that the information in your answers should not pertain to my specific case, but computer hardware in general. Furthermore, the information should apply regardless of if the system is under-clocked, stock-clocked, and over-clocked.


The Details:

In my motherboard, there are two options that pertain to my question. The first is load-line calibration (LLC), and the second is setting the CPU voltage by manual/offset mode. After some experimenting with my manually set multiplier, I have come up with the following as a stable set of voltages in each voltage mode:

  • Manual Voltage - 1.19V at idle , drops to 1.18V under load (LLC on high).
  • Offset Voltage - 0.93V at idle, 1.19V under load, voltage spikes to 1.25V under load transitions (LLC is off).

Now I understand why the voltages result from each setting (like Vdroop), and why I need to turn LLC on/off in each case, but there are two sides to the hypothetical coin here. While my load temperatures are about equal in each case, the CPU idles a few degrees cooler in offset voltage mode (due to the lower idle voltage).

That being said, in offset mode, I noticed an interesting side effect - load transitioning causes the voltage to spike up to 1.25V. I also noticed that the voltage stays at 1.25V when starting the computer (until Windows is fully loaded and SpeedStep begins to work... brownie points if you can also tell me why this happens). With LLC enabled on any setting in offset mode, the load and idle voltages remain the same, but the peak transition voltage gets a lot higher (over 1.3V).

Conversely, when I set the voltage to manual mode (with LLC enabled, since without it Vdroop causes it to be unstable at idle), the CPU is constantly at ~1.17-1.18V, in both idle/load/startup. My point is that I don't see any voltage spikes between load transitioning - the voltage is almost constant, all the time.

Again, note that in both cases, my load temperatures are the same (a perfectly acceptable 65°C under a stress test, mid to high 50's under normal full load). Thus, I am not worried about temperatures (even at idle), but rather the longevity of the CPU with respect to these voltage settings.


The Question:

For the long-term use and stability of a computer, with respect to CPU degradation and longevity, is it better to use an offset voltage (which results in a lower idle but higher transition voltage) or manual voltage (roughly constant voltage)? Will the offset voltage spikes (although within my manufacturer's specified voltages) harm the CPU or cause it to degrade faster over time?

Assume the system is under load 60% of the time it is on (which is why I want to use offset mode - cooler and less power at idle).

Reason for bounty: I would appreciate some hard evidence (datasheets, research papers, studies, or any proof really) in favour of one method or another, specifically pertaining to fluctuating versus constant voltage.

share|improve this question
    
Are we talking about an 24/7 system or a regular office desktop on only during office hours? –  Robert Jul 12 '11 at 11:04
    
This is my personal desktop, and assume it is in use for 12-16 hours per day. That being said, I don't think the answer should depend on how long the computer is turned on per day (everything should be normalized for how long the computer was turned in total, not per day). Just so you know though, I've ensured that the machine is stable and well within acceptable operating temperatures for loading it 24-hours-a-day. –  Breakthrough Jul 12 '11 at 12:51
    
Not quite an answer... but I would leave it at stock (including cooler) and expect greatest longevity that way. Overclocking scares me after having eaten 6 months of my life over a bad purchase. –  PriceChild Jul 12 '11 at 14:20
5  
@PriceChild You couldn't pay me to use a stock cooler. Overclocked or not, there is not a single advantage to a stock cooler over an aftermarket one. The stock ones are always small, cheap, and don't cool nearly as good as most aftermarket solutions. I'd also like to note that the voltages I listed above are actually less then the ones I get when I set everything to Auto in my motherboard. –  Breakthrough Jul 12 '11 at 14:23

4 Answers 4

up vote 5 down vote accepted

Both temperature and voltage kills a CPU. A high voltage spike can kill it quickly. In your case I wouldn't worry though. The voltage spikes you have are still low. The Intel spec sheets specify the max vCore VID as 1.52v for that processor. Now, I wouldn't actually run the processor at anything near that, but on the other hand I doubt a voltage spike up to that would kill it either.

Faced with the very same question you are posing I went with the offset. I figured the bigger killer would be the extra idle voltage and temperature. And I have a much higher overclock than you do. With the overclock you have I would still expect to get 10 years of life out of the processor either way.

share|improve this answer
1  
+1 for referencing the Intel datasheet. Just curious, do you have any sources or references for CPU lifetime expectancy? With the clock speed I have, I'd say the processor is at stock frequency pretty much - I just want to know if those little voltage blips, even though they are a small magnitude, will do any harm. –  Breakthrough Jul 12 '11 at 18:40
1  
Sorry, and a little FYI - Intel specifies the maximum VID as 1.52V, not Vcore. I just mention this because under load, CPU-Z reports my VID as ~1.34V, but HWMonitor/HWInfo64 reports the CPU core voltage as 1.19V. If I recall correctly, Vcore = VID - Vdroop. –  Breakthrough Jul 12 '11 at 18:43

The primary killer of all electronic components is heat. Heat is created by friction. Friction is caused by use. High use, extreme use, and use outside the normal operating parameters create more friction and thus more heat.

So long as you have an effective-enough cooling solution for the use you envision, modifying the operating parameters should not affect the CPU to a significant extent.

Stock coolers are normally insufficient to handle overclocked processors for extended periods of time, and when they are used, the higher heat will cause problems and premature failure.

Water coolers are best for CPU longevity even if not overclocking as the temperature modulation between low use and high use is kept to much narrow range.

Aftermarket air coolers with sufficient cooling rates are recommended for price/performance value when overclocking because they can keep the CPU temperatures within acceptable ranges even under abnormal operating conditions.

Once you have selected a cooler that is capable of handling your CPU at the desired operating parameters, the form of voltage selection should be set at the point of maximum system stability.

share|improve this answer
3  
I'm going to argue that this answer is misleading. I could have a nitrogen cooled setup, but if I double the voltage of my processor, even though it's still running under 0°C, the transistors will begin to degrade and break down. While I do agree that high temperatures can kill a CPU, the same can be said for high voltages - and I have read several reports already of people degrading their shiny new 2600k's by overvolting them too much. –  Breakthrough Jul 12 '11 at 14:20
    
But is the damage from overvoltage due directly to the overvoltage or from the processor's checmical makeup being incapable of dissipating the heat effectively through the various processor layers before reaching the surface and the cooling system? Is there some evidence that precludes heat being the ultimate cause of failure? I merely addressed one issue I have observed and am relatively knowledgeable regarding. Hence, if it is misleading, it was only so unwittingly. –  music2myear Jul 12 '11 at 14:30
2  
my only issue was with the last paragraph in your answer. It would be unwise to recommend that users set their "voltage selection [...] at the point of maximum system stability". Transistor breakdown is caused by electrons tunneling through the dielectric layer, slowly breaking it down. For more information, see this IEEE Spectrum publication, "Transistor Aging". Again, while temperature is an important player, so is voltage. –  Breakthrough Jul 12 '11 at 14:34
2  
Friction has nothing to do with this. The CPU has no moving parts, and therefore, no friction. –  psusi Jul 21 '11 at 13:57
1  
Friction and electrical resistance are two different things. From dictionary.com: "surface resistance to relative motion, as of a body sliding or rolling." There are other ways of generating heat as well: sitting in the sun will heat something up, but there is no friction there either. Burning fuel also produces a lot of heat without friction. –  psusi Jul 22 '11 at 1:36

Just to add an extension to Mr Alpha's answer, I also found this interesting guide on [H]ard|Forum in regards to overclocking on the Sandy Bridge chipsets. In regards to SpeedStep settings:

It possible to overclock while keeping speedstep enabled and use the offset voltage method for Vcore increase this will provide the overall best in temperature and eatsink performance as well as generally efficiency and extension of the lifespan of the motherboards its components and the CPU. This is due to the Turbo Multiplier working and exucting ramping as Intel intended. Example leaving all CStates and Speedstep enabled will allow the CPU to idle down in Vcore and frequency (1600MHz) and ramp up when under load to 4.8GHz when needed.

Since I have a modest overclock, and have all power-saving settings enabled (there are no stability or performance issues in doing so), I believe that keeping the CPU voltage in offset mode would be the best choice in this case.

For anyone going down this route as well, do keep an eye on your temperatures and voltages using a temperature monitoring tool (e.g. HWMontior or HWiNFO). Pay special attention to the maximum CPU core voltage, and see if that fits within your needs. Also note the effects of Vdroop (see my comments on Mr Alpha's answer for details).

share|improve this answer

In response to music2myear.

It's not just heat that kills a processor. Breakthrough is correct, the interconnects degrade with increased voltage:

V=IR

If you increase the voltage (while resistance remains constant), the current will increase proportionally. Increased current running through interconnects leads to electromigration which transports conductor material away from the interconnect due to momentum transfer between electrons and metal ions.

From a simplistic point of view you could think of it as a river eroding its path. Basically, if the increased current degrades the interconnects, they will eventually fail. All CPU's have a finite lifetime, but increasing voltage, and hence current will decrease its lifespan.

That said, I recommend offset voltage, it will result in far less current passing through your processor, and the small voltage spikes are of no consequence at 1.25V.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.