What kills processors?

[Smartazz]

What really kills processors during overclocking? Voltage, temperature or both?

 

[Matthias99]

Originally posted by: Smartazz
What really kills processors during overclocking? Voltage, temperature or both?

I'm not sure that anyone has done exhaustive studies on this sort of thing.

Running at a very high temperature certainly increases the likelyhood of having some kinds of stability problems, and may decrease the lifespan of the chip somewhat. However, keep in mind that most modern processors have maximum operating temperatures of something like 80-90 degrees C, so I doubt that running one at even 60 or 70 for an extended period of time is going to kill it much faster than normal.

Significant overvolting can cause fatal transistor-level problems in a CPU, especially if it is flawed internally in some way. Eventually effects like electron tunnelling will cause some gates to no longer operate reliably, or short out. I'm not sure that small amounts of voltage increase (less than 10% over the stock voltage) are going to drop the lifespan of a CPU by much. You may be talking something like a CPU that would normally last for 10 years only lasting for 5 years -- but are you going to be using this CPU in 5 years?

I think the general assumption you have to make is that running things out of spec may cause them to break down, potentially very quickly. If you can't afford to replace it, don't mess around with that sort of thing. But if you maintain reasonable temperatures and don't go crazy with overvolting, most CPUs are going to run long enough that they will probably be obsolete (or at least cheaply replaceable) before they fail.

 

[Smartazz]

Yeah, well GPUs can safely hit 80C-90C degrees, I don't see why a CPU couldn't do that and GPUs do last a long time too.

 

[solas989]

Both can kill a cpu.

Temp - each manufacturer has certain limits on the cpu temp - anything greater will cause potential phycical failure. In General Guide. This does not mean that an overheated cpu will die right then and there, it is like slowly killing the cpu, how quickly it dies depends on the cpu. Usually the CPU crashes and the pc is rendered frozen.

Voltage - As Matthias99 said, too much volts could in time kill the cpu. He explained it well.

GPU vs CPU - different uses, manufacturing methods etc. It all depends on the chip. GPU are designed to withstand 80-90C in some case.

In general, cooler is always better.

Just my $0.02

Solas

 

[Jiggz]

Heat used to be the predominant reason but not anymore since modern cpu's now have built in over-temp protection. Over-voltage either caused intentionally (overclocking) or failure of the mobo power regulator or power protector circuit or PSU voltage spikes is the only thing I can think of that can internally damage a cpu. Of course, this does not include damaged caused by mishandling.

 

[snowdogg187]

good replies here but I see them both as enemies

 

[cubby1223]

Originally posted by: Smartazz
Yeah, well GPUs can safely hit 80C-90C degrees, I don't see why a CPU couldn't do that and GPUs do last a long time too.

GPUs typically run at 500-600mhz. CPUs are up at 2-3ghz.

 

[DARQ MX]

Well think about. Mostly Temp does... but with the temp thermistors built into the chips now. Any overheating with make the thing shut down. So over voltage would kill it more...

 

[Roguestar]

Originally posted by: Smartazz
Yeah, well GPUs can safely hit 80C-90C degrees, I don't see why a CPU couldn't do that and GPUs do last a long time too.

GPUs run at less than a third of the speed of most high-end CPUs and aren't nearly as complex in architecture as a CPU. A CPU is a multi-function powerhouse whereas these days GPUs are mostly giant FPU number-crunchers.

 

[Sunner]

I remember pm(or maybe it was Sochan, one of our sadly departed Intel employees anyway) replying to a similar question with a very lengthy response once.
I doubt many people(yours truly included) understood more than half of it, but in short, I seem to remember the answer being that as long as the CPU isn't run out of spec as far as temp goes(at the time I believe that was in the 80-90 degrees Celsius range for both Intel and AMD CPU's), that wasn't a problem.
Nor was increased clock speed in itself.
Overvolting however apparently did shorten the life span of the CPU, increasing the wear exponentially as you increased the voltage.

Again, this is just what I recall from a post that was probably 2 years ago or more, so I might be somewhat off, but I believe that was the gist of it, and considering the source, I tend to believe in it

 

[Rubycon]

RWS 350 Magnum Pellet Gun

Gives new meaning to the phrase "My processor is shot". :laugh:

 

[Special K]

I am currently looking at an academic paper on CPU reliability analysis that gives a model for the MTTF of a processor based on five common failure types as based on data from industry. It would be pretty difficult to type out the equations here, but both temperature and voltage make an appearance in the equations. However, temperature appears in all of the failure types, while voltage only appears in one - failure due to time-dependent dielectric breakdown.

However, because dynamic power dissipation is proportional to the square of the voltage, increasing the voltage will increase dynamic power dissipation, which will in turn increase temperature and therefore impact the overall lifetime of the processor.

 

[Imyourzero]

I thought I remember reading that it's voltage, not necessarily heat, that diminishes the life of a CPU. I can see how they are directly proportional under normal circumstances, but there are special cases and that's what I wonder about. For example, let's say you implement a high tech cooling solution (i.e. not air) to achieve a stable overclock. What if you have to increase the voltage on a given CPU to 1.55 or 1.6 but you're able to keep the CPU nice and cool? Will the life still be shortened because of the increased voltage even though temps are well within normal operating spec?

 

[Special K]

Originally posted by: Imyourzero
I thought I remember reading that it's voltage, not necessarily heat, that diminishes the life of a CPU. I can see how they are directly proportional under normal circumstances, but there are special cases and that's what I wonder about. For example, let's say you implement a high tech cooling solution (i.e. not air) to achieve a stable overclock. What if you have to increase the voltage on a given CPU to 1.55 or 1.6 but you're able to keep the CPU nice and cool? Will the life still be shortened because of the increased voltage even though temps are well within normal operating spec?

That makes sense - my analysis was speaking from the point of view that the cooling method was fixed - i.e., given the same processor and cooling configuration, increasing the voltage will lead to an increase in temperature.

I suppose if you could offset the temperature increase with a better cooling method, then you would only have the decreased lifetime due to TDDB, which directly depends on the voltage.

 

[Rubycon]

Originally posted by: Imyourzero
I thought I remember reading that it's voltage, not necessarily heat, that diminishes the life of a CPU. I can see how they are directly proportional under normal circumstances, but there are special cases and that's what I wonder about. For example, let's say you implement a high tech cooling solution (i.e. not air) to achieve a stable overclock. What if you have to increase the voltage on a given CPU to 1.55 or 1.6 but you're able to keep the CPU nice and cool? Will the life still be shortened because of the increased voltage even though temps are well within normal operating spec?

The problem is people are doing this but the methods of cooling are rather unusual for pc cooling (i.e. expendable refrigeration such as dry ice or liquid nitrogen) and thus continuous operation at these voltages is not possible or practical. I would imagine operation with that much overvoltage is going to cause a breakdown faster than a moderate overvoltage to get a cpu running on more conventional cooling.

 

[blamb425]

so, if voltage IS the problem, then that would mean cpu's would last just as long if they were o/c'ed to the highest speed possible at stock volts

 

[corkyg]

What comes first, voltage or heat. In OC'ing, the voltage is increased, and that in turn creates higher speeds which create heat. The simple answer is BOTH, but start with voltage as the basic initiating cause.

With today's vastly increased native speeds, why overclock? It won't be reflected in most serious apps.

 

[Matthias99]

Originally posted by: corkyg
What comes first, voltage or heat. In OC'ing, the voltage is increased, and that in turn creates higher speeds which create heat. The simple answer is BOTH, but start with voltage as the basic initiating cause.

You can OC without overvolting. And upping the voltage but keeping the same clockspeed will produce substantially more heat. Clock rate and voltage are totally independent, although you may need to increase the voltage to make the processor run stably at a higher speed.

With today's vastly increased native speeds, why overclock? It won't be reflected in most serious apps.

If you just use your system to browse the web and send email... yes, it's kind of pointless, although some people just like playing around with their hardware.

If you encode video regularly, higher CPU speeds will reduce encoding times. Some games are also sensitive to CPU speed. Distributed computing workloads will also complete faster.

 

[GuitarDaddy]

Originally posted by: Rubycon

Originally posted by: Imyourzero
I thought I remember reading that it's voltage, not necessarily heat, that diminishes the life of a CPU. I can see how they are directly proportional under normal circumstances, but there are special cases and that's what I wonder about. For example, let's say you implement a high tech cooling solution (i.e. not air) to achieve a stable overclock. What if you have to increase the voltage on a given CPU to 1.55 or 1.6 but you're able to keep the CPU nice and cool? Will the life still be shortened because of the increased voltage even though temps are well within normal operating spec?

The problem is people are doing this but the methods of cooling are rather unusual for pc cooling (i.e. expendable refrigeration such as dry ice or liquid nitrogen) and thus continuous operation at these voltages is not possible or practical. I would imagine operation with that much overvoltage is going to cause a breakdown faster than a moderate overvoltage to get a cpu running on more conventional cooling.

I hang out quite frequently over at extreme forums where phase change (refrigerated) cooling is the norm and can be used for long periods of time as opposed to DI or LN2, and I can tell you that prolonged extreme overvolting at high levels can and will kill a chip rather quickly. But the folks that are that extreme change CPU's about as often as they change socks so it really doesn't bother them. But for the real hardcore guys that use cascade, DI and LN2 to shoot for WR's it doesn't seem to be a problem because they are only doing it for a relatively short period of time.

Bottom line is extreme overvolting for extended use will kill the CPU regardless of what cooling method you use.

 

[Rubycon]

Originally posted by: GuitarDaddy

I hang out quite frequently over at extreme forums where phase change (refrigerated) cooling is the norm and can be used for long periods of time as opposed to DI or LN2, and I can tell you that prolonged extreme overvolting at high levels can and will kill a chip rather quickly. But the folks that are that extreme change CPU's about as often as they change socks so it really doesn't bother them. But for the real hardcore guys that use cascade, DI and LN2 to shoot for WR's it doesn't seem to be a problem because they are only doing it for a relatively short period of time.

Bottom line is extreme overvolting for extended use will kill the CPU regardless of what cooling method you use.

Yep, I'm over there too.

All CPU's have margin built into them. The further you overclock, the less margin you have - for a hot day, dusty heatsink, etc.

C2D's that overclock 20% with default voltage and retail cooling are amazing. Left at stock speed, they would probably continue to run at full load in a very hot room with minimal cooling. This margin is good with servers and so-called mission critical tasks.

Folks buying bottom of the barrel chips and getting performance that rivals the X series costing upwards of 5X as much find an overclock a bargain. Of course if they have to spend the difference on an elaborate water cooling setup it seems silly. BUT the experience garnered through such a path is often worth it as it teaches many different fundamentals.

For a gamer that wants to push their hardware to the max and get every last MHz out of the core, it really doesn't matter. Just don't make a post about how you lost all your nudies and warez when it burns down!

Like a drag racer, the extreme overclocker isn't particularly concerned with expense or chip lifetime. They want a number and want it fast. They'll change chips like a basshead changes fuses. :laugh:

 

[crimson117]

Originally posted by: Smartazz
What really kills processors?

I KILL PROCESSORS

 

[Binky]

The #1 killer of CPUs is.....BENT PINS! Duh!

 

[Special K]

Originally posted by: Matthias99

Originally posted by: corkyg
What comes first, voltage or heat. In OC'ing, the voltage is increased, and that in turn creates higher speeds which create heat. The simple answer is BOTH, but start with voltage as the basic initiating cause.

You can OC without overvolting. And upping the voltage but keeping the same clockspeed will produce substantially more heat. Clock rate and voltage are totally independent, although you may need to increase the voltage to make the processor run stably at a higher speed.

With today's vastly increased native speeds, why overclock? It won't be reflected in most serious apps.

If you just use your system to browse the web and send email... yes, it's kind of pointless, although some people just like playing around with their hardware.

If you encode video regularly, higher CPU speeds will reduce encoding times. Some games are also sensitive to CPU speed. Distributed computing workloads will also complete faster.

Clock rate and voltage are not totally independent. The faster you clock the processor, the less time the circuits have to charge and discharge the internal capacitances. If you continue to increase the frequency while leaving the voltage the same, eventually a circuit somewhere in the processor will no longer be able to charge/discharge that capacitor to an acceptable voltage level within the alotted time (i.e., the clock cycle). This in turn will eventually lead to what most people call "instability" or "a glitch".

Increasing the voltage allows the transistors to source/sink more current, allowing the processor to charge/discharge the capacitors faster.

 

[Jiggz]

Originally posted by: Imyourzero
I thought I remember reading that it's voltage, not necessarily heat, that diminishes the life of a CPU. I can see how they are directly proportional under normal circumstances, but there are special cases and that's what I wonder about. For example, let's say you implement a high tech cooling solution (i.e. not air) to achieve a stable overclock. What if you have to increase the voltage on a given CPU to 1.55 or 1.6 but you're able to keep the CPU nice and cool? Will the life still be shortened because of the increased voltage even though temps are well within normal operating spec?

Yes! Because you are putting unnecessary stress to the dielectric material between circuits. Applying higher voltage than normally designed to any electrical equipment will suffer two form of casualties, heat and failure of insulating material between circuits, i.e dielectric strength or characteristic.

 

[Matthias99]

Originally posted by: Special K
Clock rate and voltage are not totally independent. The faster you clock the processor, the less time the circuits have to charge and discharge the internal capacitances. If you continue to increase the frequency while leaving the voltage the same, eventually a circuit somewhere in the processor will no longer be able to charge/discharge that capacitor to an acceptable voltage level within the alotted time (i.e., the clock cycle). This in turn will eventually lead to what most people call "instability" or "a glitch".

Increasing the voltage allows the transistors to source/sink more current, allowing the processor to charge/discharge the capacitors faster.

Originally posted by: corkyg
What comes first, voltage or heat. In OC'ing, the voltage is increased, and that in turn creates higher speeds which create heat. The simple answer is BOTH, but start with voltage as the basic initiating cause.

This quote (which I was responding to) is, at best, very misleading. Clock rate and voltage are independent in the sense that you can raise or lower one without raising or lowering the other. If the voltage is too low relative to the clock rate it will cause instability, which is sort of what I said before.