Does overclocking reduce life span/damage cpu

[Aquinus]

But let me clarify things for you... VOLTAGE doesn't matter. It is the CURRENT that will degrade a CPU.

That's not entirely accurate because voltage alone can damage a CPU if the transitor can't handle the voltage. Or in other words, the voltage on the collector of the transistor exceeds the breakdown voltage of the semiconductor, you will catastrophically kill your CPU or any IC. Under normal operation, this shouldn't occur. The big issue is when people go screwing with LLC while pumping high voltages through their CPU, so it's not even the core voltage that accelerates the damage, it's the spikes in voltage after CPU load changes (while the VRMs catch up.)

I agree with everything else you said though. I just wanted to make sure it was known that voltage *can* damage a CPU, even if that's not how most CPUs get damaged over time.

Let me put it another way: A CPU with a clogged heat sink running at stock will probably die faster than an overclocked CPU with good airflow. It really comes down to heat and if you're getting rid of it fast enough.

 

[EarthDog]

the key: keep your processor under safe voltages, also properly chilled, the same for the motherboard, most people cares about properly cooling CPU but not motherboards, Power phases, memory dims and slots are hot zones, decent cooling will bring more lifetime to your hardware…!

Unless you rock an AMD FX Hex/Octo and overclock it, there is little reason to specifically cool the VRMs. Memory is the absolute LAST thing that needs cooled...

 

[Aquinus]

Unless you rock an AMD FX Hex/Octo and overclock it, there is little reason to specifically cool the VRMs. Memory is the absolute LAST thing that needs cooled...

Well, if you don't have enough airflow and the DIMMs temperature gets a bit on the hotter side, they will die faster. They might need less cooling, but they're no less susceptible to heat damage than any other component. It needs less cooling but, cooling them is most definitely not less important. All components should be running cool if you care about longevity. At work I've seen DIMMs in servers fail due to high temperatures. It most definitely does happen.

 

[CAPSLOCKSTUCK]

Thermal throttling is your friend,

the only major component i havent cooked is a cpu

 

[LightningJR]

I have had my 2500K at 4.5Ghz - 4.8Ghz for over 3 years now. I have noticed that it requires more voltage to hit those rated speeds now and 4.8Ghz isn't attainable without better cooling. Whether that's because of my CPU or Mobo I couldn't tell you but if you know how to overclock do your research and be smart you should be fine for a long time with an overclock.

My CPU may just need a new mobo since my mobo is sub par anyway but it could be that I have degraded my CPU somewhat since I bought it.

 

[CrAsHnBuRnXp]

Ive been on a 2500k overclocked originally at 4.7GHz but then I started having some stability issues and knocked it down to 4.5GHz for 24/7 use since 2011. Yes it degrades and shortens the life span on the CPU, but it's generally not something you need to worry about. Youll end up upgrading your CPU before it ever dies on you due to a stable, well cooled, overlcock.

 

[Ikaruga]

But let me clarify things for you... VOLTAGE doesn't matter. It is the CURRENT that will degrade a CPU.

I don't understand, wouldn't you get more Current in the CPU if you would "give" it more Voltage? Anyways, in my experience, the "evil" spikes degrade it the fastest.

Well, if you don't have enough airflow and the DIMMs temperature gets a bit on the hotter side, they will die faster. They might need less cooling, but they're no less susceptible to heat damage than any other component. It needs less cooling but, cooling them is most definitely not less important. All components should be running cool if you care about longevity. At work I've seen DIMMs in servers fail due to high temperatures. It most definitely does happen.

Can you link some source for that please? I find it hard to believe that a part which only needs 1.65V and consumes around 2-3W (per stick) would have any kind of heat issues (ok, lets say you are doing some extreme overclocking and each stick eats 10W, then you might want to cool those chips a little perhaps ), so why would you need to cool DDR3 ram modules?

 

[peche]

Unless you rock an AMD FX Hex/Octo and overclock it, there is little reason to specifically cool the VRMs. Memory is the absolute LAST thing that needs cooled...

i respect your opinion ... but not share it...

 

[cadaveca]

It really comes down to heat and if you're getting rid of it fast enough.

Actually, Intel was pretty specific about the 4790K launch to say that temps were NOT a factor in the lifespan of the Haswell chips, outside of solder-melting temps. The CPU will throttle LONG before temps are an issue.

Hence my stance in the past that everyone complaining about high temps and de-lidding were all blowing it out of proportion. The "poor TIM" was not poor...it is in fact what can safe the chip before temps exceed safety margins, and by changing the TIM when you pull the heatspreader, you break that protection mechanism.


Also, Votlage is EMF - electro motive force - potential difference. It enables current flow, but does NOT do anything than indicate a difference in potential for current to flow.

Votlage can exist without current, but current CANNOT exist without that difference in voltage. Those spikes in voltage increase the ability for current to flow and it is the CURRENT that actually does the damage...voltage is really nothing BUT potential for current flow.

I don't understand, wouldn't you get more Current in the CPU if you would "give" it more Voltage? Anyways, in my experience, the "evil" spikes degrade it the fastest.

Yes, it increases the potential for current to flow. But how much is based on the resistance of the circuit... I= E/R. But again, voltage is a POTENTIAL for current flow, not a physical thing. It is a difference between two points, and without those two points, has zero meaning, because there is zero potential for flow without a beginning and end.

 

[Frick]

Can you link some source for that please? I find it hard to believe that a part which only needs 1.65V and consumes around 2-3W (per stick) would have any kind of heat issues (ok, lets say you are doing some extreme overclocking and each stick eats 10W, then you might want to cool those chips a little perhaps ), so why would you need to cool DDR3 ram modules?

IIRC cadaceca has said in a review that a stick got pretty toasty... For everyday farting around it's useless though. Like those silly OCZ Reapers. I loved that look but boy was it pointless.

 

[Ikaruga]

Hence my stance in the past that everyone complaining about high temps and de-lidding were all blowing it out of proportion. The "poor TIM" was not poor...it is in fact what can safe the chip before temps exceed safety margins, and by changing the TIM when you pull the heatspreader, you break that protection mechanism.

But some just de-lidding the chip to apply a better TIM there, so it will run cooler and last longer, and not to reach higher overclocks. I think it's a good thing if your chip runs 15C lower after a de-lid. I understand that it might not affect longevity much, but a hotter CPU affects its surroundings too, and that matters imho.

IIRC cadaveca has said in a review that a stick got pretty toasty... For everyday farting around it's useless though. Like those silly OCZ Reapers. I loved that look but boy was it pointless.

Yes, as I said, overclocking is a different story, you obviously need adequate cooling if you push things, but other than that, I think DDR3 RAM cooling is just aesthetics/marketing.

 

[silentbogo]

Can you link some source for that please? I find it hard to believe that a part which only needs 1.65V and consumes around 2-3W (per stick) would have any kind of heat issues (ok, lets say you are doing some extreme overclocking and each stick eats 10W, then you might want to cool those chips a little perhaps ), so why would you need to cool DDR3 ram modules?

Just because the power consumption is small doesn't mean that you can easily dismiss the heat, generated by the electronic part.
Here's an example: you've probably heard about RaspberryPi embedded computers and you know that they are small. Max power requirements for Model B is around 5V 500mA, which is barely 2.5W peak for all of the components, including CPU, RAM and Ethernet controller. Even though not all of this power goes to Broadcom chip, at 100% load CPU heats up high enough to burn your fingers!

Same applies to RAM. The hottest one I had was a Corsair XMS2. I had an old case with LCD status display and a few built-in temperature probes. When OCed to 900MHz I'd have BSODS and freezes from, what I've assumed at the moment, was its natural limit. But when I've attached a thin-film thermistor to RAM heatsink, I was surprised to see that at nominal voltage and less than 10% overclock it was spiking at almost 80C!!! Replacing the thermal compound and adding a simple in-case gooseneck fan dropped temps to 55-60 and allowed stable work at ~920MHz. After few years it still degraded to 890MHz max.

 

[Ikaruga]

Just because the power consumption is small doesn't mean that you can easily dismiss the heat, generated by the electronic part.
Here's an example: you've probably heard about RaspberryPi embedded computers and you know that they are small. Max power requirements for Model B is around 5V 500mA, which is barely 2.5W peak for all of the components, including CPU, RAM and Ethernet controller. Even though not all of this power goes to Broadcom chip, at 100% load CPU heats up high enough to burn your fingers!

Same applies to RAM. The hottest one I had was a Corsair XMS2. I had an old case with LCD status display and a few built-in temperature probes. When OCed to 900MHz I'd have BSODS and freezes from, what I've assumed at the moment, was its natural limit. But when I've attached a thin-film thermistor to RAM heatsink, I was surprised to see that at nominal voltage and less than 10% overclock it was spiking at almost 80C!!! Replacing the thermal compound and adding a simple in-case gooseneck fan dropped temps to 55-60 and allowed stable work at ~920MHz. After few years it still degraded to 890MHz max.

A CPU is not memory, you can't compare the temp of those under "full load", and XMS2 was a ddr2 chip, those were a lot hotter because of the higher power requirements. Look at new graphics cards where they push ram chips as much as possible, yet they just put them naked on most of the cards. That's not because those engineers are clueless, but because they know that it's not needed.

 

[cadaveca]

But some just de-lidding the chip to apply a better TIM there, so it will run cooler and last longer, and not to reach higher overclocks. I think it's a good thing if your chip runs 15C lower after a de-lid. I understand that it might not affect longevity much, but a hotter CPU affects its surroundings too, and that matters imho.

When you change the ability of the TIM to react, you also change the ability of the temp sensor to react to the current flowing through it. Silicon is a semi-conductor, whose electrical properties change with temperature. In the past, it was generally accepted that a change of 10c could lessen the lifespan of a component by half. However, it depends on the state of the silicon, and what doping methods are used, which affect the resistance of the circuit made by the CPU, that dictates whether than "10c = 1/2 life" is valid or not.

 

[Ikaruga]

When you change the ability of the TIM to react, you also change the ability of the temp sensor to react to the current flowing through it. Silicon is a semi-conductor, whose electrical properties change with temperature. In the past, it was generally accepted that a change of 10c could lessen the lifespan of a component by half. However, it depends on the state of the silicon, and what doping methods are used, which affect the resistance of the circuit made by the CPU, that dictates whether than "10c = 1/2 life" is valid or not.

I fully understood your point ofc and agreed with it, and no it's probably not valid in my opinion, but what I was trying to say is that after I delided my CPU, the temps went down around it too, and my whole motherboard runs a lot cooler now. I did not and I will not push it any further than how it was stable and safe with the original TIM, so I hope it will never get to throttling where matters what you described could arise (it's a Haswell refresh and only runs at 4.6Ghz, so it never reached risky temps with the original TIM either).

 

[cadaveca]

I get where you are coming from, for sure.

But, since the chip was designed to operate at higher temperatures, lowering them isn't going to have any significant impact on the lifespan of the chip. Yes, it can lower temps on other components as well, but since the majority of components (outside of memory and SDDs) are more than happy to operate between 80c-100c, going much lower isn't really going to have much of an impact.

In the example of the 4790K, Intel specifically designed these CPUs with a 4.6 GHz+ target. That means that these CPUs are capable of withstanding the current needed to maintain those clocks. Intel reps were very specific that because not all chips can handle these clockspeeds, Intel had found a way to reliably find CPUs within their production processes to pull these chips from.

With that idea in mind, not all chips are really capable of meeting those clockspeeds, unless they are a 4790K. Yes, some outside of that binning will do it...but some cannot. You need to understand WHY those other chips cannot keep those clocks to understand that safety, or lack thereof, of overclocking.

Yet, as I said before, companies would not have disclaimers about damage to components when OC'ing, unless you could PROVE IN A COURT OF LAW that OC could damage stuff. So no matter how you look at it, YES, OC can damage components.

 

[silentbogo]

Here's what Micron has to say about DDR3:

DDR3 is specified to operate at the standard 85°C case or an extended temperature range of 95°C with 2X the refresh rate.

By 2x refresh rate they mean that you have to refresh the contents of memory cells twice as often to keep the data intact.

Since technical document only covers DDR3 operating within normal specifications(DDR3-800, DDR3-1067, DDR3-1333), you can imagine what's happening with DDR3 sticks in your gaming rig

At the higher DDR3 clock rates and with minimal air flow, DRAM will approach or exceed the 85°C level

https://www.micron.com/~/media/documents/products/presentation/ddr3_thermals_nonnda.pdf

If you read through the entire document, you'll see that with increased density (how many modules are on your motherboard) and insufficient airflow TJ also falls drastically even with the heatsink attached (the ability to disperse heat).

There's no testing data for faster RAM, but you can also scroll down to FBDIMM testing section: due to additional heat source on the board it adds some perspective to what the thermal characteristics of non-buffered but much faster RAM can be:

A full module heat spreader decreases the DRAM and register temperatures; but to get the full benefits of the heat sink, increased air flows are required

In a hypothetical average gaming computer there is usually an adequate airflow for CPU, VGA and chipset, but if you don't have a fan moving the air around your DDR3 planks, you'll get pockets of hot air trapped in-between, which in critical cases may cause performance issues and reduce its lifespan.

So there is a reason why high-performance RAM comes with heatsinks and there is market for memory cooling devices.

Videocards are no different. High-end videocards have heatspreaders attached to VRAM, while low-end cards can afford to skip heatspreaders because GPU ICs generate less heat and HSF provides adequate airflow to keep temperature within recommended specs.

 

[Ikaruga]

I get where you are coming from, for sure.

But, since the chip was designed to operate at higher temperatures, lowering them isn't going to have any significant impact on the lifespan of the chip. Yes, it can lower temps on other components as well, but since the majority of components (outside of memory and SDDs) are more than happy to operate between 80c-100c, going much lower isn't really going to have much of an impact.

In the example of the 4790K, Intel specifically designed these CPUs with a 4.6 GHz+ target. That means that these CPUs are capable of withstanding the current needed to maintain those clocks. Intel reps were very specific that because not all chips can handle these clockspeeds, Intel had found a way to reliably find CPUs within their production processes to pull these chips from.

With that idea in mind, not all chips are really capable of meeting those clockspeeds, unless they are a 4790K. Yes, some outside of that binning will do it...but some cannot. You need to understand WHY those other chips cannot keep those clocks to understand that safety, or lack thereof, of overclocking.

Yet, as I said before, companies would not have disclaimers about damage to components when OC'ing, unless you could PROVE IN A COURT OF LAW that OC could damage stuff. So no matter how you look at it, YES, OC can damage components.

We need to test these things for science. Talking about it without testing is meh. We need an experiment with two (lets say) G3258s, one is running hot while the other is well cooled, and see if the hot starts to ask for more juice to maintain it's overclock.

 

[cadaveca]

We need to test these things for science. Talking about it without testing is meh. We need an experiment with two (lets say) G3258s, one is running hot while the other is well cooled, and see if the hot starts to ask for more juice to maintain it's overclock.

Well therein lies the point of contention... no two chips are going to react to said conditions in the exact same way, because they will never be 100% physically identical.

But, I can say that the hot chip will throttle before reaching the point where temps may damage it, so doing such a task would be futile, nevermind it would probably take the better part of a decade.

 

[Ikaruga]

Here's what Micron has to say about DDR3:



By 2x refresh rate they mean that you have to refresh the contents of memory cells twice as often to keep the data intact.

Since technical document only covers DDR3 operating within normal specifications(DDR3-800, DDR3-1067, DDR3-1333), you can imagine what's happening with DDR3 sticks in your gaming rig



https://www.micron.com/~/media/documents/products/presentation/ddr3_thermals_nonnda.pdf

If you read through the entire document, you'll see that with increased density (how many modules are on your motherboard) and insufficient airflow TJ also falls drastically even with the heatsink attached (the ability to disperse heat).

There's no testing data for faster RAM, but you can also scroll down to FBDIMM testing section: due to additional heat source on the board it adds some perspective to what the thermal characteristics of non-buffered but much faster RAM can be:


In a hypothetical average gaming computer there is usually an adequate airflow for CPU, VGA and chipset, but if you don't have a fan moving the air around your DDR3 planks, you'll get pockets of hot air trapped in-between, which in critical cases may cause performance issues and reduce its lifespan.

So there is a reason why high-performance RAM comes with heatsinks and there is market for memory cooling devices.

Videocards are no different. High-end videocards have heatspreaders attached to VRAM, while low-end cards can afford to skip heatspreaders because GPU ICs generate less heat and HSF provides adequate airflow to keep temperature within recommended specs.

That's a very nice documents, thanks for the link. I'm not going to question or counter the findings of that document, but I'm honestly puzzled now. I have two kits from Crucial and G.Skill both are without heat spreaders which I only bought to test this heat spreader theory some time ago. I vigorously tortured and tested these modules and never measured temps above 60-70C, so I don't know what to think now.

Well therein lies the point of contention... no two chips are going to react to said conditions in the exact same way, because they will never be 100% physically identical.

But, I can say that the hot chip will throttle before reaching the point where temps may damage it, so doing such a task would be futile, nevermind it would probably take the better part of a decade.

Throttling is bad, if your CPU needs to throttle itself, you need better cooling, so I was thinking about safe high temps vs low temps (let's say 80C vs 40C), but yes, obviously this is something what can't really be tested because it would take a decade indeed.

 

[silentbogo]

That's a very nice documents, thanks for the link. I'm not going to question or counter the findings of that document, but I'm honestly puzzled now. I have two kits from Crucial and G.Skill both are without heat spreaders which I only bought to test this heat spreader theory some time ago. I vigorously tortured and tested these modules and never measured temps above 60-70C, so I don't know what to think now.

It all depends on the type of measuring device and the way you measure it. Small fraction of unbuffered memory has a built-in temperature sensor, which gives the closest approximation of what the real temperature might be, but if you measure it externally, you will get lower temps and larger error margin, When I was measuring temps on my XMS2, I've attached 2 thermistors to the inner side of the heatsink with thermal paste. I've used the same input line on my CPU prior to that and I know that it differed from CPU readings by ~10-12C around the same temperature range. So, if I get average 70C on thermistors it can only mean that I have no less than 80 on the RAM itself. It's not totally accurate and reliable, but at least it is something.

Built-in sensors lie too... Right now it is around 30C in my house and just started to cool down outside.
I'm struggling at the workdesk in my underwear, but my GPU is trying to convince me that its core temperature is 32C

 

[Aquinus]

But, I can say that the hot chip will throttle before reaching the point where temps may damage it, so doing such a task would be futile, nevermind it would probably take the better part of a decade.

I believe we're talking about reducing the life span of CPUs as well, not just instantly killing it. Thermal limits may keep heat from killing it instantly, but it very well could degrade over time when exposed to larger temperature changes than under stock conditions. People see this when it takes more voltage to hold the same clock. We've seen that and it happens. You're right though, a CPU in most motherboards will shut itself off well before it gets so hot to instantly destroy it. My point was that bigger temperature fluctuations from over-volting are going to cause it to degrade faster than running it at stock given the swing in temperature is larger and you're not exceeding the breakdown voltage of the transitions in the CPU.

If you want to kill a CPU instantly, you get a hammer... or a firearm... or a _________ .

 

[cadaveca]

Throttling is bad, if your CPU needs to throttle itself, you need better cooling, so I was thinking about safe high temps vs low temps (let's say 80C vs 40C), but yes, obviously this is something what can't really be tested because it would take a decade indeed.

Obviously you've never met a stock-cooled 4770K. They throttle all the time, as do laptop chips. CPUs running 90c++ was common until we had thermal monitoring, and then people starting cooking eggs on their PCs, and temps all of a sudden became a thing that normal users though about. I have been OC'ing, really, since 586 days, when soldering was the way it was done.

I'm not sure exactly why everyone has an aversion to parts running 100c, or throttling... if they are designed ot do this, then they do it, without any problems.

 

[qubit]

18 years on the p2 300 (SLOT 1 )i bought with the specific intent of overclocking is still running at 550Mhz ( now inherited by my friends daughter for her personal computer)

Epic respect.

Votlage can exist without current, but current CANNOT exist without that difference in voltage.

Ok, just to be pedantic, you technically can have current with zero voltage if the resistance is zero, but this will only be the case with a superconductor, of course.

The calculation V / R=I would then suggest an infinite current flow, but of course this doesn't happen in this universe as there will be other limiting factors, such as internal resistance of the power source and resistance of the non superconducting wires connecting it to that source. And of course, if the whole lot were superconducting, then there would still be a limit to the current flow as the accuracy of that calculation breaks down as a model of nature in this edge case.

 

[LightningJR]

Epic respect.

Ahh i do miss the old days of overclocking where I had no idea what I was doing

My first time overclocking was the Pentium 200Mhz to 233Mhz using the dip switch on my Compaq Deskpro, you talk about about smiles when I saw it actually worked. Hooked straight away, then I needed to find out if I could overclock my S3 Virge, I believe it was possible, downloading Powerstrip on 56K and figuring it all out. Nostalgia at it's best. Wow, I had a 3.1GB HDD back then....I have 32GB of RAM now..