Do The Stability Of Clock-Speed Of Cores In A CPU Depend On CPU Temperature

[Kai Vegeta]

If the answer is yes, please can you clarify me that do CPUs have specific point beyond which if we OC the CPU the clock-speed of cores will loose stability even if temperature is in control??

If the answer is no, then what should we do to get stability while achieving higher clock speeds??

 

[qubit]

As the temperature of a CPU (or anything with transistors for that matter) heats up, it's ability to work at high speed is impaired. It's not an on-off process, but a gradual one, which is why CPUs are rated at certain clock speeds and certain max temperatures, which also take into account preventing damage to it.

Hence, a hard overclock combined with increased voltage and high temperatures will shorten the life of the CPU significantly.

 

[Kai Vegeta]

As the temperature of a CPU (or anything with transistors for that matter) heats up, it's ability to work at high speed is impaired. It's not an on-off process, but a gradual one, which is why CPUs are rated at certain clock speeds and certain max temperatures, which also take into account preventing damage to it.

Hence, a hard overclock combined with increased voltage and high temperatures will shorten the life of the CPU significantly.

That means we can oc a CPU as much as we can until the temperature remains in control which is 60-75C and get stable results for all cores

 

[GhostRyder]

If the answer is yes, please can you clarify me that do CPUs have specific point beyond which if we OC the CPU the clock-speed of cores will loose stability even if temperature is in control??

If the answer is no, then what should we do to get stability while achieving higher clock speeds??

Yes, while many chips will have a similar range even if temps hit very low or high (Before hitting a throttle point) too high of temps can cause instability. For instance, Maxwell (Yes a GPU but the same principle for this instance) architecture cards from Nvidia can achieve beyond the usual range of up to 1500ish MHz with extremely low temps (Not as much cramming voltage). Same can be said for CPU's but most of the time as long as you keep temps in a certain range your going to get roughly the same overclocks even if you get the chip lower in temps because voltage requirement become ridiculous after a certain point.

 

[qubit]

Just because the temperature is stable doesn't mean the CPU will overclock much. Also, the temps can be kept down with powerful cooling when lots of voltage is applied. However, the excess voltage will still damage the CPU over time. The original scenario I painted is one of the worst that can happen, but there are lots of inbetween ones.

Rule of thumb is to see how high it will clock with moderate voltage increases and good temps.

 

[BarbaricSoul]

temperature is only one of many factors that limit a CPU's OC'ability

 

[trog100]

an intel cpu will run at 100 C it will simply downclock to stop it going any higher.. the temps and stability are not directly related..

the chip can be super cool and still unstable or "hot" and still stable.. everything else being equal the chip wont be any more stable at 60 C than it is at 80 C..

trog

 

[R-T-B]

I have seen stable chips become unstable when prime95'ing though as soon as they break 90C, so there is some effect from temp.

 

[Kai Vegeta]

Yes, while many chips will have a similar range even if temps hit very low or high (Before hitting a throttle point) too high of temps can cause instability. For instance, Maxwell (Yes a GPU but the same principle for this instance) architecture cards from Nvidia can achieve beyond the usual range of up to 1500ish MHz with extremely low temps (Not as much cramming voltage). Same can be said for CPU's but most of the time as long as you keep temps in a certain range your going to get roughly the same overclocks even if you get the chip lower in temps because voltage requirement become ridiculous after a certain point.

So if I throw in a very heavy water cooling loop I can overclock until it reaches normal temperature. Is EK good.

Just because the temperature is stable doesn't mean the CPU will overclock much. Also, the temps can be kept down with powerful cooling when lots of voltage is applied. However, the excess voltage will still damage the CPU over time. The original scenario I painted is one of the worst that can happen, but there are lots of inbetween ones.

Rule of thumb is to see how high it will clock with moderate voltage increases and good temps.

As you said life will decrease question is how much I only used two Inlet chips one is Pentium 4 which I used for 12years(2000-2012) after which I sold it but chip was still running without problem and other is i3 which I am still using from 6years(2010-still in use). If Intel is providing same quality their chip should do 10year in ease(normal condition) and if we do OC how much will its life drop please can you give me a near estimate because even if it gives me half life which is of five years it will be good as I will upgrade my PC after 5years.

 

[qubit]

I'm sorry, but it's not possible to give lifetime figures like that.

 

[R-T-B]

I'm sorry, but it's not possible to give lifetime figures like that.

Indeed it varies wildly with node, process, ASIC quality, all sorts of crap

 

[Sasqui]

In solid state electronics, there's a phenomenon known as thermal runaway current.

https://en.m.wikipedia.org/wiki/Thermal_runaway

...The behavior of virtually all transistors is affected by temperature.

 

[Kai Vegeta]

In solid state electronics, there's a phenomenon known as thermal runaway current.

https://en.m.wikipedia.org/wiki/Thermal_runaway

...The behavior of virtually all transistors is affected by temperature.

Got it Sasqui and I am providing 16 intake fans specifically for air intake if I build the rig I was talking about in my other thread
https://www.techpowerup.com/forums/threads/will-this-build-be-good-for-5-years.220971/

And the intake air will be nearly 15C as I am setting it up in central air conditioned room so it will always have cool air flowing in all corners of the room so it should solve the problem.

 

[de.das.dude]

This is especially true with AMD chips, they love lower temps. Keep them under low temps and they happily overclock!
keep in mind that overclocking requires bump in voltage which means more heat, so this fact is true only when you use more voltage. if you are trying to use stock voltage it will usually make little difference.

 

[Aquinus]

In solid state electronics, there's a phenomenon known as thermal runaway current.

https://en.m.wikipedia.org/wiki/Thermal_runaway

...The behavior of virtually all transistors is affected by temperature.

Your thinking about the wrong physical property. Read that link you posted yourself:

The leakage current of logic switching transistors increases with temperature. In rare instances, this may lead to thermal runaway in digital circuits. This is not a common problem, since leakage currents usually make up a small portion of overall power consumption, so the increase in power is fairly modest — for an Athlon 64, the power dissipation increases by about 10% for every 30 degrees Celsius.[9] For a device with a TDP of 100 W, for thermal runaway to occur, the heat sink would have to have a thermal resistivity of over 3 K/W (kelvins per watt), which is about 6 times worse than a stock Athlon 64 heat sink. (A stock Athlon 64 heat sink is rated at 0.34 K/W, although the actual thermal resistance to the environment is somewhat higher, due to the thermal boundary between processor and heatsink, rising temperatures in the case, and other thermal resistances.[citation needed].) Regardless, an inadequate heat sink with a thermal resistance of over 0.5 to 1 K/W would result in the destruction of a 100 W device even without thermal runaway effects.

When temperature increases, electron mobility (how quickly current can flow though a medium,) and threshold voltage (when the transistor activates,) gets worse. Improved electron mobility means improved conductivity and improved threshold voltage means that the transistor voltage required to switch the transistor on drops but only is noticeable between two temperatures when the ΔT is relatively large. The benefit of the latter is that since any transistor takes time to switch between an on and off state, starting the transition earlier which leaves more time for the transistor to transition to whatever state it's going to. Pair this with a "more conductive" circuit and you have higher stable switching frequencies. However, if threshold voltage gets too low, the transistor will cold bug iirc.

So the question isn't if you'll overclock better between 60 and 70*C but rather what kind of clocks would be stable at 10*C compared to 60*C. If the OP really wants to get better overclocking, that means a bigger
ΔT between the CPU and the medium that keeps it cool. In order to seriously improve overclocking by taking advantage of physical properties and thermodynamics, you would need to have to work phase-change somewhere into the equation to pump heat out of the cooling system if this is to be a long term thing, which means nothing like DICE or LN2.

I apologize if I did a terrible job of explaining that, I've yet to have my coffee.

I'll leave this here though:
http://www.springer.com/cda/content...407478-c1.pdf?SGWID=0-0-45-1268751-p174130080

tl;dr: Yes but, you probably will only be able to get higher clocks due to temperature by adding phase change to the equation.

 

[Kai Vegeta]

Your thinking about the wrong physical property. Read that link you posted yourself:


When temperature increases, electron mobility (how quickly current can flow though a medium,) and threshold voltage (when the transistor activates,) gets worse. Improved electron mobility means improved conductivity and improved threshold voltage means that the transistor voltage required to switch the transistor on drops but only is noticeable between two temperatures when the ΔT is relatively large. The benefit of the latter is that since any transistor takes time to switch between an on and off state, starting the transition earlier which leaves more time for the transistor to transition to whatever state it's going to. Pair this with a "more conductive" circuit and you have higher stable switching frequencies. However, if threshold voltage gets too low, the transistor will cold bug iirc.

So the question isn't if you'll overclock better between 60 and 70*C but rather what kind of clocks would be stable at 10*C compared to 60*C. If the OP really wants to get better overclocking, that means a bigger
ΔT between the CPU and the medium that keeps it cool. In order to seriously improve overclocking by taking advantage of physical properties and thermodynamics, you would need to have to work phase-change somewhere into the equation to pump heat out of the cooling system if this is to be a long term thing, which means nothing like DICE or LN2.

I apologize if I did a terrible job of explaining that, I've yet to have my coffee.

I'll leave this here though:
http://www.springer.com/cda/content...407478-c1.pdf?SGWID=0-0-45-1268751-p174130080

tl;dr: Yes but, you probably will only be able to get higher clocks due to temperature by adding phase change to the equation.

Thank you very much for spending your time on this.

I will be using 12 rads 4 x 4fan 140mm only for CPU and 8 which will be divided as 2 x 2fan 140mm for each GPU and will be doing 4-way SLI. All the rads will have push-pull configuration. And will we setting up 16 fans for good air intake which will be cold nearly at 15C. I will also be providing separate reservoir-pump for CPU and GPU. So with configuration can I do high overclocking.

 

[Vineet Reddy]

Is this loop even possible?

 

[Kai Vegeta]

Is this loop even possible?

Don't know I have the same question. Most probably it should but not sure

 

[Aquinus]

Thank you very much for spending your time on this.

I will be using 12 rads 4 x 4fan 140mm only for CPU and 8 which will be divided as 2 x 2fan 140mm for each GPU and will be doing 4-way SLI. All the rads will have push-pull configuration. And will we setting up 16 fans for good air intake which will be cold nearly at 15C. I will also be providing separate reservoir-pump for CPU and GPU. So with configuration can I do high overclocking.

The GPU setup makes sense however, instead of going from the last GPU to the pump/res then to the rad, to have a rad after the last GPU then pump directly into the first GPU instead.

The CPU loop makes very little sense. Having 4 rads in series won't help you very much because each stage becomes increasingly less efficient because you're trying to cool already cooled water. The amount of heat that flows out of the system depends on the difference in temperatures (ΔT.) After each radiator the temperature of the liquid in relation to the ambient temperature get smaller so each rad becomes less useful because that difference in temperature is what moves the heat. Also, having the res/pump right after the CPU probably is unwise, you'll be subjecting the pump to more heat than it would be otherwise if it were at the end of the loop (CPU before rads, prior to ending the res.)

Lastly, I would probably opt for a single loop. The GPUs are going to need more cooling capacity than the CPU will, so it makes more sense to allow them to share that cooling capacity.

 

[alucasa]

Looks possible. Feasible? Not so sure.

 

[trog100]

the basic idea of simply adding more rads and fans isnt a valid one.. the weak spot will be getting the heat from the chip into the block.. most of the rads and fans you are dreaming up will be doing pretty much bugger all..

dare i ask exactly what your objective is.. ??

trog

 

[Brusfantomet]

Thank you very much for spending your time on this.

I will be using 12 rads 4 x 4fan 140mm only for CPU and 8 which will be divided as 2 x 2fan 140mm for each GPU and will be doing 4-way SLI. All the rads will have push-pull configuration. And will we setting up 16 fans for good air intake which will be cold nearly at 15C. I will also be providing separate reservoir-pump for CPU and GPU. So with configuration can I do high overclocking.

For the CPU loop i would get one or two BIG radiators (Watercool MO-RA3 or the Aqua computer airplex GIGANT) and only one 460 rad per GPU. What kind of pumps are you planing?

 

[Aquinus]

the basic idea of simply adding more rads and fans isnt a valid one.. the weak spot will be getting the heat from the chip into the block.. most of the rads and fans you are dreaming up will be doing pretty much bugger all..

dare i ask exactly what your objective is.. ??

trog

That's where a higher ΔT between the the heat carrier (liquid and block,) and the heat producer (the CPU,) comes into play. Conduction of heat can never exceed the difference in temperature between the source and the "sink".

@Kai Vegeta: You can never cool something below ambient temperature because heat conduction requires a temperature differential. With that many radiators, you could have a phase change system when an evaporator before the CPU in the liquid loop and a condenser after the first radiator. This has the benefit of cooling the liquid going into the CPU potentially below ambient temperature while using the subsequent radiators to disperse the re-heated air from the phase change loop. Of course, depending on the strength of such a system, you may need to make some form of anti-freeze for your loop and you'll want to cover up anywhere susceptible to condensation or frost but, only a solution like that could possibly offer a long term solution to higher overclocks because it's going to take a lot more than above ambient temperatures to make a serious dent in your overclocking ceiling. You'll need 10*C or lower under load (in my opinion,) for it to truly make a difference. Also keep in mind that this makes your computer depend on a phase change system in order to keep it cold enough to operate properly so, it's a really a lot of work for what I would consider minimal gain.

I don't intend to be mean when I say this but, your idea feels like it's coming from someone with more money than brains. You can't simply throw more of something (like radiators,) at a problem and expect it to be better. You need to simplify to make it more feasible and realistic. While you can do what you're suggesting, it won't get you much further and is simply a waste when there are better ways to spend that money on hardware.

 

[Folterknecht]

an intel cpu will run at 100 C it will simply downclock to stop it going any higher.. the temps and stability are not directly related..

the chip can be super cool and still unstable or "hot" and still stable.. everything else being equal the chip wont be any more stable at 60 C than it is at 80 C..

trog

I have seen stable chips become unstable when prime95'ing though as soon as they break 90C, so there is some effect from temp.

In an inverview with TomsHardware [US] Intel engineers responsible for Devils Canyon development stated, that for these chips anything above 80°C will have a noticable negative effect on the ability to overclock (further).

 

[trog100]

In an inverview with TomsHardware [US] Intel engineers responsible for Devils Canyon development stated, that for these chips anything above 80°C will have a noticable negative effect on the ability to overclock (further).

i think they are saying in a roundabout way that once you hit 85 C its time to stop or think about better cooling.. having said that its quite easy to push a devils canyon chip up to its throttle point of 100 C by running things like prime95..

with a stock cooler even at stock speeds prime95 will take the 4790K temps up to 100 C or very close..

intels own diagnostic software records a pass at 99 C.. one degree below TJ max.. so does real temp..

trog