CPU Temp vs Heat: Easy to understand thermodinamic lesson

[samuel]

A CPU that pulls nearly twice as much power runs 20° cooler? Work on your physics as well.

 

Sorry to put you as a bad example but you are the most recent user that claimed this.

Pulled power is totally transformed into heat. This is the universal law of energy conservation. Yes there is some electromagnetism to it but next to nothing in terms of pulled watts

.But heat is not temperature. They are related but there is more to it: surface and time!

Humans managed to get up to  " a whopping 4 trillion degrees Celsius, 250,000 times hotter than the center of the sun." With this kind of heat, the Earth should explode, right? It would if that tiny spark would be the size of a pool and kept alive a second or two.

Another good example is the light bulb: It's burning hot but I cannot use it instead of an radiator to heat up my room. The room represents the cooler and the bulb is a small but high performance CPU!

 

This is why a Haswell CPU has bad temps: The chip surface is so small that any cooler will have a bad time transporting it from that small die, but not dissipating it!

 

A 15C 30T 150W TDP Xeon can be cooled quite easily because of the huge die size provided by LGA2011.

Same exact thing happens with GTX 480: 600W heat at 70 degreeds

 

Please don't start a flame about amd vs intel cpus in this thread! Flame!

[Real_PhillBert]

As a mechanical engineer, your post made my head hurt.

 

I read your post a half a dozen times and I think I see what you are trying to say. Are you referring to Linus's comments on the WAN show a few weeks ago when he was talking about cooling a CPU vs a GPU?

 

Like you mentioned power coming into the PC is (with some small exceptions) all converted into waste heat energy. This is true. The more wattage you pull from the wall, the more heat your cooling system will have to dissipate. Then you go on to talk a bit about "surface and time" I can only assume you are talking about surface area and time. If you assume a constant load on your PC you can effectively ignore time, yes there will be a period of time where your heat sinks are getting up to temp but to make the math easier we can look only at a period of time where the system has reached thermodynamic equilibrium. 

 

Your next paragraph talking about temps is a bit confusing. Why exactly would earth explode? Then you talk about a light bulb being hot but cannot be used to heat up your room. This is not correct. Any heat source will effect the temperature of the surroundings, however a 60w light bulb does not radiate enough energy to make a noticeable impact on the temp of your room. If you had a space heater that was 60 watts it would increase the temp of your room by exactly the same amount. This is because in your light bulb example, your room is acting as the heat sink, a very inefficient one, but a heat sink non the less. It operates exactly the same as the cooler you have on your GPU or CPU.

 

Now lets move on to more PC related examples, like the Haswell temps that you talk about. This is where you get things right again, just not quite for the right reasons. You are thinking of energy as heat again, you cant do that. Haswell produces more heat even though it uses less wattage (energy) because it has a higher energy density. In other words there is more energy (wattage) being converted per square MM than there may be on a LGA 2011 socket. When you have a higher energy density, you have higher temperatures. Now for the cooler "transporting" heat from the CPU and dissipating it, here again you are correct. You have to break the cooler down into different components in order to better examine how this works. In order to get some of the energy out of the CPU your CPU cooler will have a large base to contact as much of the CPU as possible. The problem becomes the heat transfer (k value) between the two surfaces. This happens again between the CPU Coolers base, and the heat pipes. Then again between the heat pipes and the Cooler's heat sink, finally one more time between the heat sink and the surrounding air. The problem is that the transfer is never really efficient. At least not in the context we are discussing here. 

 

Please do not try to compare temps/watt of CPU's to GPU's. GPU's have their own dedicated PCB and usually the cooler on a stock card attaches directly to the GPU core, resulting in much fewer inefficient transitions. In most cases the GPU on the card is in direct contact with the heat sink. Plus since GPU's have their own PCB its more predictable so GPU coolers (heat sinks) can be designed to be larger and still have fewer interference's. 

 

In short lower wattage does mean less heat energy output, but not necessarily lower temps. 

[samuel]

As a mechanical engineer, your post made my head hurt.

 

I read your post a half a dozen times and I think I see what you are trying to say. Are you referring to Linus's comments on the WAN show a few weeks ago when he was talking about cooling a CPU vs a GPU?

 

Like you mentioned power coming into the PC is (with some small exceptions) all converted into waste heat energy. This is true. The more wattage you pull from the wall, the more heat your cooling system will have to dissipate. Then you go on to talk a bit about "surface and time" I can only assume you are talking about surface area and time. If you assume a constant load on your PC you can effectively ignore time, yes there will be a period of time where your heat sinks are getting up to temp but to make the math easier we can look only at a period of time where the system has reached thermodynamic equilibrium. 

 

Your next paragraph talking about temps is a bit confusing. Why exactly would earth explode? Then you talk about a light bulb being hot but cannot be used to heat up your room. This is not correct. Any heat source will effect the temperature of the surroundings, however a 60w light bulb does not radiate enough energy to make a noticeable impact on the temp of your room. If you had a space heater that was 60 watts it would increase the temp of your room by exactly the same amount. This is because in your light bulb example, your room is acting as the heat sink, a very inefficient one, but a heat sink non the less. It operates exactly the same as the cooler you have on your GPU or CPU.

 

Now lets move on to more PC related examples, like the Haswell temps that you talk about. This is where you get things right again, just not quite for the right reasons. You are thinking of energy as heat again, you cant do that. Haswell produces more heat even though it uses less wattage (energy) because it has a higher energy density. In other words there is more energy (wattage) being converted per square MM than there may be on a LGA 2011 socket. When you have a higher energy density, you have higher temperatures. Now for the cooler "transporting" heat from the CPU and dissipating it, here again you are correct. You have to break the cooler down into different components in order to better examine how this works. In order to get some of the energy out of the CPU your CPU cooler will have a large base to contact as much of the CPU as possible. The problem becomes the heat transfer (k value) between the two surfaces. This happens again between the CPU Coolers base, and the heat pipes. Then again between the heat pipes and the Cooler's heat sink, finally one more time between the heat sink and the surrounding air. The problem is that the transfer is never really efficient. At least not in the context we are discussing here. 

 

Please do not try to compare temps/watt of CPU's to GPU's. GPU's have their own dedicated PCB and usually the cooler on a stock card attaches directly to the GPU core, resulting in much fewer inefficient transitions. In most cases the GPU on the card is in direct contact with the heat sink. Plus since GPU's have their own PCB its more predictable so GPU coolers (heat sinks) can be designed to be larger and still have fewer interference's. 

 

In short lower wattage does mean less heat energy output, but not necessarily lower temps. 

 

I was indeed talking about surface area, missed a word there.

 

The earth exploding example is about time component: the heat radiated by a 250 000 warmer than the sun 's core will be at least dangerous when the surface area is macroscopic on a undetermined period. But our spark is too small and most importantly, is not maintained!

 

Heat is a type of energy. Right?

 

60W Light bulb is hot (3500 Degreeds hot) but the heat emanated is not enough to heat the room considerably. Just like 4770k is too small and too isolated to be cooled as efficient as that 480.

I did not mention anything related to heatpipes and cpu heatspreader. I wanted to keep it simple. CPU,GPU,NB all of them are heat sources that need cooling. Doesn't matter if it's suspended on a dedicated PCB or PVC! That guy now knows that he was wrong saying what he said and that's all to it!

 

 

 

Feel free to post equations and theorems. I'm a crane operator, I know my stuff you know yours.

[Faa]

That quote was my response to this guy
 

Very good points. It's true that games really don't stress the CPU that much at all.
My 8320 is happy to run at 4.6Ghz @ 50c while playing Crysis 3. My 2600 overclocked to 4.4Ghz ran 20 degrees hotter than that in Crysis 3.
And Sandy Bridge was the coolest of the bunch compared to Haswell & Ivy.

He claims his 2600 runs at 70° and his 8350 at 50° with a higher clock speed which uses a shitload more power. Thing is a 2600K at 4.3GHz with 1.25V I could keep the temps below 70° when stressing with prime95 small fft's under a stock cooler

[samuel]

 

That quote was my response to this guy

 

 

He claims his 2600 runs at 70° and his 8350 at 50° with a higher clock speed which uses a shitload more power. Thing is a 2600K at 4.3GHz with 1.25V I could keep the temps below 70° when stressing with prime95 small fft's under a stock cooler

 

 

I know that. What I want you to understand is that even if the heat released (or power consumed if you want) is higher on the AMD side, the temps can be higher on intel because of all what that i said in OP.

[Faa]

I know that. What I want you to understand is that even if the heat released (or power consumed if you want) us higher on the AMD side, the temps can be higher on intel because of all that i said in OP.

I know, sandy bridges are soldered but Haswell & Ivy aren't and that gives us a complety different story than "cpu using more power -> more heat". If he said that his 8350 runs cooler than his haswell, yeah he is perfectly right and I wouldn't even quote him on that.

Basically Haswells that run at 150W (measured from the source) would run eg at 90° orsomething but if you'd take a 3930K/8350 and make it run at 130W it would just run somewhere around 40-50°. 

[samuel]

I know, sandy bridges are soldered but Haswell & Ivy aren't and that gives us a complety different story than "cpu using more power -> more heat". If he said that his 8350 runs cooler than his haswell, yeah he is perfectly right and I wouldn't even quote him on that.

Basically Haswells that run at 150W (measured from the source) would run eg at 90° orsomething but if you'd take a 3930K/8350 and make it run at 130W it would just run somewhere around 40-50°. 

My 8320 will make the highest end noctua air cooler hot to the touch at 1.5v.

A 4770 will burn its internals at 1.5v  but will not make the highest end nokiatua hot to the touch because of failing to move that relatively small heat to the cooler. Or most likely will throttle.

[Faa]

My 8320 will make the highest end noctua air cooler hot to the touch at 1.5v.

A 4770 will burn its internals at 1.5v  but will not make the highest end nokiatua hot to the touch because of failing to move that relatively small heat to the cooler. Or most likely will throttle.

Same on my 3930K, just means you have lots of heat being transfered. Haswell barely puts any heat into your cooler and thats why I never recommend anything better than a evo 212 for them.

[samuel]

-

red your post more carefully and I must say that I agree with almost everything (and yes, you are better than me at explaing the phenomen) but my own examples are not that bad I think. 

[Real_PhillBert]

red your post more carefully and I must say that I agree with almost everything (and yes, you are better than me at explaing the phenomen) but my own examples are not that bad I think. 

 

I think we are really trying to say the same thing, just two different ways of going about it.