Why are radiators for water cooling systems so huge?

[heimdali]

Hi,

 

first, Linus, I would like to thank you for making all these great videos and putting so much effort into them and for all you´re sharing with us.  (And yes, I just signed up because I really wanted to ask this question ...)

 

Now I have a question:  Why are the radiators used for water cooling so huge and yet have what appears to me as extremely poor performance?

 

Many years ago, I tried out water cooling and created a self-made system that used an aquarium pump, a heat exchanger from the cabin heater of a car and a self-made water block which was basically a block of aluminium with three holes drilled through it through which the water was flowing.  Readily made parts just started to become barely available.  That must have been like 20 years ago.  (I deprecated water cooling when I took the machine out of service because it´s just too much of a hassle and has serious reliability issues.)

 

I thought I needed to have a fan on the radiator and found that I didn´t.  The system got slightly warm to the touch without any fan, and it was running 24/7, even at ambient temperatures around 35C.

 

Quite a few years later, I had an air heater in my truck that was rated for an output of 5kW.  For those who don´t know about these, they are very cool devices that burn fuel they take from the tank of the vehicle to heat up an air stream which is then used to heat up the interior without the engine running.  There also versions that heat the engine coolant instead.

 

Based on that experience, I would rate the heat exchangers used in cabin heaters of cars capable of an output of between 5 and 10kW.  The heat exchanger I used for the water cooling was somewhat smaller than a radiator you could fit two 120mm fans on.

It was a very simple design, using thin aluminium fins to expand the surface of maybe three or four aluminium tubes through which the water would flow from side to side.

 

So I´m wondering about two things:

 

 

1.)  What is actually the heat dissipation capacity of the radiators used for water cooling?

I am aware that the heat dissipation capacity does vary with the amount and temperature of the water and air flowing through the radiator.  I would like to see a test comparing different radiators used for water cooling with a couple radiators used in cabin heaters of cars --- and for that, the variables that determine the heat dissipation capacity of the radiators would need to be the same with all of them in order for them to be comparable.  I´ll leave open to debate how to measure exactly how much heat a radiator dissipates.

 

2.)  Well, 1.) kinda leads to this point:  What is the efficiency of a given radiator?

I suppose the efficiency of a radiator could be defined as a relation between the effort required to make the radiator dissipate a given amount of heat at various temperature differences and absolute temperatures of the mediums involved.  The "effort required" could be something like the electricity required to drive the pumps and fans.  Different radiators probably each have their own sweet spot at which the balance between "effort required" and heat dissipated is at a point that could be considered an optimum. --- Considering that a vast number of possible combinations of radiators, fans and pumps are available, a ton of videos could be made about this.

 

 

And I make a bold bold claim here:  The radiators used for computer water cooling solutions are extremely inefficient and have very poor cooling performance compared to other radiators, like those used for cabin heaters in cars.

 

Who of you has a computer the water-cooled CPU of which puts out even 3kW of heat?  Who would seriously try to get the interior of their car cozyly warm with a heater that has only 200W or maybe even 500W, especially when heating up realtively large amounts of ambient air from maybe -5C to 23C, constantly for hours while driving?

 

So what the heck are you guys doing with your water cooling systems that you get the temperatures so high and apparently have trouble fitting enough fans onto enough radiators to keep them in check?

 

Edited October 18, 2017 by heimdali

[Sakkura]

You need to take into account that CPU coolers (and GPU coolers) are transporting heat away from a very small heat source. That's more difficult than transporting it away from a larger heat source.

[heimdali]

Just now, Sakkura said:

You need to take into account that CPU coolers (and GPU coolers) are transporting heat away from a very small heat source. That's more difficult than transporting it away from a larger heat source.

Yes, that is very true.  The radiator itself is not the part that transports the heat away.

 

Still I guess you might want to use a radiator capable of dissipating as much heat as the water block puts into the coolant without needing a fan (fans are cool but comsume electricity, cost money, can be noisy, can fail and may introduce a lot of dust over time ...).

[W-L]

4 minutes ago, heimdali said:

-SNIP-

It all has to do with surface area from a general standpoint, the more surface are the more potential for cooling. Most design their loops for silent operation having a lower fin density and lower airflow than a heat exchanger.  

[heimdali]

1 minute ago, W-L said:

It all has to do with surface area from a general standpoint, the more surface are the more potential for cooling. Most design their loops for silent operation having a lower fin density and lower airflow than a heat exchanger.  

Well, I just saw one of Linus´ videos where he pointed out the folded fins those radiators have.  The one I used didn´t; it had straight fins, and it was even less high than 1U.  That begs another question:  What´s more efficient, folded fins or straight ones ...

[Damascus]

15 minutes ago, heimdali said:

1.)  What is actually the heat dissipation capacity of the radiators used for water cooling?

About 300-500 watts per 120mm square.  My loop has a theoretical max dispersion of about 2kw with its 360mm and 2 240mm radiators.  They are also passive, not requiring any catalyzing energy (ie gas) to dissipate heat. 

 

15 minutes ago, heimdali said:

That must have been like 20 years ago.

 

I thought I needed to have a fan on the radiator and found that I didn´t.  The system got slightly warm to the touch without any fan, and it was running 24/7, even at ambient temperatures around 35C.

Your old CPU probably had a power draw of around 35-50 watts max, a 7700K clocked high could take 150-200+

A 7980XE or high core TR4 can take 600W, 12 times as much as the hottest Pentium and 60 times as much as one of the older ones.  No wonder it takes more to run! 

 

[cj09beira]

the reason is pretty simple, in cars the temp of the engine is much higher, so heat transfer is easily done,  so in cpus its harder too keep them cool, cpus also have much smaller area helping making things worse

[W-L]

1 minute ago, heimdali said:

Well, I just saw one of Linus´ videos where he pointed out the folded fins those radiators have.  The one I used didn´t; it had straight fins, and it was even less high than 1U.  That begs another question:  What´s more efficient, folded fins or straight ones ...

It's more complex than that, it depends on the airflow going through the rad and what application it's in, the accordion folding helps to increase the surface area where the heat from the tubes can interface and transfer to the air, the closer the folds and denser they are the more surface area you have. With that also comes with the downside of restriction which ties back to the airflow going through it and what is able to push through it. 

 

They are essentially two different designs but both achieve the same results, straight fin designs are popular for their ease of manufacture and are widely used.

[heimdali]

3 minutes ago, Damascus said:

About 300-500 watts per 120mm square.  My loop has a theoretical max dispersion of about 2kw with its 360mm and 2 240mm radiators.  They are also passive, not requiring any catalyzing energy (ie gas) to dissipate heat. 

 

Your old CPU probably had a power draw of around 35-50 watts max, a 7700K clocked high could take 150-200+

A 7980XE or high core TR4 can take 600W, 12 times as much as the hottest Pentium and 60 times as much as one of the older ones.  No wonder it takes more to run! 

 

Doesn´t the theoretical maximum heat dissipation depend on factors like the amount of air (which is a gas, too) and water flowing about the radiator?  Of course, you can crank up the flow of coolant only to a point at which the radiator can´t withstand the pressure anymore, and there´s likewise a (technical) limit to the amount of air you can flow about it.  Crank up both, and might get above 2kW.

 

The CPU was an AMD Athlon, which had a bad reputation for requiring lots of power and generating lots of heat.  It was plugged into the board much like you plug in a RAM stick.  I don´t remember exactly how it was called, and I can´t tell how much power it needed.  Ever since the water cooling experiment, I am always very doubtful about reputations like that and claims about CPUs generating incredible amounts of heat

 

[Damascus]

3 minutes ago, heimdali said:

Doesn´t the theoretical maximum heat dissipation depend on factors like the amount of air (which is a gas, too) and water flowing about the radiator?  Of course, you can crank up the flow of coolant only to a point at which the radiator can´t withstand the pressure anymore, and there´s likewise a (technical) limit to the amount of air you can flow about it.  Crank up both, and might get above 2kW.

 

The CPU was an AMD Athlon, which had a bad reputation for requiring lots of power and generating lots of heat.  It was plugged into the board much like you plug in a RAM stick.  I don´t remember exactly how it was called, and I can´t tell how much power it needed.  Ever since the water cooling experiment, I am always very doubtful about reputations like that and claims about CPUs generating incredible amounts of heat

 

To that first bit, I could probably dissipate a max of 2400-2500w based on the spec of my radiators, the power of my pumps (a D5 and some swiftech stuff) and the insanity of my fans (2150rpm gentle typhoons) but it would be loud and negate the reason behind such a loop - quiet.

 

Also, Athlons only took 42 - 60w so a HEDT processor of today is still 10-14 times the power draw

[heimdali]

8 minutes ago, cj09beira said:

the reason is pretty simple, in cars the temp of the engine is much higher, so heat transfer is easily done,  so in cpus its harder too keep them cool, cpus also have much smaller area helping making things worse

See it this way:

 

cabin interior temperature: 23C, coolant temperature: ~88C (which under some circumstances you might want to be lower, and they used to be more like 85C in the past)

ambient temperature: 23C, CPU temperature ~80C (which you probably want to be lower)

 

The point is that the delta of these temperatures is about the same.  So yes, if you want to get the CPU down to 60, it becomes more difficult to do that at the same amient temperature, and you have to look at how well you can transfer the heat from CPU to the waterblock, and it becomes the easier the lower the temperature of the coolant is.

 

What kind of coolant temperatures do you see?

 

[heimdali]

2 minutes ago, Damascus said:

To that first bit, I could probably dissipate a max of 2400-2500w based on the spec of my radiators, the power of my pumps (a D5 and some swiftech stuff) and the insanity of my fans (2150rpm gentle typhoons) but it would be loud and negate the reason behind such a loop - quiet.

 

Also, Athlons only took 42 - 60w so a HEDT processor of today is still 10-14 times the power draw

It´s more a theoretical question, not considering the noise it makes The fans used for cabin heating in cars were around 180 to 250W (and very inefficient).  That may have changed because they have dust filters and all kinds of fancy stuff nowadays.  Imagine putting a 250W fan on your radiator ...  Could you heat a car with it?

 

10--14 times like 420--840W?  Really that much?  I admit that I didn´t buy any HEDT CPU for a long time now.  Aren´t they supposed to have become more energy efficient?

 

If that Athlon used only 40--60W and what you guys have draws that much, I would have to say the opposite and think that these radiators are indeed very efficient.

 

[Damascus]

Just now, heimdali said:

10--14 times like 420--840W?  Really that much?  I admit that I didn´t buy any HEDT CPU for a long time now.  Aren´t they supposed to have become more energy efficient?

That was the range, HEDT maxes out at around 650w while mainstream hangs out between 100-200

[SquintyG33Rs]

you forget that car radiators are much much thicker than watercooling gear. the radiators built for computers are designed to fit into computer cases natively and are thus thinner and slimmer than what cars use or even motorcycles. also the heat output of a vehicle is much larger and they don't need to really turn on their cooling until they reach close to 100C which is absolutely not the case for PCs

[heimdali]

1 minute ago, Damascus said:

That was the range, HEDT maxes out at around 650w while mainstream hangs out between 100-200

With a TDP of 140W or so?  I start to feel misinformed, and/or not up to date at all.  Ugh ...

 

[SquintyG33Rs]

Just now, heimdali said:

With a TDP of 140W or so?  I start to feel misinformed, and/or not up to date at all.  Ugh ...

that's what they are said to use at stock. but since they forced themselves to put xeon level cores at mainstream level speeds they effectively use exponentially more power when even small OC is applied.

stock on all cores they only run in the 2~3GHz range.

[heimdali]

2 minutes ago, SquintyG33Rs said:

you forget that car radiators are much much thicker than watercooling gear. the radiators built for computers are designed to fit into computer cases natively and are thus thinner and slimmer than what cars use or even motorcycles. also the heat output of a vehicle is much larger and they don't need to really turn on their cooling until they reach close to 100C which is absolutely not the case for PCs

The radiators that do the engine cooling are much larger, yet the ones for cabin heaters aren´t.  I didn´t use a radiator for an engine (although I thought about doing that, but it isn´t very practical )

 

[SquintyG33Rs]

Just now, heimdali said:

The radiators that do the engine cooling are much larger, yet the ones for cabin heaters aren´t.  I didn´t use a radiator for an engine (although I thought about doing that, but it isn´t very practical )

 

so moore comparable to the motorcycles ones? i'm not much of a car guy....

[heimdali]

1 minute ago, SquintyG33Rs said:

so moore comparable to the motorcycles ones? i'm not much of a car guy....

I´m not a motorcycle guy, but I think the ones for motorcycles are somewhat larger.

 

On the one I used, you might fit about one-and-a-half 120mm fans, and it was maybe 3/4 of an inch thick.  I got it from a junkyard ... none of the tubing I could get was a good fit, and the connectors couldn´t be replaced ... fun ...

 

[heimdali]

8 minutes ago, SquintyG33Rs said:

that's what they are said to use at stock. but since they forced themselves to put xeon level cores at mainstream level speeds they effectively use exponentially more power when even small OC is applied.

stock on all cores they only run in the 2~3GHz range.

Ok, with that kind of power consumption, I´m getting a whole new perspective on water cooling and these radiators.  And they can´t be inefficient at all when they can handle that.

 

[SquintyG33Rs]

2 minutes ago, heimdali said:

I´m not a motorcycle guy, but I think the ones for motorcycles are somewhat larger.

 

On the one I used, you might fit about one-and-a-half 120mm fans, and it was maybe 3/4 of an inch thick.  I got it from a junkyard ... none of the tubing I could get was a good fit, and the connectors couldn´t be replaced ... fun ...

 

did you see scrapyard wars season 2? or was it 3... where they ghetto watercooled

[SquintyG33Rs]

1 minute ago, heimdali said:

Ok, with that kind of power consumption, I´m getting a whole new perspective on water cooling and these radiators.  And they can´t be inefficient at all when they can handle that.

 

yeah they mostly seem inefficient because thermal dynamics themselves are inefficient at low delta T. and people go overkill on how much radiator they put for their components because they're pursuing aesthetics more than the rest

[heimdali]

5 minutes ago, SquintyG33Rs said:

did you see scrapyard wars season 2? or was it 3... where they ghetto watercooled

Yes, I did, and it actually made some sense to me doing that under the given circumstances because it would seem to potentially give a crucial advantage through overclocking otherwise not possible.

 

Let me add this:  I generally dislike "reality shows" (if I may call it that), but, to my surprise, I really really really like these scrapyard wars.  I couldn´t stop watching, and I´d love to see more of those.

 

And I´ve tried overclocking a few times without any success whatsoever, so I can´t speak to that.  Fortunately, I didn´t fry anything when trying it, either, yet it never ran stable and mostly didn´t run at all.  IIRC, the first CPU I tried to overclock was a 486DX16 ... it just was so slowwww ... I ended up replacing it with DX100, which was also slow, but at least bearable.

 

[heimdali]

17 minutes ago, SquintyG33Rs said:

yeah they mostly seem inefficient because thermal dynamics themselves are inefficient at low delta T. and people go overkill on how much radiator they put for their components because they're pursuing aesthetics more than the rest

They´d fall over if they could see what I had built.  My sole purpose was to get the thing quiet.  I even used two small c-clamps to fix the water block on the CPU, and one of them kept falling off, so I had to decide that one is enough and perhaps even better because it wouldn´t get so heavy as to break the slot the CPU plugged into.  It worked.

 

So ok, my claim is apparently mistaken and false.  Still I´d like to see a comparison as described in the op, even if it might not make much, if any, sense

 

[Damascus]

1 hour ago, heimdali said:

With a TDP of 140W or so?  I start to feel misinformed, and/or not up to date at all.  Ugh ...

 

The 7980XE is an 18 core monster that can OC to 5ghz on all cores, stock TDP is meaningless in this circumstance