What are all the problems with air-cooling? Let's go over what you've stated so far:
Smaller: Then where are you going to put the radiator, radiator fans, and liquid?
Quieter: What about the radiator fans?
9 fans: Yes, but 9 quiet fans.
Able to move more heat: True. However, air-cooling hasn't reached it's limits yet.
There are no serious problems with air-cooling. It cools sufficiently and keeps the G5 reasonably quiet. Considering that we are not likely to see a fanless computer anytime soon (like the cube and second generation CRT iMacs), I agree that it has the best balance of proven reliability, performance, and cost.
But I do think that Apple should consider water cooling in it's future products for one reason: size. As always, correct me if I'm wrong. I am not an engineer.
Here's the way I see things: As we all know, Apple decided to go with 9 quiet fans instead of fewer loud fans in order to cool the G5 quietly. Obviously, each fan takes up some physical space within the aluminum case. And not only do the fans take of space, the 4 cooling zones within the case also requires a certain amount of dead air space to insure an uninterrupted air flow from the front to the back of the case. Otherwise, there wouldn't be any cooling.
Water cooling can centralize all of the computers heat and move it to one radiator with a streamlined cooling tunnel. The tunnel will be extremely efficient. It doesn't have any odd shaped objects such as hard drives or PCI cards that interrupt air flow thus causing noise and reduce cooling. In effect, efficiency of the radiator will allow fewer fans to be used and less dead air space will be needed. Inside the case, the internal components of the computer can be packed closer together which will allow the case to be shrunk to a smaller size.
The biggest problem with water-cooling is the unproven track record and probably cost. However, as I mentioned before, those things tend to get better over time.
One possible and theoretical bonus, the external radiator/cooling tunnel can be easily modified as needed. You don't have to re-engineer the whole motherboard/case just to make a change in cooling requirements. Which means that R&D costs will be cheaper and time to market will be faster.
I am not so sure there is really a "requirement" for a certain amount of "dead space" in the case. As long as the air can flow w/o great restriction, that should be quite sufficient. This is possible because the system is fan-blown in the first place. It is forced convection, thus the air will be forced through the system whether things are in the way or not. Now if it was blowing air through there at 60 mph or there were no fans at all (thus utilizing a fully natural convection effect), then streamlining things would be more pertinent.
If you can cool a chip that's hotter than 2 G5's combined by using water-cooling and 2 fans, wouldn't that benefit the G5's by allowing it clock higher? Or reducing the number of fans? Move the CPU around so the case can be more compact?
Not really. The CPU takes up a lot of the space it takes up partly because its daughtercard is highly elaborate (see the thread "G5 Pictures" for a shot of the daughtercard), and Apple has never factory overclocked CPUs (this is a good thing). There is no PPC equivalent to the old Celerons that were so easy to "overclock" because Intel underclocked them in the factory for marketing reasons. The 970s, at their stated clock speeds, are going as fast as it's safe for them to go - the cooling system is there to see that they're able to go that fast in the first place.
Now, as the fab process continues to shrink, and chips become denser and denser, liquid cooling of CPUs will look more and more attractive: What matters, as far as determining whether liquid cooling is advantageous, is not so much absolute heat as heat density: the ratio between the temperature and the area. Superdense chips will be too small to effectively disperse their heat into a heat sink, so a closed-circuit liquid channel can become invaluable for spreading the heat around to all corners of the heat sink.
Notice that there is no explicit or necessary opposition between liquid cooling and other options. The combination of liquid and a heat sink and a fan is quite attractive.
The main reason the G5's case is so large is so that there will be large channels for air to move through, and the reason for that is quiet. If you shrink the channel sizes without reducing the amount of air that has to move through them in order to cool the components, you increase the noise. The MDD PowerMacs that are held up as a model of internal efficiency sparked outraged boycotts and irate customers because of the amount of noise they generated, so this is not a minor concern. The need to hook up an external FW drive is rarely a deal-breaker, but a computer that's too loud to use for your work always is.
Quote:
But reliability and costs tend to get better over time. I still think there is a possible liquid-cooled Mac in the future. Maybe Apple will take the wait-and-see approach and watch how the NEC and Hitachi turns out. They've done that many times before. They were late adopters of USB2, IDE, PCI, AGP etc. But I've always liked it when Apple was the first on the marketplace (e.g. Wi-Fi, firewire, GUI etc.) If water-cooling catches on, I hope they don't wait too long.
There already are liquid-cooled Macs; the idea is to use what works, where it works. The idea that a PC should be purely "liquid cooled" like the Hitachi PC is - discounting the fact that nothing is ever purely liquid cooled - a marketing gimmick. Look at the cost relative to the feature set.
I share your confidence that Apple will adopt liquid cooling when it makes sense for them to, because they already have, for years now. They just don't consider it a marketing bullet point.
At what cost could it be added by Apple? At what cost could it be added aftermarket?
You can buy aftermarket watercooling kits for certain x86 machines starting from around $170 to around $250. I don't know of any water-cooling kits for Macs.
I have no idea how much it would cost Apple to add watercooling to it's computers but it is likely to be much more integrated and elegant than these aftermarket parts so it may cost more than a x86 watercooler.
You can buy aftermarket watercooling kits for certain x86 machines starting from around $170 to around $250. I don't know of any water-cooling kits for Macs.
I have no idea how much it would cost Apple to add watercooling to it's computers but it is likely to be much more integrated and elegant than these aftermarket parts so it may cost more than a x86 watercooler.
Thanks for the numbers. What I'm getting at is the fact that it is still cost prohibitive at this time. And as long as there are aftermarket solutions, everyone has a solution.
At what cost could it be added by Apple? At what cost could it be added aftermarket?
This is something I don't understand... what part of a watercolling system is so expemsive?
These systems that can be purchased now are so expensive because they are being sold in volume. I honestly see no reason why it would cost more than the fan system. And the benefits are numerous.
BUT... you must use a water based system for this to be cheap... because that fluorinert stuff costs 200 dollars for 1 gallon! What makes this stuff so pricey I would love to know...
Someone estimates you need about 2 gallons for your normal home pc ( I wonder about this myself... I think if you engineered a pc at the factory to be totally compact you could maybe get away with much less fluid).
It's a real shame this stuff is so much money... it really simplifies the whole liquid cooling process.
This is something I don't understand... what part of a watercolling system is so expemsive?
I agree. Current aftermarket fluid-cooling systems are not indicative of the design or costs associated with mass-produced versions. They typically involve tubes, fittings, and reservoirs that could all be replaced by a single machined piece of aluminum no larger than the current G5's heatsinks. Also, a properly designed system should need no more than a few liquid ounces of coolant, certainly not gallons!
Liquid cooling really isn't that complicated. Drill a few channels through an aluminum heat sink, fill it with a tiny bit of coolant, bolt a pump to the side and use a single set screw for the plug. There would be no chance of a leak. Tiny pumps are roughly the same cost as tiny fans. The fluid-filled heatsink would take the same amount of work by a computer controlled mill as the current heatsinks.
The idea just isn't that ludicrous.
Yet I don't think that industry is missing out on anything here. Engineers are perfectly aware of the possibilities of liquid cooling. Occasionally, after weighing the tradeoffs, they conclude that liquid cooling is the way to go and include it on a shipping product. (like some of the apple laptops)
Ironically, the radiator fan in my 1990 accord gave out a few days ago. Thanks to heat transfer from the radiator to the surrounding air... i'm posting this from work today.
Must be a heat pipe... I read somewhere that heatpipes have 100 times the thermal conductivity of diamonds and diamonds have 4 times the thermal conductivity of copper... which is the most conductive metal.
Wow... so this is pretty cool. I was wondering why the heat sink was so big... I figured that a heatsink that large would be pretty ineffectual because of their limitations.
But why is it still so noisy then? Mayb ethe power supply?
A couple questions... where did you get that pic Dfiler? And where is that heatsink located? That isn't one of the largest heatsinks(with the G5 logo on it) is it?
Quote:
also a properly designed system should need no more than a few liquid ounces of coolant, certainly not gallons!
The system I was speaking of was Totally submerged in fluorinnert. No plumbing required. I think this is the most simple solution to the problem. But one of the more expensive.
Imagine a bucket like design similar to the Cube where you slide the bulk of the computer out of the fluid filled bucket.
Must be a heat pipe... I read somewhere that heatpipes have 100 times the thermal conductivity of diamonds and diamonds have 4 times the thermal conductivity of copper... which is the most conductive metal.
Surely, a statement like that has to be in a limited, specific context. It has to be. You will find that the thermal conductivity of water is far less than copper. So that eliminates the possibility that the statement is true in a literal sense (a water-based system being 400x more conductive than copper- sure, why not just invent fusion power, as well). Perhaps if you substituted the term "heat transfer" in place of conductivity, that would be a more applicable statement. That will also be highly dependent upon what the heatpipe is sinking to. In the most extreme cases, I could imagine a heatpipe being that effective (MASSIVE cooling occuring at the other end of that heatpipe). However in typical use, using a typical heatsink at the end of that heatpipe, the heat transfer will simply be as good as you would expect for a humble, air-cooled heatsink. In other words, the capability of that heatpipe will only be as good as the heatsink functioning at the end of it. It is by no means unconditionally 400x "better" just by virtue of it being a heatpipe with water inside it.
[Modern equipment fairly groans with problems like that, and diamonds -- even commercial diamonds -- are too expensive use on a commercial scale. But during the last few decades a new technology has come into use. It's the heat pipe -- a man-made conductor that carries heat a hundred times better than even diamond.
Conductivity refers to a coefficient of inverse resistance in a solid or a non-moving liquid/gas. It is applicable in conjunction with surface area wrt overall heat transfer.
Heat transfer is the end result of heat being moved from one point/regime/medium to another, whether it be by modes of conduction, convection, radiation, or active fluid transfer (in conjunction with convective properties).
Since a heatpipe utilizes a moving medium, thermal conductivity isn't the most applicable term as to how it works and why it does what it does so well. You need to look toward liquid mobility, thermal convection, and phase change to understand why it works the way it does. Given that the interface that typically exists between the heatpipe and the CPU is a copper or aluminum plate, that should be indication enough that the series thermal resistance of the entire system will not be anywhere near 400x the conductivity of copper.
Ironically, the heatpipe in particular can be vulnerable to the very same heat spread issues that you knock air-cooled heatsinks for (since it relies on a solid metal plate as the interface). Perhaps the only saving grace is the phase change behavior of the water inside gives the additional benefit of "locking" the temperature in the 100 deg C range. In that sense, a phase-change device makes for a terrific heat buffer component. Finally, you don't use a heatpipe just to achieve "uber heat transfer property". You use it because there is literally no physical clearance directly above the CPU to accomodate a properly-sized heatsink. So you "pipe" the heat somewhere else where you can accomodate the heatsink.
This is completely aside from your earlier approach involving an active water-circulation system as a means of heat transfer. Naturally, this can be designed to directly address heat spread limitation issues.
The bottomline once again, is that these are all neat things to work with, but they have specific applications where they make the most sense. You don't just use them because they "cool" better (whether or not that is true, indeed) or "have 400x the conductivity".
Also, please be advised that this explanation had very little to do with being "Einstein". Einstein would be good to tell you something about astrophysics, and if you think astrophysics is involved in cooling your CPU, then I don't know what to tell you. I do freely admit that an engineering background enabled me to give the above passage, however. It allows one to cut through the BS far easier than attempting to devine actual knowledge from reading the marketing buzzwords printed on the packaging of your favorite CPU cooler product.
Well, regardless of whether or not that is a G5 daugherboard, that copper tube is almost certainly a heat pipe. From Aavid Thermalloy's web site, I see that a 1/4" diameter heat pipe can carry 100W and distribute it across a large heat sink. This is the kind of thing used in my Pismo, btw.
Also from Aavid Thermalloy's website:
Quote:
A heat pipe is a closed evaporator-condenser system consisting of a sealed, hollow tube whose inside walls are lined with a capillary structure or wick. Thermodynamic working fluid, with substantial vapor pressure at the desired operating temperature, saturates the pores of the wick. When heat is applied to the heat pipe, its fluid heats and evaporates. As the evaporating fluid fills the hollow center of the wick, it diffuses throughout the heat pipe. Condensation of the vapor occurs wherever the temperature is even slightly below that of the evaporation area. As it condenses, the vapor gives up the heat it acquired during evaporation. This effective thermal conductance helps maintain constant temperatures.
Attaching a heat sink to a portion of the heat pipe makes condensation take place at this point of heat loss and establishes a vapor flow pattern. Capillary action within the wick returns the condensate to the evaporator (heat source) and completes the operating cycle. This system, proven in aerospace applications, transmits thermal energy at rates hundreds of times greater and with a far superior energy-to-weight ratio than can be gained from the most efficient solid conductor.
emphasis mine
According to the manufacturer, 400x improvement in thermal transfer does not seem out of the question. And interestingly, the benefit here seems not so much to move the heat from an inaccessible place; from what I read in Avid's site, it looks like the heat pipe can spread the heat across a much larger heat sink than could be used if it were just directly-connected to the processor. Self-pumped, so it's very efficient and inherently silent, and a simple, massive vaned heat sink which can use low-velocity air from a quiet, tuned-down fan to dump the heat out the case. That's what, in engineering, we call "elegant."!
I know that Apple has used liquid cooling in the past (my laptop), and it seems--if this photo is truly of a G5--that they still do. Pretty cool!
Conductivity refers to a coefficient of inverse resistance in a solid or a [i]non-moving liquid/gas. It is applicable in conjunction with surface area wrt overall heat transfer.
What is this BS? lol
Quote:
Heat transfer is the end result of heat being moved from one point/regime/medium to another, whether it be by modes of conduction, convection, radiation, or active fluid transfer (in conjunction with convective properties).
Duh.. your point is?
Quote:
that should be indication enough that the series thermal resistance of the entire system will not be anywhere near 400x the conductivity of copper.
Then what is it's thermal conductivity efficiency compared to copper?
Quote:
Ironically, the heatpipe in particular can be vulnerable to the very same heat spread issues that you knock air-cooled heatsinks for (since it relies on a solid metal plate as the interface).
Except it isn't solid copper... hence -- no heat spread issues.
Quote:
This is completely aside from your earlier approach involving an active water-circulation system as a means of heat transfer. Naturally, this can be designed to directly address heat spread limitation issues.
When did I say a watercooling system could only use a active circulation system? Never. Though one would certainly be more efficient than a heat pipe. But reagardless... a heat pipe is more efficient than a heatsink -- which is what you have been denying from the start.
Heheh, it was fun... sitting there with an image I knew would set you guys off again.
I found it somewhere on the net and dragged it to my desktop. Since then, I haven't been able to find the original source again. Sorry. It does seem to justify apple's advisory not to run the G5s sideways.
Honestly, I really hadn't read up on heat pipes prior to Fawke's excellent post. It seems similar in concept to the liquid cooling systems that I was envisioning. However, the heat pipe cleverly replaces pump or convection driven fluid-flow with condensation and evaporation. Two orders of magnitude better heat transfer means that heatsinks can be more effective, larger in size, and even relocated slightly.
Thoughts on whether heatpipes will become common-place or here-and-then-gone like the fanless cube?
None of that makes sense to you? Then how can you so confidently ascertain what is a thermal conductivity property and what is not?
Quote:
Duh.. your point is?
The point is you aren't even using the terminology properly. So how can you be so sure about what you are really saying. Do you think "heat transfer" and "thermal conductivity" can be used interchangeably in a technical discussion? Honestly, yes or no?
Quote:
Then what is it's thermal conductivity efficiency compared to copper?
It is that of water, technically. What you could say is that the effect of the entire unit is effectively equivalent to a thermal conductivity of xyz... However, that will be based on a number of external factors wrt being true or not. It isn't true all by itself- not by a long shot. What you have done is confuse a statement citing the ultimate capability of a heatpipe under ideal conditions with that of typical conditions. I could say the engine in my car is capable of 300 hp (up from 160) in "race form". Does that then mean that my engine is then a 300 hp engine. No, it is still a 160 hp engine as it has been implemented in my car.
Quote:
Except it isn't solid copper... hence -- no heat spread issues.
So how does it contact the CPU? Air? Magic vacuum? If it isn't copper, then that only makes your situation worse.
Quote:
When did I say a watercooling system could only use a active circulation system? Never. Though one would certainly be more efficient than a heat pipe. But reagardless... a heat pipe is more efficient than a heatsink -- which is what you have been denying from the start.
You are objecting to the wrong element of that statement. Where did I say you can only speak in terms of an active circulation system? I simply referred to your earlier stance when you were talking in terms of an active circulation system, where heat spread actually worked in your favor. You got stomped there, ultimately. So you fished around some (until you acquired that juicy 400x conductivity claim) and decided to rehash heat pipes to support your point that water-cooling is always better. So instead of just conceding that all of these techniques have their circumstances which make them optimal or sub-optimal solutions, you appear to have resigned to endlessly looping around with fragmented pieces of information in hopes that you will find something that can justify your original stance. Instead of understanding the physics behind what makes something good in a particular situation, you just keep coming back with "tantalizing passages" from other people's writings with no regard to the context in which they can be applied or even if they are relevant to your original point.
None of that makes sense to you? Then how can you so confidently ascertain what is a thermal conductivity property and what is not?
Because I know you are just trying to bs your way out of admitting you were wrong. I have quoted a college professors lecture that shows how you were wrong... unless you can prove that you are more qualified than him I will take his word for it. If you are more qualified than him and you are arguing with me about something like this in the manner that you are, then I suggest you go get yourself some medical attention from a "qualified" professional.
Quote:
How can you quantify that 400x number so exactly?
Why does it have to be exact? And I didn't... a college professor that is friends with the guy that invented heat pipes did.
But I never said that this system was that efficient, that is just Randy trying to pick a fight.
Comments
Originally posted by ryaxnb
What are all the problems with air-cooling? Let's go over what you've stated so far:
Smaller: Then where are you going to put the radiator, radiator fans, and liquid?
Quieter: What about the radiator fans?
9 fans: Yes, but 9 quiet fans.
Able to move more heat: True. However, air-cooling hasn't reached it's limits yet.
There are no serious problems with air-cooling. It cools sufficiently and keeps the G5 reasonably quiet. Considering that we are not likely to see a fanless computer anytime soon (like the cube and second generation CRT iMacs), I agree that it has the best balance of proven reliability, performance, and cost.
But I do think that Apple should consider water cooling in it's future products for one reason: size. As always, correct me if I'm wrong. I am not an engineer.
Here's the way I see things: As we all know, Apple decided to go with 9 quiet fans instead of fewer loud fans in order to cool the G5 quietly. Obviously, each fan takes up some physical space within the aluminum case. And not only do the fans take of space, the 4 cooling zones within the case also requires a certain amount of dead air space to insure an uninterrupted air flow from the front to the back of the case. Otherwise, there wouldn't be any cooling.
Water cooling can centralize all of the computers heat and move it to one radiator with a streamlined cooling tunnel. The tunnel will be extremely efficient. It doesn't have any odd shaped objects such as hard drives or PCI cards that interrupt air flow thus causing noise and reduce cooling. In effect, efficiency of the radiator will allow fewer fans to be used and less dead air space will be needed. Inside the case, the internal components of the computer can be packed closer together which will allow the case to be shrunk to a smaller size.
The biggest problem with water-cooling is the unproven track record and probably cost. However, as I mentioned before, those things tend to get better over time.
One possible and theoretical bonus, the external radiator/cooling tunnel can be easily modified as needed. You don't have to re-engineer the whole motherboard/case just to make a change in cooling requirements. Which means that R&D costs will be cheaper and time to market will be faster.
Originally posted by BeigeUser
If you can cool a chip that's hotter than 2 G5's combined by using water-cooling and 2 fans, wouldn't that benefit the G5's by allowing it clock higher? Or reducing the number of fans? Move the CPU around so the case can be more compact?
Not really. The CPU takes up a lot of the space it takes up partly because its daughtercard is highly elaborate (see the thread "G5 Pictures" for a shot of the daughtercard), and Apple has never factory overclocked CPUs (this is a good thing). There is no PPC equivalent to the old Celerons that were so easy to "overclock" because Intel underclocked them in the factory for marketing reasons. The 970s, at their stated clock speeds, are going as fast as it's safe for them to go - the cooling system is there to see that they're able to go that fast in the first place.
Now, as the fab process continues to shrink, and chips become denser and denser, liquid cooling of CPUs will look more and more attractive: What matters, as far as determining whether liquid cooling is advantageous, is not so much absolute heat as heat density: the ratio between the temperature and the area. Superdense chips will be too small to effectively disperse their heat into a heat sink, so a closed-circuit liquid channel can become invaluable for spreading the heat around to all corners of the heat sink.
Notice that there is no explicit or necessary opposition between liquid cooling and other options. The combination of liquid and a heat sink and a fan is quite attractive.
The main reason the G5's case is so large is so that there will be large channels for air to move through, and the reason for that is quiet. If you shrink the channel sizes without reducing the amount of air that has to move through them in order to cool the components, you increase the noise. The MDD PowerMacs that are held up as a model of internal efficiency sparked outraged boycotts and irate customers because of the amount of noise they generated, so this is not a minor concern. The need to hook up an external FW drive is rarely a deal-breaker, but a computer that's too loud to use for your work always is.
But reliability and costs tend to get better over time. I still think there is a possible liquid-cooled Mac in the future. Maybe Apple will take the wait-and-see approach and watch how the NEC and Hitachi turns out. They've done that many times before. They were late adopters of USB2, IDE, PCI, AGP etc. But I've always liked it when Apple was the first on the marketplace (e.g. Wi-Fi, firewire, GUI etc.) If water-cooling catches on, I hope they don't wait too long.
There already are liquid-cooled Macs; the idea is to use what works, where it works. The idea that a PC should be purely "liquid cooled" like the Hitachi PC is - discounting the fact that nothing is ever purely liquid cooled - a marketing gimmick. Look at the cost relative to the feature set.
I share your confidence that Apple will adopt liquid cooling when it makes sense for them to, because they already have, for years now. They just don't consider it a marketing bullet point.
Originally posted by bunge
At what cost could it be added by Apple? At what cost could it be added aftermarket?
You can buy aftermarket watercooling kits for certain x86 machines starting from around $170 to around $250. I don't know of any water-cooling kits for Macs.
I have no idea how much it would cost Apple to add watercooling to it's computers but it is likely to be much more integrated and elegant than these aftermarket parts so it may cost more than a x86 watercooler.
Originally posted by BeigeUser
You can buy aftermarket watercooling kits for certain x86 machines starting from around $170 to around $250. I don't know of any water-cooling kits for Macs.
I have no idea how much it would cost Apple to add watercooling to it's computers but it is likely to be much more integrated and elegant than these aftermarket parts so it may cost more than a x86 watercooler.
Thanks for the numbers. What I'm getting at is the fact that it is still cost prohibitive at this time. And as long as there are aftermarket solutions, everyone has a solution.
At what cost could it be added by Apple? At what cost could it be added aftermarket?
This is something I don't understand... what part of a watercolling system is so expemsive?
These systems that can be purchased now are so expensive because they are being sold in volume. I honestly see no reason why it would cost more than the fan system. And the benefits are numerous.
BUT... you must use a water based system for this to be cheap... because that fluorinert stuff costs 200 dollars for 1 gallon! What makes this stuff so pricey I would love to know...
Someone estimates you need about 2 gallons for your normal home pc ( I wonder about this myself... I think if you engineered a pc at the factory to be totally compact you could maybe get away with much less fluid).
It's a real shame this stuff is so much money... it really simplifies the whole liquid cooling process.
Originally posted by iSegway
This is something I don't understand... what part of a watercolling system is so expemsive?
I agree. Current aftermarket fluid-cooling systems are not indicative of the design or costs associated with mass-produced versions. They typically involve tubes, fittings, and reservoirs that could all be replaced by a single machined piece of aluminum no larger than the current G5's heatsinks. Also, a properly designed system should need no more than a few liquid ounces of coolant, certainly not gallons!
Liquid cooling really isn't that complicated. Drill a few channels through an aluminum heat sink, fill it with a tiny bit of coolant, bolt a pump to the side and use a single set screw for the plug. There would be no chance of a leak. Tiny pumps are roughly the same cost as tiny fans. The fluid-filled heatsink would take the same amount of work by a computer controlled mill as the current heatsinks.
The idea just isn't that ludicrous.
Yet I don't think that industry is missing out on anything here. Engineers are perfectly aware of the possibilities of liquid cooling. Occasionally, after weighing the tradeoffs, they conclude that liquid cooling is the way to go and include it on a shipping product. (like some of the apple laptops)
Ironically, the radiator fan in my 1990 accord gave out a few days ago. Thanks to heat transfer from the radiator to the surrounding air... i'm posting this from work today.
See anything pertaining to this discussion?
Wow... so this is pretty cool. I was wondering why the heat sink was so big... I figured that a heatsink that large would be pretty ineffectual because of their limitations.
But why is it still so noisy then? Mayb ethe power supply?
A couple questions... where did you get that pic Dfiler? And where is that heatsink located? That isn't one of the largest heatsinks(with the G5 logo on it) is it?
also a properly designed system should need no more than a few liquid ounces of coolant, certainly not gallons!
The system I was speaking of was Totally submerged in fluorinnert. No plumbing required. I think this is the most simple solution to the problem. But one of the more expensive.
Imagine a bucket like design similar to the Cube where you slide the bulk of the computer out of the fluid filled bucket.
Originally posted by iSegway
Must be a heat pipe... I read somewhere that heatpipes have 100 times the thermal conductivity of diamonds and diamonds have 4 times the thermal conductivity of copper... which is the most conductive metal.
Surely, a statement like that has to be in a limited, specific context. It has to be. You will find that the thermal conductivity of water is far less than copper. So that eliminates the possibility that the statement is true in a literal sense (a water-based system being 400x more conductive than copper- sure, why not just invent fusion power, as well). Perhaps if you substituted the term "heat transfer" in place of conductivity, that would be a more applicable statement. That will also be highly dependent upon what the heatpipe is sinking to. In the most extreme cases, I could imagine a heatpipe being that effective (MASSIVE cooling occuring at the other end of that heatpipe). However in typical use, using a typical heatsink at the end of that heatpipe, the heat transfer will simply be as good as you would expect for a humble, air-cooled heatsink. In other words, the capability of that heatpipe will only be as good as the heatsink functioning at the end of it. It is by no means unconditionally 400x "better" just by virtue of it being a heatpipe with water inside it.
Perhaps if you substituted the term "heat transfer" in place of conductivity, that would be a more applicable statement
What is the difference between "conduct" and "transfer" Einstein?
http://forums.appleinsider.com/newre...threadid=26847
[Modern equipment fairly groans with problems like that, and diamonds -- even commercial diamonds -- are too expensive use on a commercial scale. But during the last few decades a new technology has come into use. It's the heat pipe -- a man-made conductor that carries heat a hundred times better than even diamond.
Heat transfer is the end result of heat being moved from one point/regime/medium to another, whether it be by modes of conduction, convection, radiation, or active fluid transfer (in conjunction with convective properties).
Since a heatpipe utilizes a moving medium, thermal conductivity isn't the most applicable term as to how it works and why it does what it does so well. You need to look toward liquid mobility, thermal convection, and phase change to understand why it works the way it does. Given that the interface that typically exists between the heatpipe and the CPU is a copper or aluminum plate, that should be indication enough that the series thermal resistance of the entire system will not be anywhere near 400x the conductivity of copper.
Ironically, the heatpipe in particular can be vulnerable to the very same heat spread issues that you knock air-cooled heatsinks for (since it relies on a solid metal plate as the interface). Perhaps the only saving grace is the phase change behavior of the water inside gives the additional benefit of "locking" the temperature in the 100 deg C range. In that sense, a phase-change device makes for a terrific heat buffer component. Finally, you don't use a heatpipe just to achieve "uber heat transfer property". You use it because there is literally no physical clearance directly above the CPU to accomodate a properly-sized heatsink. So you "pipe" the heat somewhere else where you can accomodate the heatsink.
This is completely aside from your earlier approach involving an active water-circulation system as a means of heat transfer. Naturally, this can be designed to directly address heat spread limitation issues.
The bottomline once again, is that these are all neat things to work with, but they have specific applications where they make the most sense. You don't just use them because they "cool" better (whether or not that is true, indeed) or "have 400x the conductivity".
Also, please be advised that this explanation had very little to do with being "Einstein". Einstein would be good to tell you something about astrophysics, and if you think astrophysics is involved in cooling your CPU, then I don't know what to tell you. I do freely admit that an engineering background enabled me to give the above passage, however. It allows one to cut through the BS far easier than attempting to devine actual knowledge from reading the marketing buzzwords printed on the packaging of your favorite CPU cooler product.
Also from Aavid Thermalloy's website:
A heat pipe is a closed evaporator-condenser system consisting of a sealed, hollow tube whose inside walls are lined with a capillary structure or wick. Thermodynamic working fluid, with substantial vapor pressure at the desired operating temperature, saturates the pores of the wick. When heat is applied to the heat pipe, its fluid heats and evaporates. As the evaporating fluid fills the hollow center of the wick, it diffuses throughout the heat pipe. Condensation of the vapor occurs wherever the temperature is even slightly below that of the evaporation area. As it condenses, the vapor gives up the heat it acquired during evaporation. This effective thermal conductance helps maintain constant temperatures.
Attaching a heat sink to a portion of the heat pipe makes condensation take place at this point of heat loss and establishes a vapor flow pattern. Capillary action within the wick returns the condensate to the evaporator (heat source) and completes the operating cycle. This system, proven in aerospace applications, transmits thermal energy at rates hundreds of times greater and with a far superior energy-to-weight ratio than can be gained from the most efficient solid conductor.
emphasis mine
According to the manufacturer, 400x improvement in thermal transfer does not seem out of the question. And interestingly, the benefit here seems not so much to move the heat from an inaccessible place; from what I read in Avid's site, it looks like the heat pipe can spread the heat across a much larger heat sink than could be used if it were just directly-connected to the processor. Self-pumped, so it's very efficient and inherently silent, and a simple, massive vaned heat sink which can use low-velocity air from a quiet, tuned-down fan to dump the heat out the case. That's what, in engineering, we call "elegant."!
I know that Apple has used liquid cooling in the past (my laptop), and it seems--if this photo is truly of a G5--that they still do. Pretty cool!
Conductivity refers to a coefficient of inverse resistance in a solid or a [i]non-moving liquid/gas. It is applicable in conjunction with surface area wrt overall heat transfer.
What is this BS? lol
Heat transfer is the end result of heat being moved from one point/regime/medium to another, whether it be by modes of conduction, convection, radiation, or active fluid transfer (in conjunction with convective properties).
Duh.. your point is?
that should be indication enough that the series thermal resistance of the entire system will not be anywhere near 400x the conductivity of copper.
Then what is it's thermal conductivity efficiency compared to copper?
Ironically, the heatpipe in particular can be vulnerable to the very same heat spread issues that you knock air-cooled heatsinks for (since it relies on a solid metal plate as the interface).
Except it isn't solid copper... hence -- no heat spread issues.
This is completely aside from your earlier approach involving an active water-circulation system as a means of heat transfer. Naturally, this can be designed to directly address heat spread limitation issues.
When did I say a watercooling system could only use a active circulation system? Never. Though one would certainly be more efficient than a heat pipe. But reagardless... a heat pipe is more efficient than a heatsink -- which is what you have been denying from the start.
I found it somewhere on the net and dragged it to my desktop. Since then, I haven't been able to find the original source again. Sorry. It does seem to justify apple's advisory not to run the G5s sideways.
Honestly, I really hadn't read up on heat pipes prior to Fawke's excellent post. It seems similar in concept to the liquid cooling systems that I was envisioning. However, the heat pipe cleverly replaces pump or convection driven fluid-flow with condensation and evaporation. Two orders of magnitude better heat transfer means that heatsinks can be more effective, larger in size, and even relocated slightly.
Thoughts on whether heatpipes will become common-place or here-and-then-gone like the fanless cube?
Originally posted by iSegway
What is this BS? lol
None of that makes sense to you? Then how can you so confidently ascertain what is a thermal conductivity property and what is not?
Quote:
Duh.. your point is?
The point is you aren't even using the terminology properly. So how can you be so sure about what you are really saying. Do you think "heat transfer" and "thermal conductivity" can be used interchangeably in a technical discussion? Honestly, yes or no?
Quote:
Then what is it's thermal conductivity efficiency compared to copper?
It is that of water, technically. What you could say is that the effect of the entire unit is effectively equivalent to a thermal conductivity of xyz... However, that will be based on a number of external factors wrt being true or not. It isn't true all by itself- not by a long shot. What you have done is confuse a statement citing the ultimate capability of a heatpipe under ideal conditions with that of typical conditions. I could say the engine in my car is capable of 300 hp (up from 160) in "race form". Does that then mean that my engine is then a 300 hp engine. No, it is still a 160 hp engine as it has been implemented in my car.
Quote:
Except it isn't solid copper... hence -- no heat spread issues.
So how does it contact the CPU? Air? Magic vacuum? If it isn't copper, then that only makes your situation worse.
Quote:
When did I say a watercooling system could only use a active circulation system? Never. Though one would certainly be more efficient than a heat pipe. But reagardless... a heat pipe is more efficient than a heatsink -- which is what you have been denying from the start.
You are objecting to the wrong element of that statement. Where did I say you can only speak in terms of an active circulation system? I simply referred to your earlier stance when you were talking in terms of an active circulation system, where heat spread actually worked in your favor. You got stomped there, ultimately. So you fished around some (until you acquired that juicy 400x conductivity claim) and decided to rehash heat pipes to support your point that water-cooling is always better. So instead of just conceding that all of these techniques have their circumstances which make them optimal or sub-optimal solutions, you appear to have resigned to endlessly looping around with fragmented pieces of information in hopes that you will find something that can justify your original stance. Instead of understanding the physics behind what makes something good in a particular situation, you just keep coming back with "tantalizing passages" from other people's writings with no regard to the context in which they can be applied or even if they are relevant to your original point.
"Water-cooled" systems ARE more efficient than "air-cooled" systems.
You can keep crying about it and trying to derail the discussion... but the reality is that you are wrong.
I just love the fact that you are now arguing that a heat pipe isn't 400 times as efficient as a solid piece of copper the same size. LMAO!
Tooo funny. Keep crying about it though. It is entertaining.
None of that makes sense to you? Then how can you so confidently ascertain what is a thermal conductivity property and what is not?
Because I know you are just trying to bs your way out of admitting you were wrong. I have quoted a college professors lecture that shows how you were wrong... unless you can prove that you are more qualified than him I will take his word for it. If you are more qualified than him and you are arguing with me about something like this in the manner that you are, then I suggest you go get yourself some medical attention from a "qualified" professional.
How can you quantify that 400x number so exactly?
Why does it have to be exact? And I didn't... a college professor that is friends with the guy that invented heat pipes did.
But I never said that this system was that efficient, that is just Randy trying to pick a fight.