Isolating hot components from each other - is there a point?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Isolating hot components from each other - is there a point?
I have for some time wondered about the different cooling ‘philosophies’ that seem to be circulating. Mainly whether or not to fear having all hot components ‘share’ the same airflow, or isolate them into different airflow areas. I have also heard people voice opinions against the BTX airflow-design claiming that the fact that the fact that all the intake air blows through the CPU heatsink, and therefore would be much hotter than in a traditional system, would be problematic for GFX-cooling.
For example a lot of people advocate ducting the PSU, giving it it’s own airflow path isolated from that of the hdd-gfx-cpu path (for the Antec P180-concept and MiceC in one of his recent projects). The philosophy apparently is that ‘shielding’ the PSU from the hot air from the CPU (mainly) would make it less likely to ramp up in speed - which of course seems logical.
But what I can’t understand is that this isn’t out-weighted by the fact that the PSU would contribute to CPU-airflow with it’s own fan – creating ‘suckage’ over the CPU area – instead of just ‘keeping all of it’s airflow to itself’.
And the same philosophy seems to be behind the critics of the BTX-design. They seem to believe that the fact that the CPU and graphics system ‘share’ the same intake and overall airflow somehow make them hotter / harder to cool than if they were isolated.
On the contrary my intuition would say that placing ALL hot components in the same airflow path, which is as straight as possible (which by the way seems to be what intel tried when designing BTX), would be by far the most efficient way of cooling a computer.
I am no engineer, but in my head concentrating hot components (to a limit of course) and cooling them with the same airflow seems much more efficient (and manageable) than giving each component it’s own. As far as I can see it’s also more or less the philosophy of ATX as well, btw.
Is there something that I’m missing here?
And another question: does the order of the components in the airflow path play any role in the efficiency in a given airflow. For example does it make sense that the CPU should be after the GFX or wise-versa? – I guess someone experienced with water-cooling should be able to answer that question.
- I can see an extreme case where the order would logically play a role: I guess a VERY hot processor with a VERY efficient heatsink using a VERY low airflow through it would heat the airflow to a level where it could hypothetically be actually HEATING a very cool hard drive…
For example a lot of people advocate ducting the PSU, giving it it’s own airflow path isolated from that of the hdd-gfx-cpu path (for the Antec P180-concept and MiceC in one of his recent projects). The philosophy apparently is that ‘shielding’ the PSU from the hot air from the CPU (mainly) would make it less likely to ramp up in speed - which of course seems logical.
But what I can’t understand is that this isn’t out-weighted by the fact that the PSU would contribute to CPU-airflow with it’s own fan – creating ‘suckage’ over the CPU area – instead of just ‘keeping all of it’s airflow to itself’.
And the same philosophy seems to be behind the critics of the BTX-design. They seem to believe that the fact that the CPU and graphics system ‘share’ the same intake and overall airflow somehow make them hotter / harder to cool than if they were isolated.
On the contrary my intuition would say that placing ALL hot components in the same airflow path, which is as straight as possible (which by the way seems to be what intel tried when designing BTX), would be by far the most efficient way of cooling a computer.
I am no engineer, but in my head concentrating hot components (to a limit of course) and cooling them with the same airflow seems much more efficient (and manageable) than giving each component it’s own. As far as I can see it’s also more or less the philosophy of ATX as well, btw.
Is there something that I’m missing here?
And another question: does the order of the components in the airflow path play any role in the efficiency in a given airflow. For example does it make sense that the CPU should be after the GFX or wise-versa? – I guess someone experienced with water-cooling should be able to answer that question.
- I can see an extreme case where the order would logically play a role: I guess a VERY hot processor with a VERY efficient heatsink using a VERY low airflow through it would heat the airflow to a level where it could hypothetically be actually HEATING a very cool hard drive…
You're essentially correct, but you're oversimplifying the situation. The order in which components are cooled by reused air is critically important.
Each component has two particular statistics which are important here:
1. How hot it can stand operating within.
and
2. How much heat it generates.
A component which demands operating at a cool temperature should obviously be at the "front" of the line for airflow, and should receive fresh air. In particular, hard drives should be cooled by the coolest air. Power supplies are also most reliable when cool, and that's why PSU ducting/separation is popular.
A component which generates a lot of heat warms air a lot so it should be placed at the "back" of the line for airflow. In particular, the CPU tends to generate a lot of heat compared to the other components, as well as the GPU in a gaming computer. It's folly to try and cool components with this pre-heated air.
Traditionally, the PSU has also been a major heat generator, so its exhaust should be immediately exhausted rather than reused. However, many newer power supplies are very efficient, so they exhaust relatively cool air which can potentially be usefully reused. However, it's a challenge to lay out things to reuse this air.
Now, the other thing about separating airflow paths is that it makes it easier to precisely adjust different airflow levels for each hot component. For example, with the GPU, CPU, and PSU in separate airflow paths, it's possible to independently adjust the speeds of three fans to provide each one with exactly how much airflow it needs. You can theoretically acheive the same thing with one airflow path, but it's not so easy in practice.
Each component has two particular statistics which are important here:
1. How hot it can stand operating within.
and
2. How much heat it generates.
A component which demands operating at a cool temperature should obviously be at the "front" of the line for airflow, and should receive fresh air. In particular, hard drives should be cooled by the coolest air. Power supplies are also most reliable when cool, and that's why PSU ducting/separation is popular.
A component which generates a lot of heat warms air a lot so it should be placed at the "back" of the line for airflow. In particular, the CPU tends to generate a lot of heat compared to the other components, as well as the GPU in a gaming computer. It's folly to try and cool components with this pre-heated air.
Traditionally, the PSU has also been a major heat generator, so its exhaust should be immediately exhausted rather than reused. However, many newer power supplies are very efficient, so they exhaust relatively cool air which can potentially be usefully reused. However, it's a challenge to lay out things to reuse this air.
Now, the other thing about separating airflow paths is that it makes it easier to precisely adjust different airflow levels for each hot component. For example, with the GPU, CPU, and PSU in separate airflow paths, it's possible to independently adjust the speeds of three fans to provide each one with exactly how much airflow it needs. You can theoretically acheive the same thing with one airflow path, but it's not so easy in practice.
Efficient how? In terms of noise, energy, economical use of cool air, etc?On the contrary my intuition would say that placing ALL hot components in the same airflow path, which is as straight as possible (which by the way seems to be what intel tried when designing BTX), would be by far the most efficient way of cooling a computer.
The fundamental argument against all components sharing the same airflow is that heat transfer scales proportional to the temperature differential between the cooling fluid (air) and the heated surface (CPU/GPU etc). If you have all components in the same airflow the temperature differential is less, because the air is already heated up, so the heat transfer is less.
http://en.wikipedia.org/wiki/Convection
See the local convective heat flux equation.
I would guess I would define efficient in this context as transfering as much heat as possible with as little noise as possible, which would be the same as getting as much cooling out of the airflow you have as possible.jaganath wrote:Efficient how? In terms of noise, energy, economical use of cool air, etc?On the contrary my intuition would say that placing ALL hot components in the same airflow path, which is as straight as possible (which by the way seems to be what intel tried when designing BTX), would be by far the most efficient way of cooling a computer.
The fundamental argument against all components sharing the same airflow is that heat transfer scales proportional to
the temperature differential between the cooling fluid (air) and the heated surface (CPU/GPU etc). If you have all components in the same airflow the temperature differential is less, because the air is already heated up, so the heat transfer is less.
http://en.wikipedia.org/wiki/Convection
See the local convective heat flux equation.
I don’t believe you’ve got my point. Of course if you give separate the components and give EACH as much airflow as if they were in the same path it’s logical that they would be cooled better because as you say: the difference of temperature between the coolant and the heated surface will be optimised.
But how would the following two setups compare?
1. You divide the CPU, the GPU and the PSU in three different airflow paths with each their own fan and therefore xCFM airflow and zDB noise.
2. You keep the three components together in the same airflow path with 3xCMF airflow and 3zDB noise.
In my head the amount of heat transferred should be exactly the same if we hold all other things equal. I would believe that the heattransfer, y, would be 1y for each of the three airflow paths in system '1' and 3y for the airflow path in system '2'.
In other words as I can see the greater amount of air passing over the three components should outweigh the fact that it has passed over some other first.
I’m not good enough physics myself to be able to see if that is true or false using the formulas for convection so I would actually appreciate it if someone could tell me if my intuision matches the laws of physics - a matematical proof would be greatly appresiated
IsaacKuo
I see you have a point that I hadn’t considered: if everything is in the same airflow path it seems logical to me that the things ‘first’ in the path will be kept coolest and how that can be a problem. I guess you have a good point about being able to adjust the airflow each different component gets.
-
- *Lifetime Patron*
- Posts: 618
- Joined: Mon May 23, 2005 7:05 am
- Location: State College, PA
I think the point jaganath's trying to make is that any given heatsink+fan combination is going to result in X heatrise for Y amount of heat generated, so if the ambient air temperature is higher, then the temperature of the component will be correspondingly higher (ignoring any effect of temperature differential on cooling efficiency. If that higher temperature is still within safe limits, then there's no problem with components sharing airflow. If not, then the HS will need a higher fanspeed to maintain the same temperature which = more noise.
This applies for CPU, PSU, GPU and pretty much any heatsink. So essentially, if all the components are sharing the same air, they will more likely as not require higher airflow to maintain the same temperature, particularly in the low airflow set ups most people on here have. More airflow = more noise. Even though BTX can use less fans than a compartmentalised case those fewer fans will generally have to spin faster to maintain the same temps.
Obviously, all of the above isn't set in stone and certain case/ component/ airflow combinations will work better than others, but it generally holds fast in the quiet computing context.
This applies for CPU, PSU, GPU and pretty much any heatsink. So essentially, if all the components are sharing the same air, they will more likely as not require higher airflow to maintain the same temperature, particularly in the low airflow set ups most people on here have. More airflow = more noise. Even though BTX can use less fans than a compartmentalised case those fewer fans will generally have to spin faster to maintain the same temps.
Obviously, all of the above isn't set in stone and certain case/ component/ airflow combinations will work better than others, but it generally holds fast in the quiet computing context.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
I've been agreement with the one fan/one airflow concept for quite a while....and have made several examples of this concept. Here's one attempt. The rest of the photos in that album show the whole story.
The big problem is modifying the PSU for more efficient airflow, enough to cool the rest of the computer as well as the PSU....and to do so without a noisy fan. The other problem is the HD temps, which will be too high for my liking without a separate fan. Laptop drives solve this problem.
Here's another example. Either of these two computers could operate easily with only one fan, except for the HD temp problem.
The big problem is modifying the PSU for more efficient airflow, enough to cool the rest of the computer as well as the PSU....and to do so without a noisy fan. The other problem is the HD temps, which will be too high for my liking without a separate fan. Laptop drives solve this problem.
Here's another example. Either of these two computers could operate easily with only one fan, except for the HD temp problem.
Hi,
This question has no answer. There is no single right way.
Seriouly.... because the devil is in the detail. The best arrangement depends entirely on exactly how much haet each component happens to create, and how hot it can run.
Just consider the following simple question. If you have a separate air supply for the CPU, will that make the CPU run cooler or hotter?
What will happen is that the air coming into the CPU will be cooler, as nothing else is heating it up first, but (for the same total airflow through the system) the airflow through the CPU heatink must be less, meaning it heats up more when cooling the CPU. meaning the average temperature of the CPU coolant goes up again.
And that's the nub of the problem - just that one change has two effects working in the opposite direction. You need to know whic is the greater; and to do that you have to ask exactly how the airflow through the case is orgainized, and the temperatures/heat output of all the major components.
Another example - suppose you have a machine with one air inlet and two outlets - the PSU and the CPU cooler ducted straight out of the case to stop the hot air re-circulating. Then you partially block the PSU air outlet. The effect will be to reduce total airflow through the case (so the case temp must rise), but also this will increase internal air pressure and force more air through the CPU cooler. The same problem - you have to do the sums on your particular machine to see which effect will dominate.
It is possible to make some simplyfing assumptions and get answers for "typical" systems. Let's consider just one such system: quiet, high powered, with several HD and a modest graphics card - a typical audio production system that I build quite a few of.
The CPU alone will be responsible for over 60% of the heat pruduction and the airflow resistance of the heatsink will be many times greater than all the other airflow resitances put together. Imagine you could change this system to provide the CPU with it's own private air path (straight in and straight out). What would happen?
Let's assume your CPU is producing 100W and you have 18CFM flowing through the heatsink, so the air would gain 10C passing through the heatsink. That means the average air temp in the heatsink (which I call CPU coolant temp) is 6C above case temp, which is itself typically 4C above room temperature. 6C + 4C = 10C, so average CPU coolant temperature is 10C above room temp. If, instead, you had a air at room temperature coming into the CPU heatsink, you'd be able to turn down the CPU fan until you had only 9CFM flowing through the heatsink to maintain the same CPU coolant temp.
Either way the CPU would be 25C-30C above room temperature at full load because the core CPU temp is (upto) 10C above its surface temp and you need 5C-10C tempertaure differnce between the heatsink and the coolant to transfer 100W.
OK... this assumes the effective CPU coolant temperature is the average between the input and output temps. That's an approximation, but you'd need a computer model of the air flow and heat flow through the cooler to do better.
But theh key pont is simple:
Half the airflow = half the decibels = 1/10th of the noise (typically)
The PSU fan is probably powerful enough to produce the airflow needed to cool the rest of the machine. There is no high airflow resistance like the CPU heatink anywhere else. Lets assume thre rest of the machine generates 50W, so 18CFM would hold the case air temp to 5C above room temp...
This design probably leaves you with just two fans - both 120mm running below 800rpm, and the machine running entirely stable at 55C-60C in a room at 30C.
The problem, of course is to provide both a straight in and a straight out airpath for the CPU. You can easily have one or the other, but both is harder. Given that, you have a number of possible options, dominated by the following two points:
1. The PSU fan normally pushes air out of the case
2. After the CPU the graphics card and mobo are the prime sources of heat
And the only way to find out which is best is to do the sort of sums I've shown above using actual data from your system.
Peter
This question has no answer. There is no single right way.
Seriouly.... because the devil is in the detail. The best arrangement depends entirely on exactly how much haet each component happens to create, and how hot it can run.
Just consider the following simple question. If you have a separate air supply for the CPU, will that make the CPU run cooler or hotter?
What will happen is that the air coming into the CPU will be cooler, as nothing else is heating it up first, but (for the same total airflow through the system) the airflow through the CPU heatink must be less, meaning it heats up more when cooling the CPU. meaning the average temperature of the CPU coolant goes up again.
And that's the nub of the problem - just that one change has two effects working in the opposite direction. You need to know whic is the greater; and to do that you have to ask exactly how the airflow through the case is orgainized, and the temperatures/heat output of all the major components.
Another example - suppose you have a machine with one air inlet and two outlets - the PSU and the CPU cooler ducted straight out of the case to stop the hot air re-circulating. Then you partially block the PSU air outlet. The effect will be to reduce total airflow through the case (so the case temp must rise), but also this will increase internal air pressure and force more air through the CPU cooler. The same problem - you have to do the sums on your particular machine to see which effect will dominate.
It is possible to make some simplyfing assumptions and get answers for "typical" systems. Let's consider just one such system: quiet, high powered, with several HD and a modest graphics card - a typical audio production system that I build quite a few of.
The CPU alone will be responsible for over 60% of the heat pruduction and the airflow resistance of the heatsink will be many times greater than all the other airflow resitances put together. Imagine you could change this system to provide the CPU with it's own private air path (straight in and straight out). What would happen?
Let's assume your CPU is producing 100W and you have 18CFM flowing through the heatsink, so the air would gain 10C passing through the heatsink. That means the average air temp in the heatsink (which I call CPU coolant temp) is 6C above case temp, which is itself typically 4C above room temperature. 6C + 4C = 10C, so average CPU coolant temperature is 10C above room temp. If, instead, you had a air at room temperature coming into the CPU heatsink, you'd be able to turn down the CPU fan until you had only 9CFM flowing through the heatsink to maintain the same CPU coolant temp.
Either way the CPU would be 25C-30C above room temperature at full load because the core CPU temp is (upto) 10C above its surface temp and you need 5C-10C tempertaure differnce between the heatsink and the coolant to transfer 100W.
OK... this assumes the effective CPU coolant temperature is the average between the input and output temps. That's an approximation, but you'd need a computer model of the air flow and heat flow through the cooler to do better.
But theh key pont is simple:
Half the airflow = half the decibels = 1/10th of the noise (typically)
The PSU fan is probably powerful enough to produce the airflow needed to cool the rest of the machine. There is no high airflow resistance like the CPU heatink anywhere else. Lets assume thre rest of the machine generates 50W, so 18CFM would hold the case air temp to 5C above room temp...
This design probably leaves you with just two fans - both 120mm running below 800rpm, and the machine running entirely stable at 55C-60C in a room at 30C.
The problem, of course is to provide both a straight in and a straight out airpath for the CPU. You can easily have one or the other, but both is harder. Given that, you have a number of possible options, dominated by the following two points:
1. The PSU fan normally pushes air out of the case
2. After the CPU the graphics card and mobo are the prime sources of heat
And the only way to find out which is best is to do the sort of sums I've shown above using actual data from your system.
Peter
-
- SPCR Reviewer
- Posts: 1850
- Joined: Sun Sep 21, 2003 11:23 am
- Location: Vancouver, BC, Canada
Re: Isolating hot components from each other - is there a po
The key here is that, unless you're using a low-powered CPU and a heatsink with widely spaced fins (think Ninja), the effect of the extra airflow from the PSU is negligible because it isn't high pressure enough to flow between the fins of the heatsink. Air tends to take the path of least resistance, which means, if there's a heatsink in the way, most of it tends to go around the heatsink. There's a reason why CPU heatsinks have a dedicated fan: It's the only way to make certain that enough air flows through the heatsink instead of around it.The I wrote: But what I can’t understand is that this isn’t out-weighted by the fact that the PSU would contribute to CPU-airflow with it’s own fan – creating ‘suckage’ over the CPU area – instead of just ‘keeping all of it’s airflow to itself’.
As far as I can tell, a power supply fan can play a role in evacuating heat from the system, but it won't have much direct effect on the CPU temperature unless the airflow it generates is forced through the heatsink with a duct. In other words, having an extra fan generating airflow in the general vicinity of the CPU heatsink is a poor trade for the addition noise it generates in the PSU because it really has very little effect on providing airflow where it is needed.
Perhaps it would be the most efficient way in terms of heat removed per amount of airflow, but it doesn't take noise into account. Let's assume that there are two ways of doubling airflow through a system. One way is to use two identical fans, the other is to have a single fan spinning twice as fast. All other things being equal (they generally aren't, but bear with me), the two fan configuration is almost certainly going to sound quieter because of the way that mulitple noise sources add up. The accepted rule is that, for each identical noise source that is added to the first, the total noise level goes up by 3 dB (not sure how this is affected by A-weighting). From experience I can tell you that doubling the airflow output of a single fan often increases noise by 10 dBA or more.The I wrote:On the contrary my intuition would say that placing ALL hot components in the same airflow path, which is as straight as possible (which by the way seems to be what intel tried when designing BTX), would be by far the most efficient way of cooling a computer.
However, there are more complications. At the speeds that many SPCR users run them at, our favourite fans are inaudible from one meter. You can add two of these together and the result is still close to inaudible. On the other hand, cranking just one fan up to produce double the airflow is almost certain to be audible. This is a significant subjective difference that cannot be tracked just by looking at the numbers.
Therefore, for practical reasons it is easier to achieve high airflow quietly with multiple fans than with a single fan. This approach lends itself well to isolating components, and does not work well with a linear, BTX-style setup. The reason: The airflow from multiple fans is less focussed and spread over a wider area than the airflow from a single fan. For a linear setup to work, all of the airflow must be channeled into a focussed space that will force all of the air through each of the heatsinks in the chain. Focussing airflow can be achieved through ducting, the ducting introduced impedance that forces the fan to spin more quickly to achieve the same amount of airflow. Airflow from a single fan is already focussed and is thus better suited to a BTX-style setup.
Another issue with a linear airflow setup is the fact that point sources of heat (CPU and GPU) require heatsinks that introduce impedance into the airflow path. Putting them in a line effectively adds the impedance of each heatsink together, requiring more pressure to achieve the same airflow. When the heatsinks are separated, each impedance is dealt with by its own fan, each of which isn't required to produce as much pressure to force the same amount of air through the whole chain.
Yes. The earlier components in the chain effectively increase the ambient temperature for the later components, meaning more airflow is needed to cool the later components to the same level as they would be if they were cooled with fresh air. I believe Isaac Kuo has already mentioned this effect.The I wrote:And another question: does the order of the components in the airflow path play any role in the efficiency in a given airflow. For example does it make sense that the CPU should be after the GFX or wise-versa? – I guess someone experienced with water-cooling should be able to answer that question.
When it comes down to it, I do think that a BTX-style cooling system could be effective for a low-powered system (<100W AC). These systems are not at all difficult to build these days. There's no question that a single low speed fan is quieter than two low speed fans. With big enough heatsinks and linear airflow, it is possible to build a system that is cooled quietly by a single fan. The Shuttle Zen is a good example of this kind of design, although it is atypical because it doesn't have a power supply. Note that even BTX systems are not "one fan" systems; they tend to have a PSU fan in addition to the system fan.
However, for high powered systems with many hot components (discrete GPU), I believe that the low-noise benefits of a modular cooling system outweigh the "efficiency" of a linear system. Take a look at the comparison between the P180 and the SLK3000B for a real world example of how effective a PSU duct can be at reducing noise.
Re: Isolating hot components from each other - is there a po
Or, to put it more succinctly, it's better to have three fans you can't hear than one fan you can.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
matter of opinion....more fans equal more cost, more trouble to mount, more chances for bearing noise, more difficulty to filter, more chances for failure, and so on. We have yet to see the perfect airflow in a case which results in the quietest/coolest computer. But IMHO, that perfect airflow will be achieved with fewer, rather than multiple fans.
This is because of the way fan noise scales quasi-exponentially with fan RPM; so 3 fans at 5V are highly likely to be quieter than one at 12V.Or, to put it more succinctly, it's better to have three fans you can't hear than one fan you can
http://www.cpemma.co.uk/fanlaws.html
Yes, I suppose that is true, but the more fans you have the less likely it is that they will all fail at the same time, so it is less likely that your PC will suffer a catastrophic loss of cooling (although most SPCR comps run so cool a loss of all forced cooling might not destroy the components).more chances for failure
There is inevitably a trade-off between quiet and coolness; after all if we wanted to run the PC as cool as possible we would have some 220CFM monster blasting away at it 24/7; most SPCRer's are happy to tolerate slightly higher temps in order to reduce airflow (and hence noise). The perfect airflow is like the holy grail, no-one can find it....We have yet to see the perfect airflow in a case which results in the quietest/coolest computer.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Again....debatable. To me, it's better to have one fan you cannot hear, rather than three fans you cannot hear. Most electronic devices (other than computers) that need a fan, have only one fan. There can be a fail-safe device to detect fan failure, maybe shut off the device.
IMHO.....there is a perfect airflow for computers, but it's not being used very much. I'm referring to a positive airflow blowing ambient air on the hot spots, with the heated air naturally rising out the top. This technique requires a minimum number of fans, and makes use of natural convection (which is silent), aided by positive case pressure. In this perfect setup, the MB sits horizontal, so the rising heated air does not affect anything else.
Of course, even with this airflow, the Devil is in the details....
IMHO.....there is a perfect airflow for computers, but it's not being used very much. I'm referring to a positive airflow blowing ambient air on the hot spots, with the heated air naturally rising out the top. This technique requires a minimum number of fans, and makes use of natural convection (which is silent), aided by positive case pressure. In this perfect setup, the MB sits horizontal, so the rising heated air does not affect anything else.
Of course, even with this airflow, the Devil is in the details....
Hi,
I have to disagree with this.
For the same aiflow two fans will be running at half the speed of one fan (same size). Typically this will reduce noise levels by 10 times, and the chances of failure or bearing nouise are less because thay are running more slowly. Filtering is easier because the larger total area means lower backpressure for given airflow.
Mounting... OK, depends on whether the case is set up for it.
As for cost - you can get 120mm fans for about $1.50....
Peter
Bluefront wrote:matter of opinion....more fans equal more cost, more trouble to mount, more chances for bearing noise, more difficulty to filter, more chances for failure, and so on. We have yet to see the perfect airflow in a case which results in the quietest/coolest computer. But IMHO, that perfect airflow will be achieved with fewer, rather than multiple fans.
I have to disagree with this.
For the same aiflow two fans will be running at half the speed of one fan (same size). Typically this will reduce noise levels by 10 times, and the chances of failure or bearing nouise are less because thay are running more slowly. Filtering is easier because the larger total area means lower backpressure for given airflow.
Mounting... OK, depends on whether the case is set up for it.
As for cost - you can get 120mm fans for about $1.50....
Peter
Surely the reason for isolating the components is that they all have different cooling requirements that vary independently of one another.
For example, a hard disk must be kept at a much lower temperature than for example a CPU, but conversely needs much less airflow to keep it at that temperature. The heat ouput from a disk is also far more constant than that from a CPU. Whilst it is possible to create a system that adequately cools both, in general this is done by excessive cooling of the disk drives, which is wasted energy which is (ultimately) dissipated as heat.
For example, a hard disk must be kept at a much lower temperature than for example a CPU, but conversely needs much less airflow to keep it at that temperature. The heat ouput from a disk is also far more constant than that from a CPU. Whilst it is possible to create a system that adequately cools both, in general this is done by excessive cooling of the disk drives, which is wasted energy which is (ultimately) dissipated as heat.
Hi Bluefront,
dB is roughly proportional to CFM for any given fan over it's opertaional range.. Take a fan producing 30dB at 12 volts, and run two of them in parallel at 6V. Airflow will be the same, and the noise will drop from 30dB to 18 (15dB +15dB). 12dB less is about 1/10th.
The figures would come out differently for different fans, but 1/10th is a fair average for the sort of setups we are looking at here. It makes the point, which is:
Whatever your setup, using more/larger fans at lower speed to move the same amout of air will always produce less noise.
The trick is to design the layout to minimize the amount of airflow required - absolutely; and then devise the quietest way of moving that amount of air.
It's clearly impossible to cool a computer proding 140+ watts (say) under full load by air with less than c. 10CFM, because the air must heat up 25C to absorb 140W at 10CFM, and some system component(s) - probably the CPU - must be running at a higher temp than the exhaust air.
That puts the CPU surface at 60C under full load conditions with ambient at 30C (which is the target max ambient for fault free operation at 100% load for my machines). I think that's quite hot enough, as the core of the CPU is 10C above surface and the melt poiint could be as low as 80C.
In practice I think that 10 -20 CFM is where we need to be for a quiet but robust system... which leads inexorably to the simple question:
What is the quietest way to move 15CFM +/- ? The answer is two or three big fans at 5V.
Peter
Bluefront wrote:"Typically this will reduce noise levels by 10 times"........just where do you get that figure? But you miss the point......with proper design/airflow, you can get the same cooling, at the same noise level, with fewer fans, the minimum being one.
dB is roughly proportional to CFM for any given fan over it's opertaional range.. Take a fan producing 30dB at 12 volts, and run two of them in parallel at 6V. Airflow will be the same, and the noise will drop from 30dB to 18 (15dB +15dB). 12dB less is about 1/10th.
The figures would come out differently for different fans, but 1/10th is a fair average for the sort of setups we are looking at here. It makes the point, which is:
Whatever your setup, using more/larger fans at lower speed to move the same amout of air will always produce less noise.
The trick is to design the layout to minimize the amount of airflow required - absolutely; and then devise the quietest way of moving that amount of air.
It's clearly impossible to cool a computer proding 140+ watts (say) under full load by air with less than c. 10CFM, because the air must heat up 25C to absorb 140W at 10CFM, and some system component(s) - probably the CPU - must be running at a higher temp than the exhaust air.
That puts the CPU surface at 60C under full load conditions with ambient at 30C (which is the target max ambient for fault free operation at 100% load for my machines). I think that's quite hot enough, as the core of the CPU is 10C above surface and the melt poiint could be as low as 80C.
In practice I think that 10 -20 CFM is where we need to be for a quiet but robust system... which leads inexorably to the simple question:
What is the quietest way to move 15CFM +/- ? The answer is two or three big fans at 5V.
Peter
Hi,
No. That would be true if you had sensors everywhere controlling fans and ensuring that each of these separate components was beiong held, independantly, at the correct temperature.
That route tends to force you into having lots of little fans which is a more noisy solution than a few bigger ones.
In practice, most cooling systems ae much simpler than that anyway, and exploit the fact that most of the comonents (other than the CPU and/or mobster graphic cards) produce a modest and consistant amount of heat. So all you have to do to cool them is produce airflow that is sufficient to keep all of these other compnents at safe temperatures. No component is overcooled because the air is used to cool several components in sequence, and conveniently the last thing in the chain - the PSU - can run at a higher temperature than the HDs and mobo.
Peter
|Romeo| wrote:Surely the reason for isolating the components is that they all have different cooling requirements that vary independently of one another.
For example, a hard disk must be kept at a much lower temperature than for example a CPU, but conversely needs much less airflow to keep it at that temperature. The heat ouput from a disk is also far more constant than that from a CPU. Whilst it is possible to create a system that adequately cools both, in general this is done by excessive cooling of the disk drives, which is wasted energy which is (ultimately) dissipated as heat.
No. That would be true if you had sensors everywhere controlling fans and ensuring that each of these separate components was beiong held, independantly, at the correct temperature.
That route tends to force you into having lots of little fans which is a more noisy solution than a few bigger ones.
In practice, most cooling systems ae much simpler than that anyway, and exploit the fact that most of the comonents (other than the CPU and/or mobster graphic cards) produce a modest and consistant amount of heat. So all you have to do to cool them is produce airflow that is sufficient to keep all of these other compnents at safe temperatures. No component is overcooled because the air is used to cool several components in sequence, and conveniently the last thing in the chain - the PSU - can run at a higher temperature than the HDs and mobo.
Peter
There's always a limit to how many fans is the best, where adding more fans makes the system noisier.pcy wrote:Whatever your setup, using more/larger fans at lower speed to move the same amout of air will always produce less noise.
Personally, I have yet to put together a system which requires more than one 80mm fan running at minimum speed to cool it. I'm not a speed demon, so a Sempron 3100 or a 2.5Ghz Northwood is fast enough for me. For my computers, any more fans or any larger fans would mean more noise.
You can't say that I could replace my one fan with two fans running at lower speeds because the one fan is already running at minimum speed.
Hi Isacc,
Yes... that limit was implicit im my argument. It remains true that if anybody made larger fans that would start up at all under suitable voltages you wouold still get less noise from the bigger/slower route. There is probably some other limit here as well....
I'm surprised/impressed at what you say though... my typical sytems are X2 4600s now (used my musians - they need the power), with 2GB or 4GB RAM plus two or three HDs: and I cannot even begin to get them to stay cool at full load with the room at 30C on the airflow from one 80mm fan at 5V.
Does that include the CPU fan?
Peter
Yes... that limit was implicit im my argument. It remains true that if anybody made larger fans that would start up at all under suitable voltages you wouold still get less noise from the bigger/slower route. There is probably some other limit here as well....
I'm surprised/impressed at what you say though... my typical sytems are X2 4600s now (used my musians - they need the power), with 2GB or 4GB RAM plus two or three HDs: and I cannot even begin to get them to stay cool at full load with the room at 30C on the airflow from one 80mm fan at 5V.
Does that include the CPU fan?
Peter
Yes just one minimum speed 80mm fan. But none of my systems are anywhere near as powerful as yours! I only do shoestring budget computers. I recently got my hands on some cheap Sempron 3100 combos--those were massive upgrades for me.
As for hard drives...my only machine with multiple hard drives is my file server. I while ago I posted my single fan file server, with one 120mm fan, but that's about to change.
My switch to diskless has allowed me to concentrate all of my large 3.5" hard drives into the file server. I'm upping it to two 120mm fans to keep all those 3.5" drives nice and frosty. With all that airflow devoted to the hard drives, the rest of the components are kept cool "for free".
As for hard drives...my only machine with multiple hard drives is my file server. I while ago I posted my single fan file server, with one 120mm fan, but that's about to change.
My switch to diskless has allowed me to concentrate all of my large 3.5" hard drives into the file server. I'm upping it to two 120mm fans to keep all those 3.5" drives nice and frosty. With all that airflow devoted to the hard drives, the rest of the components are kept cool "for free".
We can all pack up and go home, this is basically all of PC silencing in one sentence. Although it's not just the layout that has to be optimised, but choice of CPU/GPU/HDD can also make a huge difference to the minimum required airflow.The trick is to design the layout to minimize the amount of airflow required - absolutely; and then devise the quietest way of moving that amount of air.
Of course there will always be some nutters going for absolute silence and no fans (ie me and justblair), but if ever SPCR were to have a motto or credo this would be it.
Hi Isaac,
Clearly you are trying to build the ultimate in quiet machines, whereas I'm trying to make the fastest machines I can build as quiet as I can.
Both equally vaild objectives, but different.
Have you tried running the machine cooled by the Zalman VGA cooler at 100% CPU for a few days. If so, how hot, relative to ambient, did it get?
I currently use ThretmalRight XP-120s with 140mm fans because the Tower Coolers (Scyth Ninja etc) will not fit in my 4U format cases. Even in free air (rather than in a case) the CPUs need the fan to run about 7-8V to hold them to 55C on 100% load in 30C ambient.
In a sense all this is irrelvant. What I was trying to do was introduce some actual numbers - heatflow and airflow calcualtions - to support the qualititive discussions about isolating the heat producing components. My assumption was that the computer produced 140W-160W on a 24/7 basis and that the exhaust air temp cannot be allowed to go more than 15C over room (or the CPU will overheat). This assumption is a pretty good statement of the requirements my machines must meet; but the reason for making any assumption at all is that you need sone basis in order to produce worked quantitive examples.
Peter
I just had a look at the stuff you posted in the general gallery.IsaacKuo wrote:Yes just one minimum speed 80mm fan. But none of my systems are anywhere near as powerful as yours! I only do shoestring budget computers. I recently got my hands on some cheap Sempron 3100 combos--those were massive upgrades for me.
Clearly you are trying to build the ultimate in quiet machines, whereas I'm trying to make the fastest machines I can build as quiet as I can.
Both equally vaild objectives, but different.
Have you tried running the machine cooled by the Zalman VGA cooler at 100% CPU for a few days. If so, how hot, relative to ambient, did it get?
I currently use ThretmalRight XP-120s with 140mm fans because the Tower Coolers (Scyth Ninja etc) will not fit in my 4U format cases. Even in free air (rather than in a case) the CPUs need the fan to run about 7-8V to hold them to 55C on 100% load in 30C ambient.
In a sense all this is irrelvant. What I was trying to do was introduce some actual numbers - heatflow and airflow calcualtions - to support the qualititive discussions about isolating the heat producing components. My assumption was that the computer produced 140W-160W on a 24/7 basis and that the exhaust air temp cannot be allowed to go more than 15C over room (or the CPU will overheat). This assumption is a pretty good statement of the requirements my machines must meet; but the reason for making any assumption at all is that you need sone basis in order to produce worked quantitive examples.
Peter
-
- *Lifetime Patron*
- Posts: 618
- Joined: Mon May 23, 2005 7:05 am
- Location: State College, PA
I think, as pcy's alluded to, is that Bluefront and IsaacKuo's comments also highlight an important divisor in this argument - how much effort you're willing to go to to achieve the perfect balance of airflow, cooilng and quiet in a PC. I don't doubt that, as Bluefront states, you can cool a mid-range or lower PC with a single low speed fan, but I doubt that more than 1 in a 100 silent PC buffs would be willing to go to the lengths that he does
So personally I think that, whilst it is possible to build a silent PC with only 1 slow fan, it's easier and arguably as effective to build a silent PC by separating the heat producing components and cooling them individually with slow fans. Not one to satisfy the purists admittedly, but certainly good enough for pragmatists
So personally I think that, whilst it is possible to build a silent PC with only 1 slow fan, it's easier and arguably as effective to build a silent PC by separating the heat producing components and cooling them individually with slow fans. Not one to satisfy the purists admittedly, but certainly good enough for pragmatists
I'm trying to make computers which are acceptably quiet. It's not my fault that my standards for "acceptably quiet" have progressively gotten more and more extreme.pcy wrote:Hi Isaac,
I just had a look at the stuff you posted in the general gallery.
Clearly you are trying to build the ultimate in quiet machines, whereas I'm trying to make the fastest machines I can build as quiet as I can.
I don't shun faster performance. I make my computers as fast as I can--but within a budget. My highest priority is spending as little as I can get away with.
Nope. I'm never going to run it at 100% CPU for any significant length of time. If I want computing power, I'm not going to use a mere 564mhz Pentium III when I've got a 2.4Ghz Pentium IV around. Now that I've got three Sempron 3100 machines around also, that PIII is relegated to just being my spare bedroom computer for light web browsing (like right now) and watching non-HDTV videos.Have you tried running the machine cooled by the Zalman VGA cooler at 100% CPU for a few days.
You made a generalization that "Whatever your setup, using more/larger fans at lower speed to move the same amout of air will always produce less noise." That generalization was wrong, because you made assumptions about what every setup would be like.My assumption was that the computer produced 140W-160W on a 24/7 basis and that the exhaust air temp cannot be allowed to go more than 15C over room (or the CPU will overheat). This assumption is a pretty good statement of the requirements my machines must meet; but the reason for making any assumption at all is that you need sone basis in order to produce worked quantitive examples.
You can't use a single example point to prove a generalization.
Anyway, we should all agree that there is no single "right answer" for all setups, and different situations have different priorities.
My file server, main workstation, and HTPC all have similar specs--they're all Sempron 3100's with the same motherboard and similar low power graphics cards. Nevertheless, they're pretty different due to different priorities.
I'm glad you didn't include me in that latter part--I'm rather lazy when it comes to PC mods. I think it shows, especially with my lax cable management and preference for crude cardstock partitions/ducts.mattthemuppet wrote:I think, as pcy's alluded to, is that Bluefront and IsaacKuo's comments also highlight an important divisor in this argument - how much effort you're willing to go to to achieve the perfect balance of airflow, cooilng and quiet in a PC. I don't doubt that, as Bluefront states, you can cool a mid-range or lower PC with a single low speed fan, but I doubt that more than 1 in a 100 silent PC buffs would be willing to go to the lengths that he does
I don't just like to minimize my spending--I also like to minimize my effort! My latest mousepad based hard drive suspension technique is largely a result of wanting something even easier/lazier than the typical suspension techniques.
Hi Isaac,
I regulalrly run my machines for long periods of time at 100%, as do my clients; some of whom use multiple machines to get the computing power they need; so I design the cooling to cope. This moves me well away from the limit you have found where one fan at min speed will produce the airflow to cool the whole system.
Absolutley.
And in particualr the question posed by this thread has no answer, because somtimes it helps to isolate the heat prducing components and sometimes it does not, depending on whether the output air fom one component (in terms of location, airflow and temperature) happens to be suitable to cool another component.
Peter
IsaacKuo wrote:Nope. I'm never going to run it at 100% CPU for any significant length of time. If I want computing power, I'm not going to use a mere 564mhz Pentium III when I've got a 2.4Ghz Pentium IV around. Now that I've got three Sempron 3100 machines around also, that PIII is relegated to just being my spare bedroom computer for light web browsing (like right now) and watching non-HDTV videos.Have you tried running the machine cooled by the Zalman VGA cooler at 100% CPU for a few days.
I regulalrly run my machines for long periods of time at 100%, as do my clients; some of whom use multiple machines to get the computing power they need; so I design the cooling to cope. This moves me well away from the limit you have found where one fan at min speed will produce the airflow to cool the whole system.
No, I made an assumption in order to do an actual worked example to illustrate (not prove) the general principal. The "more/larger at lower speed for same airlow = lower noise" rule has proved true for every combination of fans I have ever tested, and is backed by aerodynamic theory. The principal contains a limit in the phrase "to move the same amount of air"; and you have hit that limit - you cannot obtain fans that run slowly enough to use multiple/larger fans to move only the air volume you want. This does not alter the fact that if such fans existed you could employ them to move the same amount of air with less noise.You made a generalization that "Whatever your setup, using more/larger fans at lower speed to move the same amout of air will always produce less noise." That generalization was wrong, because you made assumptions about what every setup would be like.
You can't use a single example point to prove a generalization.
Anyway, we should all agree that there is no single "right answer" for all setups, and different situations have different priorities.
My file server, main workstation, and HTPC all have similar specs--they're all Sempron 3100's with the same motherboard and similar low power graphics cards. Nevertheless, they're pretty different due to different priorities.
Absolutley.
And in particualr the question posed by this thread has no answer, because somtimes it helps to isolate the heat prducing components and sometimes it does not, depending on whether the output air fom one component (in terms of location, airflow and temperature) happens to be suitable to cool another component.
Peter
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
My approach to computer building/design differs somewhat from Issac ...... I am not interested in building a low-power computer just to make it quiet. Hell I've got a VIA/Eden platform that runs completely fanless. So what?
I use Intel 478 (mostly P-4s) almost exclusively. Plus I pay close attention to temperatures, particularly HD temps, filter incoming air on every one, and stress test extensively. This normally would lead to a noisy computer.....with no undervolting or underclocking.
Run a P4-3.4 at 100% for a while and you'll need more than one 80mm fan. But that doesn't automatically lead to a noisy computer. What it does mean however, is that you will have to pay close attention to your airflow if you want the thing to be quiet at normal usage, and only slightly more audable at 100% usage. That's why most of my projects look/are complicated.
The "Positive Aria" project in the gallery section is an example......runs a P4-3.0 quietly on one 120mm fan. How? By the use of a PICO fanless PSU and a laptop HD, along with carefull execution of a positive airflow design. And with the use of an excellent fan controller, I can run quietly at an idle, and relatively cool at 100%. Overkill? Well maybe, but I like to think I'm just more determined to build the perfect quiet computer.
You can do this with one fan, one airflow.....it can be done cheaply, but not in five minutes. It takes some planning and work.
I use Intel 478 (mostly P-4s) almost exclusively. Plus I pay close attention to temperatures, particularly HD temps, filter incoming air on every one, and stress test extensively. This normally would lead to a noisy computer.....with no undervolting or underclocking.
Run a P4-3.4 at 100% for a while and you'll need more than one 80mm fan. But that doesn't automatically lead to a noisy computer. What it does mean however, is that you will have to pay close attention to your airflow if you want the thing to be quiet at normal usage, and only slightly more audable at 100% usage. That's why most of my projects look/are complicated.
The "Positive Aria" project in the gallery section is an example......runs a P4-3.0 quietly on one 120mm fan. How? By the use of a PICO fanless PSU and a laptop HD, along with carefull execution of a positive airflow design. And with the use of an excellent fan controller, I can run quietly at an idle, and relatively cool at 100%. Overkill? Well maybe, but I like to think I'm just more determined to build the perfect quiet computer.
You can do this with one fan, one airflow.....it can be done cheaply, but not in five minutes. It takes some planning and work.
I don't use low-power computers just to make it quiet. I use low-power computers because I'm cheap. Cost effectiveness comes first, for me. I love bargains!Bluefront wrote:My approach to computer building/design differs somewhat from Issac ...... I am not interested in building a low-power computer just to make it quiet. Hell I've got a VIA/Eden platform that runs completely fanless. So what?
i agree, and ive actually done this concept with a mini-itx computer system.Bluefront wrote:IMHO.....there is a perfect airflow for computers, but it's not being used very much. I'm referring to a positive airflow blowing ambient air on the hot spots, with the heated air naturally rising out the top. This technique requires a minimum number of fans, and makes use of natural convection (which is silent), aided by positive case pressure. In this perfect setup, the MB sits horizontal, so the rising heated air does not affect anything else.
i have a single quiet, undervolted, 80mm fan that blows air directly over the CPU heatsink, the northbridge heatsink, under the motherboard, and over the VGA heatsink all at the same time, without any "reused preheated air". Then the air rises out of the case and out ventalation holes at the back of the case.
It work very well, and is very simple. Also very quiet.
Though it does require all the hot components be very close to one another, and all directly in front of the fan.