Role of natural convection in silent PC cooling
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
386-sx-33.....that's a 220mm 13blade fan that comes with a number of cases. I'm not 100% certain of the specs. It has no rpm sensor wire.....but apparently turns less than 800rpms@12v. It will start at 4V, and is very quiet across the whole rpm range. Plenty of airflow.....
Issac......That setup has a side intake. Plus the hard drive[s] are soft mounted in such a way that they cannot run on their side. Plus the whole thing is heavy, too heavy to play with like that. Plus the entire setup was designed for an upper exhaust. Believe me......running it on it's side wouldn't prove anything. I'm using that one big intake fan usually under 500rpms....and no exhaust fan except the one rear fan that only turns on about 55C {cpu temp}. It never gets that hot.
Here's what it looks like from the side.....you can see the rear intake on the side cover, and the aux exhaust fan with a plastic deflector to prevent re-circulation.
Issac......That setup has a side intake. Plus the hard drive[s] are soft mounted in such a way that they cannot run on their side. Plus the whole thing is heavy, too heavy to play with like that. Plus the entire setup was designed for an upper exhaust. Believe me......running it on it's side wouldn't prove anything. I'm using that one big intake fan usually under 500rpms....and no exhaust fan except the one rear fan that only turns on about 55C {cpu temp}. It never gets that hot.
Here's what it looks like from the side.....you can see the rear intake on the side cover, and the aux exhaust fan with a plastic deflector to prevent re-circulation.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Issac....there is simply no good way I can think of, to build a case that would have equal potential airflow through it top to bottom, or bottom to top. I think the original experiment I proposed.....a Ninja with the fan on the top or the bottom....would be a better test.
Also....a test with a Ninja's fan blowing upward, or level toward the back, would be an even better test. I don't think I've ever seen a Ninja setup with a fan on top blowing downward. The problem would be to setup such a test where the other case components did not affect the result.
Anyway.....it's hard to imagine a setup where blowing the fans downward with an exhaust at the bottom, would be better than blowing upward with the exhaust at the top......everything else being equal. As I said before, in this thread, natural convection is present and trying to work in a tower case. We should work with it, rather than try to fight it.
Also....a test with a Ninja's fan blowing upward, or level toward the back, would be an even better test. I don't think I've ever seen a Ninja setup with a fan on top blowing downward. The problem would be to setup such a test where the other case components did not affect the result.
Anyway.....it's hard to imagine a setup where blowing the fans downward with an exhaust at the bottom, would be better than blowing upward with the exhaust at the top......everything else being equal. As I said before, in this thread, natural convection is present and trying to work in a tower case. We should work with it, rather than try to fight it.
We all agree hot air rises. The main problem is to determine if it's effective enough.
To make it work best the temp difference must be as big as possible.
That's no problem when using the same logic in a fireplace to get rid of the smoke.
But our aim is to get the case temp at room temp. So no temp difference at all.
So we are killing our own concept. Surely there will be a balance. The min. case temp.
The coolest components are the most heat sensitive. I have $500 experience with that
It was somewhere in the P2 or P3 era. A very standard PC with 3 HDs.
One of them being a then high end IBM DeskStar. It's heat took down all 3 drives.
Surely heat was causing it but I think the main problem was the heat being trapped between the drives.
My fear is that the upward draft isn't strong enough to suck out trapped heat.
I also fear the twirling air effect. The CPU and GPU are generating the most heat in a system.
They both create their own upward draft. Usually on the side and a little higher are the HDs.
Also creating their own airflow. Because the 3 hot air streams have different temps they start to mix.
But also interact. The coolest HD air will drop below the hotter GPU/CPU streams. And now we a have a twirling airflow.
So we end up with air pockets again and HDs will be killed. IMO a case temp of 60C is no real problem for most components.
Their life span will decrease but they usually get replaced far earlier than their lifespan.
But HDs won't survive long in such temps. And their performance will drop while they still are alive.
Every host component they own duct seems the most effective to me.
And when being forced using fans I think they should be pointing down.
The only reason for that is that I think that sound will be muffled when it blows down on the carpet.
To make it work best the temp difference must be as big as possible.
That's no problem when using the same logic in a fireplace to get rid of the smoke.
But our aim is to get the case temp at room temp. So no temp difference at all.
So we are killing our own concept. Surely there will be a balance. The min. case temp.
The coolest components are the most heat sensitive. I have $500 experience with that
It was somewhere in the P2 or P3 era. A very standard PC with 3 HDs.
One of them being a then high end IBM DeskStar. It's heat took down all 3 drives.
Surely heat was causing it but I think the main problem was the heat being trapped between the drives.
My fear is that the upward draft isn't strong enough to suck out trapped heat.
I also fear the twirling air effect. The CPU and GPU are generating the most heat in a system.
They both create their own upward draft. Usually on the side and a little higher are the HDs.
Also creating their own airflow. Because the 3 hot air streams have different temps they start to mix.
But also interact. The coolest HD air will drop below the hotter GPU/CPU streams. And now we a have a twirling airflow.
So we end up with air pockets again and HDs will be killed. IMO a case temp of 60C is no real problem for most components.
Their life span will decrease but they usually get replaced far earlier than their lifespan.
But HDs won't survive long in such temps. And their performance will drop while they still are alive.
Every host component they own duct seems the most effective to me.
And when being forced using fans I think they should be pointing down.
The only reason for that is that I think that sound will be muffled when it blows down on the carpet.
for reference, here is what a pure convection-only PC looks like:
0dBA project Rev. 2
note he overclocked the CPU! HDD temp likely not a problem because he's using 2.5" HDDs.
0dBA project Rev. 2
note he overclocked the CPU! HDD temp likely not a problem because he's using 2.5" HDDs.
2.5" HDs can get to hot too.
Maybe not due to heat generated by themselfs but case temps.
75C is a bit high.
HDs really dislike such temps.
Somehow I think heatpipes are the way forward in silent computing.
Nowadays you see a lot of heat pipes (CPU/GPU/NB) but they all transfer the heat to the wrong place. Inside the case.
I think the best way to do this is to replace the side panel(s) of the case with a cm think aluminium panel. Then connect the heatpipes to the side panel(s).
Maybe not due to heat generated by themselfs but case temps.
75C is a bit high.
HDs really dislike such temps.
Somehow I think heatpipes are the way forward in silent computing.
Nowadays you see a lot of heat pipes (CPU/GPU/NB) but they all transfer the heat to the wrong place. Inside the case.
I think the best way to do this is to replace the side panel(s) of the case with a cm think aluminium panel. Then connect the heatpipes to the side panel(s).
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Convection cooling for hard drives.....a different subject, somewhat anyway. Using laptop drives is a definite advantage. I that last computer photo I posted, I tried something different. I mounted two SATA drives externally in a separate enclosure, totally fan-cooled, with an airflow system completely removed from the main case....no convection cooling.
But I also mounted a 120gb WD laptop drive in the upper chamber (the flip-up lid). This is totally cooled by the exhaust flow from the computer. Note the heat off the PSU is down-stream from the case exhaust. This exhaust is never over the mid-30s.....which is quite acceptable as a laptop drive temperature. There's plenty of airflow up there to keep the drive temps low. And there's enough room for two laptop drives if necessary. I can disconnect the SATA drives, and run the entire computer off the convection-cooled drive, assisted by positive pressure of course.
How would a 3.5" drive perform in that upper chamber? Hard to say, but I'd estimate in the low/mid 40s.
In case you're wondering....the CPU in this setup is a P4-3.4EE. That's the hottest Northwood that was built. As with every setup, if you used a cooler CPU, you could get by with less airflow.
But I also mounted a 120gb WD laptop drive in the upper chamber (the flip-up lid). This is totally cooled by the exhaust flow from the computer. Note the heat off the PSU is down-stream from the case exhaust. This exhaust is never over the mid-30s.....which is quite acceptable as a laptop drive temperature. There's plenty of airflow up there to keep the drive temps low. And there's enough room for two laptop drives if necessary. I can disconnect the SATA drives, and run the entire computer off the convection-cooled drive, assisted by positive pressure of course.
How would a 3.5" drive perform in that upper chamber? Hard to say, but I'd estimate in the low/mid 40s.
In case you're wondering....the CPU in this setup is a P4-3.4EE. That's the hottest Northwood that was built. As with every setup, if you used a cooler CPU, you could get by with less airflow.
you are very far from the first person to have that idea:I think the best way to do this is to replace the side panel(s) of the case with a cm think aluminium panel. Then connect the heatpipes to the side panel(s).
mCubed
HushPC
Zalman TNN
all these things have been around for at least a couple of years. This is old news.
Bluefront, you fundamentally misunderstand the point of such a test. The goal isn't to try and make a case that has equal potential airflow in different directions. The goal is to keep the airflow layout completely the same and just rotate the case to change the direction of the chimney stack effect (by changing the direction of gravity).Bluefront wrote:Issac....there is simply no good way I can think of, to build a case that would have equal potential airflow through it top to bottom, or bottom to top. I think the original experiment I proposed.....a Ninja with the fan on the top or the bottom....would be a better test.
What you propose would tell absolutely NOTHING about how much benefit the stack effect provides. You're proposing changing the airflow layout without changing the direction of the stack effect. That's exactly wrong for determining what benefit the stack effect provides. Instead, you're measuring the benefit of one fan position vs another.
I would be amazed if you could make a homemade variant for less than the new Thermalright cases:So I was thinking about homemade variant.
Thermalright heatpipe cases
<$500 for all the heatpipes and machined aluminium fins?
all you have to do is swap the two around; most CPUs now use less than 75W (much less if undervolted) and even 8800GTX does not use more than 150W in stock form.Besides of that Zalman excels in tight margins.
Max. CPU power 150Watt.
Max. GPU power 75W
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
Actually, the majority of the users here would probably find those specifications high. Mainstream video cards are that power hungry, and efficient, easy to queitly cool processors aren't going to consume 150W. Besides, I don't think there are many Prescotts being bought right now.386-sx-33 wrote:I knew about the Zalman.
That was where my idea comes from.
But their case is rather expensive and ugly.
So I was thinking about homemade variant.
Besides of that Zalman excels in tight margins.
Max. CPU power 150Watt.
Max. GPU power 75W
Especially the GPU is far to low for many setups/
@ryboto
Not Prescotts but Quad cores that are already at 150W now.
Intel is adding cache and ramping up speed. So my guess the power usage will grow.
@jaganath
Very interesting link. I'm going to check their site in August
I didn't know the heat pipes could be swapped around.
I assumed the had fixed positions.
Thanks guys. I learned another thing today.
No lets hope I don't forget another thing
Not Prescotts but Quad cores that are already at 150W now.
Intel is adding cache and ramping up speed. So my guess the power usage will grow.
@jaganath
Very interesting link. I'm going to check their site in August
I didn't know the heat pipes could be swapped around.
I assumed the had fixed positions.
Thanks guys. I learned another thing today.
No lets hope I don't forget another thing
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Issac.....this "stack" effect also works in a fanless ninja mounted in a tower case. Without any fan assistance, the heat rises mostly upward by convection. There is some heat radiation in all directions, but mostly the heat comes out of the top of the heatsink. The addition of a fan on the Ninja blowing upward or downward, and the CPU temps compared, might give you an idea of how well convection helps the cooling of a Ninja/CPU.
Trying that same sort of test in the complete case setup, brings in too many variables to give any meaningful results. We all should know that heat rises off any heat-producing object, trying to go upward......but can also be forced sideways if it encounters an over-head obstruction. That case in the last photo is an example of this. The heat would go upward if it could....but apparently can be forced into a 90 degree turn without much trouble. Now if I were to put the exhaust openings on the bottom of that case, there is no way in Hell the case could cool as well. I would have hot spots/pockets of stagnant hot air throughout the case. Perhaps someone could design a setup with all the exhaust going downward....but why?
The ATX standard itself is based on heat rising, with a PSU on top blowing out the heat. With the onset of very hot computers, this "all exhaust through the PSU" thing doesn't work so well. My design gives you a fairly standard ATX case setup.....with much of the exhaust passing over the PSU, rather than through it. I think it is a perfect way to cool a standard form factor tower computer......making full use of convection.
Trying that same sort of test in the complete case setup, brings in too many variables to give any meaningful results. We all should know that heat rises off any heat-producing object, trying to go upward......but can also be forced sideways if it encounters an over-head obstruction. That case in the last photo is an example of this. The heat would go upward if it could....but apparently can be forced into a 90 degree turn without much trouble. Now if I were to put the exhaust openings on the bottom of that case, there is no way in Hell the case could cool as well. I would have hot spots/pockets of stagnant hot air throughout the case. Perhaps someone could design a setup with all the exhaust going downward....but why?
The ATX standard itself is based on heat rising, with a PSU on top blowing out the heat. With the onset of very hot computers, this "all exhaust through the PSU" thing doesn't work so well. My design gives you a fairly standard ATX case setup.....with much of the exhaust passing over the PSU, rather than through it. I think it is a perfect way to cool a standard form factor tower computer......making full use of convection.
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
Without searching google to see if my suspicions were correct, I thought I'd read that the original intel quads had a 125W thermal envelope...and it has since been decreasing.386-sx-33 wrote:@ryboto
Not Prescotts but Quad cores that are already at 150W now.
Intel is adding cache and ramping up speed. So my guess the power usage will grow.
Bluefront, there are NOT a bunch of variables involved. Rotating a computer case by 90 degrees or 180 degrees changes exactly one variable--the direction of gravity.
Why bother doing it, when we all "know" that the case will be most efficient when the exhausts are higher up than the intakes? The whole point of the experiment is to figure out HOW MUCH of a benefit there is!
In order to figure out how much of a benefit there is, you need two data points--one with the benefit, and one without.
Why bother doing it, when we all "know" that the case will be most efficient when the exhausts are higher up than the intakes? The whole point of the experiment is to figure out HOW MUCH of a benefit there is!
In order to figure out how much of a benefit there is, you need two data points--one with the benefit, and one without.
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
i hate to argue trivialities, but you do mean to say, "change the direction of gravity with respect to the orientation of the case"...cause you can't really change gravity..right?IsaacKuo wrote:Bluefront, there are NOT a bunch of variables involved. Rotating a computer case by 90 degrees or 180 degrees changes exactly one variable--the direction of gravity.
Looks you are right.ryboto wrote:Without searching google to see if my suspicions were correct, I thought I'd read that the original intel quads had a 125W thermal envelope...and it has since been decreasing.386-sx-33 wrote:@ryboto
Not Prescotts but Quad cores that are already at 150W now.
Intel is adding cache and ramping up speed. So my guess the power usage will grow.
The highest TDP I can find is 130W.
Lets hope it keeps going down!
Actually you can.ryboto wrote:i hate to argue trivialities, but you do mean to say, "change the direction of gravity with respect to the orientation of the case"...cause you can't really change gravity..right?IsaacKuo wrote:Bluefront, there are NOT a bunch of variables involved. Rotating a computer case by 90 degrees or 180 degrees changes exactly one variable--the direction of gravity.
In Australia hot air rises down
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
there's too many of you, who really thought it made sense to name a bunch of god like beings the same thing!? he's Q, she's Q, we're all Q! couldn't they have at least added a second letter, or maybe a number!? in either case, I was unaware of your abilities, I apologize.IsaacKuo wrote:Actually, no. I'm really Q and I honestly thought you could just change the universal constant of gravity.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
I was lucky enough to hear one running at Best Buy. It just had all the fans running on high I guess.....no MB however. I could definitely isolate the sound of the upper fan running like that.....loud.
You'd think that case would run fairly good with positive pressure, all the front vent holes closed off, and the upper fan removed. I'll bet that case with a large fan on the side instead of the top.....would do ok. I just hate buying an expensive case I knew I would chop up.
You'd think that case would run fairly good with positive pressure, all the front vent holes closed off, and the upper fan removed. I'll bet that case with a large fan on the side instead of the top.....would do ok. I just hate buying an expensive case I knew I would chop up.
I just did a simple test of my Antec 900 system. It's in more or less stock configuration, but with fans undervolted below the Tricool's normal minimum settings. The PSU is fanless, so it acts as a de facto intake.
Whether the case was on the side or upright, the CPU temperature was 40C. If the stack effect is having any effect, it's less than 1 degree's worth of difference for this system.
I didn't expect any difference, though. That 200mm fan moves so much air that even at its minimum speed the exhaust air feels cool. Without air significantly hotter than ambient, the chimney stack effect is going to have a negligible effect.
Whether the case was on the side or upright, the CPU temperature was 40C. If the stack effect is having any effect, it's less than 1 degree's worth of difference for this system.
I didn't expect any difference, though. That 200mm fan moves so much air that even at its minimum speed the exhaust air feels cool. Without air significantly hotter than ambient, the chimney stack effect is going to have a negligible effect.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
With sufficient exhaust from that 200mm fan, it probably wouldn't matter which direction the case was sitting. Now a real test for that setup would be to turn off/ remove the 200 mm fan, blow the other fans inward, and then see what happens to the various temperatures when you flip the case on it's side.
Even then, the test results might be suspect, depending on the other component arrangement in the case. And even if the results were inconclusive, the test might prove that you don't need a noisy 200mm exhaust fan on top.
Even then, the test results might be suspect, depending on the other component arrangement in the case. And even if the results were inconclusive, the test might prove that you don't need a noisy 200mm exhaust fan on top.
While I wouldn't call the 200mm fan noisy (not at the level I'm undervolting it at), I absolutely know I don't need it. My file server has similar hardware as my main workstation and HTPC--which is cooled just fine with a single severely undervolted 80mm fan. (The HTPC gets no benefit from the chimney stack effect--it's a horizontal case.)
The cooling airflow in my file server is massive overkill. My objective is maximum reliability and hardware lifetime, not minimum noise level.
The cooling airflow in my file server is massive overkill. My objective is maximum reliability and hardware lifetime, not minimum noise level.
-
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Well if you could cool the whole setup with one 80mm fan.....that certainly wouldn't be a good test subject for any sort of temperature test. How about setting up the same case with a hotter CPU, and turn off all the intake and exhaust fans.......just leave the CPU fan running (a blow-down type). Then see what happens to the temps with a flipped case.....
to try and bring some sort of conclusion to this thread, I would say that "natural" convection has a relatively small role to play in forced cooling for PCs, as by definition the goal of forced cooling is to keep the delta between the CPU (or GPU or whatever) and ambient as small as possible, whereas if you wanted to maximise convection you would have the opposite goal (as big a delta as possible). although of course as you slow down your fans and reduce the speed of the forced air "free" convection will have a bigger and bigger role, hence our interest in it here on SPCR.