Cooling Design

[Mist]

I posted this just as an idea on the cooling forum, but since it's of a technical nature, I thought I'd put it here to see what y'all think!

Everyone knows that because of the design of a standard cooling fan, there is a dead spot in the centre where it rotates. Now, on a HSF above a CPU, obviously, the most heat is generated in the dead centre, directly below the dead spot of the fan.

So, because of that design, the airflow doesn't go directly to the main source of heat.

I was thinking last night, (always a dangerous thing for me to do!), but I thought I'd present this here and see what any responses would be.

If a heat spreader was designed starting with a more or less standard 60cm base but then expanding to 80cm at the top where the fan goes, then the user would obviously stick an 80cm fan on it but the actual design of the heat spreader would act like a funnel and force the air more towards the centre where the most heat is immediately generated.

A further idea would be that since most PCs have their PSUs at the top of the case which also incorporates an air extraction process, then would it not be sensible to block of the side of the heat spreader which faces down ways of the motherboard, so that when the HSF is in operation, then the air generated by the cooling fan is forced directly up towards the PSU, instead of going in both directions as happens now?

This would probably also increase the cooling of the PSU since it would receiving cool air more quickly than normal?

Any comment?

Michael.

 

[Smilin]

Some random thoughts on the topic...

ignoring for a second any "dead spots" think about what a downward jet of air does when it hits a flat surface. The spot directly under the jet will have very little air flow - mostly just some turbulence. The areas around the jet will have a very fast airflow, but in a horizontal direction radiating out from where the jet hits. If you are trying to pull the heat from a surface the fast flowing air works better than the turbulent but relatively stationary air at the "impact" point. This model will get really complicated if you change the jet of air to a donut shaped jet coming down from a fan. In short, getting the air to flow over the surface you want, the way you want won't be easy.

Another thing to think about: The metal of a heatsink is where it's at. This is where the real work gets done. A really efficient heatsink (for a brief period of time) would be a solid hemisphere of copper stuck on the cpu. Heat would be pulled from the cpu very very quickly in this scenario. Problem is of course that eventually the metal will heat up and become ineffecient.

Metal is so much better at conducting heat that it doesn't matter quite so much where you pull the heat off the heatsink as long as you get rid of it somewhere. The metal will take care of getting the heat from the cpu just fine...it's getting the heat out of the metal that's the important part - I don't think it would really matter too much about "dead spots" on the heatsink.

 

[Mist]

Thanks for replying, I'm quite interested in this.

In respect of your post, would a better idea be to block off the outlets of a current CPU, so that no air can be drawn in, and have the fan on top extracting the air instead of blowing, bearing in mind that heat rises?

Surely that would be a better method of removing the heat from the metal as quickly as possible?

Michael.

 

[Smilin]

In these circumstances heat doesn't really rise. Any motion due to pressure/heat differences in the air is going to be negligible compared to the flow from fans.

Some heatsinks like some of the alpha's and some others used in 1u and 2u servers pull air from the heatsink instead of blow onto it. It depends on the heatsink.

The real thing you're trying to achive is rapid airflow over a lot of surface. Some clever ideas come to mind but I'm sure heatsink designers could shoot holes in my ideas all day long: running tubes through a block of metal like radiator tubes and then venting super high speed air through them, setting the vanes/blades on your heatsink so that the airflow spirals down into the vanes and back out..the idea being to keep the airflow smooth and non-turbulent while it's going across surfaces.

The 'heatpipe' stuff has TONS of possibilities. I've got an ice-cream scoop at home that works on the same priciple - it transfers heat from your hand through some internal liquid to the scoop. It's amazing how well it works.

edited four rotten speling.

 

[FrankSchwab]

"It's always better to suck than to blow" is what an old engineer once told me in reference to cooling cases (at least, that's what I thought he was talking about...), and in general he's right.

Blowing air into a case leads to all kinds of turbulent eddies, dead zones, and reduced efficiency because fans hate blowing into a pressurized box. Sucking air from a case creates a much smoother airflow internally, with fewer dead zones, and generally a smaller pressure differential than you'd get with blowing, leading to more airflow with the same fan.

Blowing air into a heatsink, at least in the current approach with fans on top blowing down, kinda sucks. You create a high-pressure zone under the fan where the air hits the flat surface of the heatsink, and has to turn 90 degrees to exit. The high pressure zone reduces the efficiency of the fan, and as Mist noted the center of the high-pressure zone happens to be the deadspot of the fan's airflow, and happens to be right above the processor.

Heatsink reviews all over the net try the fan in both suck and blow orientations, and get small differences that suggest a different orientation depending on the heatsink, and closeness of obstructions. Some sinks do better in suck mode, some in blow mode, generally an obstruction relatively close above the heatsink reduces the efficiency of suck mode, etc.

It just seems to me that a side-flow design would be more efficient at extracting heat for a given fan. Imagine mounting a fan on the side of a Pal8045 or similar, and placing a shroud around the top and 2 sides of the heatsink, forcing the air to travel from the fan through the sink and out the other side. Now, there is no high-pressure area, because the air doesn't have to change direction. The "dead spot" under the motor hub is now in the middle of the vertical heat sink vanes, with a goodly flow across the heatsink at the base right over the processor. Ideally, you'd put the fan a distance away from the heatsink to create a duct that evens out the airflow through the fan, eliminating the dead spot.

Or should you suck the air through the side of the heat sink? I get so confused.

/frank

 

[Howard]

Originally posted by: FrankSchwab
"It's always better to suck than to blow" is what an old engineer once told me in reference to cooling cases (at least, that's what I thought he was talking about...), and in general he's right.

Blowing air into a case leads to all kinds of turbulent eddies, dead zones, and reduced efficiency because fans hate blowing into a pressurized box. Sucking air from a case creates a much smoother airflow internally, with fewer dead zones, and generally a smaller pressure differential than you'd get with blowing, leading to more airflow with the same fan.

Blowing air into a heatsink, at least in the current approach with fans on top blowing down, kinda sucks. You create a high-pressure zone under the fan where the air hits the flat surface of the heatsink, and has to turn 90 degrees to exit. The high pressure zone reduces the efficiency of the fan, and as Mist noted the center of the high-pressure zone happens to be the deadspot of the fan's airflow, and happens to be right above the processor.

Heatsink reviews all over the net try the fan in both suck and blow orientations, and get small differences that suggest a different orientation depending on the heatsink, and closeness of obstructions. Some sinks do better in suck mode, some in blow mode, generally an obstruction relatively close above the heatsink reduces the efficiency of suck mode, etc.

It just seems to me that a side-flow design would be more efficient at extracting heat for a given fan. Imagine mounting a fan on the side of a Pal8045 or similar, and placing a shroud around the top and 2 sides of the heatsink, forcing the air to travel from the fan through the sink and out the other side. Now, there is no high-pressure area, because the air doesn't have to change direction. The "dead spot" under the motor hub is now in the middle of the vertical heat sink vanes, with a goodly flow across the heatsink at the base right over the processor. Ideally, you'd put the fan a distance away from the heatsink to create a duct that evens out the airflow through the fan, eliminating the dead spot.

Or should you suck the air through the side of the heat sink? I get so confused.

/frank

Then again, if you put a filter in front of the intake fans and have higher pressure inside the case you'll have less dust. If the pressure is lower than outside the case dust will get in through the little nooks and crannies.

I do like the way Alpha set up their PEP66 - a wing-finned heatsink and horizontal airflow seems like the best method of air cooling.

 

[Mist]

To me, the best idea would be to have the fan mounted on the side of the HS, so it would blow air through the fins and in turn, this air would then head directly towards the extraction vent on the PSU. Maybe even put fans on either side of the HS, first one to blow, the other one sucking.

Put a wide bit of tape over the top of the HS to ensure that any air being shifted is directly from the HS and not from any surrounding areas, so effectively it's a mini wind tunnel.

How does that sound?

Michael.

 

[RadioactiveHamzter]

I've seen heatsinks with the fans mounted at 45 degrees, I suppose it's pretty much the same as mounting it on the side. I'm actually quite suprised that this kind of fan mounting isn't used more often, is it just that 'dead spots' arent as problematic as some people seem to think?

 

[everman]

How about 2 fans mounted at 45degree angles and slightly angled to blow towards the back of the case (and exhaust fan). They would be above the heatsink kinda like this:
/ \ (/ = fan)
__ <-- heatsink

if that gives you an idea.

 

[Drunkguy]

Alpha makes some heatsinks that use a fan to blow across instead of down onto the heatsink.

Check out the PEP series here http://www.micforg.co.jp/en/cat_pfe.html

 

[PrinceXizor]

First off, I've seen some posts that are more applicable to case fans then to CPU fans, particulary when assoicated with pressurized cases and the whole suck vs. blow issue.

As far as a dead-zone, here are a couple points to keep in mind.

1. Any true dead-zone would be a result of completely laminar air-flow, which is virtually impossible with a traditional fan design (ask anyone who works with pumps and impellers about laminar flow). Secondly, this "dead-zone" would disappear at a certain distance away fromt he fan, you can verify this by using a normal room fan and "feeling" for the air where you think the dead-zone should be.
2. This "dead-zone", if it exists, would be zone that is relatively free of convective (active) cooling, but is still subject to radiated heat transfer.

Point two can be illustrated this way. Let's assume you have a solid heat sink for the sake of illustration. This heat sink has a particular surface temperature. Heat is being transferred from the surface of the heat sink to the layer of air closest to it. This heat transfer occurs at a particular rate dependent upon the properties of the heatsink and of air and the temperature differential b/n the two. Now if our layer one of air could transfer ITS heat to layer two of air as fast as the heatsink transfers ITS heat to layer one, there would be no need for a fan. However, this is not the case, and therefore it is faster to circulate this new hot air, elsewhere (while it dumps its heat load) and bring in new air at the original temperature to repeat the process.

Now a deadzone would still have the same reaction in step one, a transfer of heat from the surface of the heat sink to layer one of the air. Now because the air is relatively still, it would have to transfer this heat to subsequent layers of air until it reach a layer of air that is being circulated. So you would still be channeling heat away from the heatsink, just not at as fast a rate as the circulating air.

Now, when you start factoring in that the excess heat "building-up" or more accuartely, not be transferred as fast, in this small little section of the heatsink, this heat would then transfer to other areas that ARE being "serviced" by air flow.

Then you factor in that this "dead-zone" would be really small, highly localized and probably not as "dead" as we think it would be (Think of the eye of hurricane, its calm compared to the surrounding winds, but still has choppy waters and would be comparable to a gusty and windy day) and you can see why this probably isn't a big concern.

P-X