THe Sandia Cooler - Breakthrough in Air Cooling design

[sm625]

You would have to replace more than just the fan, since there has to be some sort of bearing-containing seal between the fan and the bottom plate. That seal would have no tolerance for dust and debris so it most likely would not be user serviceable.

 

[R0H1T]

You would have to replace more than just the fan, since there has to be some sort of bearing-containing seal between the fan and the bottom plate. That seal would have no tolerance for dust and debris so it most likely would not be user serviceable.

Well that's the thing, it can be made user serviceable but it probably won't be.

 

[alyarb]

What's funny is that if you read earlier in this thread, you see that the Sandia guys were basically targeting low/mid-range mass-market systems for office/industry and suchlike.

This product is being aimed at the enthusiast/gamer market.

I would rather see what they could do with this thing in a 1U rack-type situation.

edit: I know this thread is ancient and therefore not necessarily a great place for off-topicness, but reading some fairly-to-extremely intelligent people post here about real-world "thermo" (heat transfer?) involving a bunch of non-equilibrium, steady-state stuff that makes collegiate-level material less-than-useful in practice piqued my interest.

I have studied enough heat transfer to know what is a thermal boundary layer as well as a few things that can reduce the thickness of said layer (turbulence, coolant velocity, stuff like that). If there are any (economically) accessible materials available to help "unteach" someone who has been exposed to collegiate heat transfer materials, it would be great if you could post links to it here or otherwise provide some indication of where one might find such material.

After all, it could go a long way to helping readers here understand how the Sandia hsf works.

The sandia cooler operates in accordance with the same fundamentals it sounds like you wish to unlearn. not sure why you think they are circumventing some thermodynamic law or that your collegiate experience was a waste. the main problem with classroom physics is that they don't spend enough time teaching people how to think and how to set up their model in their mind's eye, so the boundary of their 'system' and reference frame can be off and that leads down a path of wrongness.

A fan or pump beating fluid down on a heat exchanger does virtually nothing to boundary layer thickness no matter how big or strong. They tested this with intel 423-478 box coolers (thats how old their concept is, tualatin-willamette period) as well as taisol style fin boxes, and then the radial fin intel coolers similar to what you get nowadays, and heat towers.

They are saying they can appreciably reduce the boundary layer if they forget about the working fluid and just keep the entire heat exchanger assembly in a state of angular acceleration. they can also eliminate the buildup of debris because they are not forcing dirty air over a stationary body.. there is no stationary body. they haven't shown all their cards. We don't know how many iterations of design we are looking at and we don't know what else Coolermaster is going to try. We just have a demonstration of the principle at work. I'd like to see how it performs in a black box full of dead air though.

If this is the real deal, and it appears to be, you can be sure there will be 1U applications and after enough time other types of heat exchangers could face extinction as well.

https://www.youtube.com/watch?v=JWQZNXEKkaU

 

[Denithor]

If it failed, would it just grind to a halt or go off-balance and potentially spin out of control grinding up your motherboard & components?

 

[KingFatty]

If it failed, would it just grind to a halt or go off-balance and potentially spin out of control grinding up your motherboard & components?

No, at that point you would use it to make salsa. You know, to eat your "chips" with.

 

[NTMBK]

If it failed, would it just grind to a halt or go off-balance and potentially spin out of control grinding up your motherboard & components?

Just make sure you're nowhere near it when it fails, unless you want to end up looking like chiseled spam.

 

[DrMrLordX]

The sandia cooler operates in accordance with the same fundamentals it sounds like you wish to unlearn. not sure why you think they are circumventing some thermodynamic law or that your collegiate experience was a waste. the main problem with classroom physics is that they don't spend enough time teaching people how to think and how to set up their model in their mind's eye, so the boundary of their 'system' and reference frame can be off and that leads down a path of wrongness.

I did not mean to come off sounding as though this cooler violated the known laws of physics; just go back and read some of aigomorla's and ctho's posts from 2013 and earlier to see what I was talking about (or read Idontcare's post from 6/22/2013). And do not think that I am actually university-trained. My experience is strictly self-taught, using whatever materials have been available within my budget (which ain't always a lot, but you can get by).

But I digress.

It will be interesting to see where coolers of this design find purchase.

 

[alyarb]

ah, i figured we were talking about the same thing

 

[RampantAndroid]

What's funny is that if you read earlier in this thread, you see that the Sandia guys were basically targeting low/mid-range mass-market systems for office/industry and suchlike.

This product is being aimed at the enthusiast/gamer market.

I would rather see what they could do with this thing in a 1U rack-type situation.

I'm not sure this would help much? They're pushing lower power devices often in 1U setups, with a lot of AC cooling going on in server rooms to keep things cool. With a better heatsink, you'd still need the same amount of air conditioning more or less, and I don't know that a better cooler would let you use hotter CPUs (or that you'd want to).

Finally, when you have a server room that's stuffed full, you already have a bunch of hard drives that are a moving part with a defined failure rate. I would suspect very few people would be willing to install a new moving heatsink without it being very well proven - they know MTBF of their hard drives, fans and such...but without knowing a ton about a new cooler they might be unwilling to spend money on it. Especially given that if a drive fails in a RAID5 array, you rebuild the array with some downtime and no data loss. If a heatsink fails on a server, you might very well be replacing the entire server?

Just my thoughts from my limited experience with servers/data centers.

 

[DownTheSky]

It will be interesting to see where coolers of this design find purchase.

Everywhere space is an issue.

Especially given that if a drive fails in a RAID5 array, you rebuild the array with some downtime and no data loss. If a heatsink fails on a server, you might very well be replacing the entire server?

I don't think it would be that hard to build a simple cage around it. Lots of normal coolers have fan grills, don't they?

 

[RampantAndroid]

Everywhere space is an issue.



I don't think it would be that hard to build a simple cage around it. Lots of normal coolers have fan grills, don't they?

Fans aren't spinning hunks of metal, though.

 

[DrMrLordX]

I'm not sure this would help much? They're pushing lower power devices often in 1U setups, with a lot of AC cooling going on in server rooms to keep things cool. With a better heatsink, you'd still need the same amount of air conditioning more or less, and I don't know that a better cooler would let you use hotter CPUs (or that you'd want to).

In the proper configuration, I think these Sandia HSFs could held 1U rackmount servers quite a bit. Maybe not all of them, but on some of the toastier ones, they could be quite useful.

Finally, when you have a server room that's stuffed full, you already have a bunch of hard drives that are a moving part with a defined failure rate. I would suspect very few people would be willing to install a new moving heatsink without it being very well proven - they know MTBF of their hard drives, fans and such...but without knowing a ton about a new cooler they might be unwilling to spend money on it. Especially given that if a drive fails in a RAID5 array, you rebuild the array with some downtime and no data loss. If a heatsink fails on a server, you might very well be replacing the entire server?

Just my thoughts from my limited experience with servers/data centers.

I was thinking MTBF would also be a big issue. There will have to be quite a bit of validation done before anyone adopts HSFs based on this design.

 

[KingFatty]

Oh man, slap a 7200 RPM hard drive motor on that sucker and let 'er rip! You get the power consumption and MTBF performance of a hard drive motor, but for cooling.

 

[Plimogz]

This may be slightly off topic, but as it pertains to fluid bearings, and I've always thought it was completely and utterly awesome, I'll go ahead:

From the Wikipedia entry on fluid bearings, pertaining more in particular to one installed at the Holtwood Hydroelectric Plant

According to the ASME [...], the first Michell/Kingsbury fluid bearing in the USA was installed in the Holtwood Hydroelectric Power Plant (on the Susquehanna River, near Lancaster, Pennsylvania, USA) in 1912. The 2.25-tonne bearing supports a water turbine and electric generator with a rotating mass of about 165 tonnes and water turbine pressure adding another 40 tonnes. The bearing has been in nearly continuous service since 1912, with no parts replaced. The ASME reported it was still in service as of 2000. As of 2002, the manufacturer estimated the bearings at Holtwood should have a maintenance-free life of about 1,300 years.

1300 years!

So, you know, it's a good bet that if you design a relatively simple device around a well implemented fluid bearing, you're going to hit MTBFs which can pass muster.

I'm not sure this would help much? They're pushing lower power devices often in 1U setups, with a lot of AC cooling going on in server rooms to keep things cool. With a better heatsink, you'd still need the same amount of air conditioning more or less, and I don't know that a better cooler would let you use hotter CPUs (or that you'd want to).

It seems to me that you can characterize a cooler according to a few simple metrics: cooling efficacy (delta temperature between component & ambient), size, weight, cost, MTBF, etc. And I think that we can pretty much throw out cost, within reason, as pertains to server rooms. So if the Sandia cooler can keep the chips cool enough using warmer ambient air (which is the point, I think -- though it could be cooler chips w/ the same ambient temp, for us air-overclockers) while being smaller to increase density, which is primordial, it will prove a winner.

Imagine, if you will, an ideal (read: magic) cooler which would keep components a mere degree above ambient. With such coolers, given the acceptable operating temperatures of most electronics, you could operate in even the warmest places without resorting to costly air refrigeration; just assure sufficient air flow (cheap), and relatively simple, inexpensive conditioning (-dust/-humidity).

 

[Flapdrol1337]

It's an interesting concept, good performance for the size, probably very reliable. But migt be noisy, expensive, and performance might not increase that much if you scale it up.

 

[Puffnstuff]

I wonder how they plan to keep the fan in place when the unit is mounted vertically. Every demonstration has been in a horizontal position. A conventional hsf can operate in either position so I'm curious to see the implementation they use for it. I also think about just how efficient the heat transfer is from the base to the fan which is supposed to dissipate it. When you think about it the concept flies in the face of what we do now with our tims. They want to use air to conduct heat and we strive to eliminate as much of it as we can with the application of our tims.

Anyway they would have to have some kind of retention system with a retail unit to keep the spinning fan in place in any orientation and that will add cost and reduce performance. Will the fan operate in any orientation? A conventional fan doesn't care if your case is upright or on its side and I run my pc both ways especially if I'm working on it. I just don't see this as a viable option just yet. I think the principle could bring us long life fans in the future without the worry of a failed bearing.

 

[RampantAndroid]

Imagine, if you will, an ideal (read: magic) cooler which would keep components a mere degree above ambient. With such coolers, given the acceptable operating temperatures of most electronics, you could operate in even the warmest places without resorting to costly air refrigeration; just assure sufficient air flow (cheap), and relatively simple, inexpensive conditioning (-dust/-humidity).

Basic physics and thermodynamics says no. In a closed room, even if you dissipate heat faster, energy is preserved. The energy still remains in the room. Sure, you might be a mere degree above ambient. But ambient was 70F when you started, and it's 130F in the room now and still climbing. My point stands: you need AC regardless.

 

[LurchFrinky]

Basic physics and thermodynamics says no. In a closed room, even if you dissipate heat faster, energy is preserved. The energy still remains in the room. Sure, you might be a mere degree above ambient. But ambient was 70F when you started, and it's 130F in the room now and still climbing. My point stands: you need AC regardless.

Actually, he didn't say that you don't need AC, just the refrigeration aspect. You still save a lot of power/money if you merely suck in outside air instead of trying to cool down hot exhaust for re-use.
Or, in other words, don't limit yourself to a closed room.

 

[Plimogz]

Obviously you would exhaust the warm air and draw in (relatively) cooler air from outside. Perhaps I didn't make that clear -- the idea isn't to build a convection oven in there.

Your point about the closed room is a straw man, in that your own counter example used AC to evacuate the heat out of the room, while you would imply that a room w/o AC would be incapable of evacuating the heat put out by the hardware. Whereas in reality, ventilation would also achieve this, while being cheaper.

So if operators could get away with no air refrigeration to achieve acceptable operating temperatures at the chips, while satisfying their need to cram the most processing power into their racks as possible, they would do so. But it simply isn't possible with conventional air coolers which aren't so efficient at removing the heat by running air across the heatsinks -- i guess because of laminar flow, boundary layers and whatnot -- so they resort to running cooler, chilled air through the rooms and hence across the fins.

 

[Shehriazad]

Not even sure why I would need this...my 25$ CPU cooler keeps my CPU at such a low temp that the "thermal reserve" reader for AMD CPUs can't even pick up a proper reading...

All I really care for is effective passive cooling...and seeing how the 3 big CPU/GPU monsters all work on energy efficiency like mad...we will get there soon enough. Even if this invention works well...what's the point if in 5-10 years all normal consumer hardware is cooled passively? (perfect case scenario)

 

[aigomorla]

Basic physics and thermodynamics says no. In a closed room, even if you dissipate heat faster, energy is preserved. The energy still remains in the room. Sure, you might be a mere degree above ambient. But ambient was 70F when you started, and it's 130F in the room now and still climbing. My point stands: you need AC regardless.

only in a closed room.

Actually, he didn't say that you don't need AC, just the refrigeration aspect. You still save a lot of power/money if you merely suck in outside air instead of trying to cool down hot exhaust for re-use.
Or, in other words, don't limit yourself to a closed room.

Yes but in some cases a closed room is the only option, or its required to keep out dust.

Obviously you would exhaust the warm air and draw in (relatively) cooler air from outside. Perhaps I didn't make that clear -- the idea isn't to build a convection oven in there.

Your point about the closed room is a straw man, in that your own counter example used AC to evacuate the heat out of the room, while you would imply that a room w/o AC would be incapable of evacuating the heat put out by the hardware. Whereas in reality, ventilation would also achieve this, while being cheaper.

So if operators could get away with no air refrigeration to achieve acceptable operating temperatures at the chips, while satisfying their need to cram the most processing power into their racks as possible, they would do so. But it simply isn't possible with conventional air coolers which aren't so efficient at removing the heat by running air across the heatsinks -- i guess because of laminar flow, boundary layers and whatnot -- so they resort to running cooler, chilled air through the rooms and hence across the fins.

again, by pulling in fresh air unless it was filtered you would introduce dust.
Dust is not something one wants in a server room as it will deposit and reduce efficiency of the said machines.
This would also reduce the down time required for maintenance in cleaning out the dust.

Having a closed room with filtered climate control can limit the amount of dust introduced to the server room.

Not even sure why I would need this...my 25$ CPU cooler keeps my CPU at such a low temp that the "thermal reserve" reader for AMD CPUs can't even pick up a proper reading...

because supposedly the design of this cool was to be low profile, and also prohibit the deposit of dust.

 

[ehume]

And now ExtremeTech has gotten into the act.

 

[Idontcare]

When you think about it the concept flies in the face of what we do now with our tims. They want to use air to conduct heat and we strive to eliminate as much of it as we can with the application of our tims.

There is a critical difference, the air we eliminate with our TIMs is essentially stagnate air. Air that doesn't move around much.

Whereas the air they are using to transport heat from one metal surface to another is air that is moving very very fast, just like the air we use to remove heat from the fins of our HSF.

Don't conflate the thermodynamics with the kinetics, both are in play but in this case they are intentionally tipping the scale well towards the kinetics and taking advantage of the thermal transport properties of air.

It may seem preposterous, but at the same time all of consumers depend on brainy scientists to do preposterous things (to us laymen) in order for the magical world of electronics to even exist.

 

[EduCat]

Cooler Master onto the same thing? http://www.techspot.com/news/59346-cooler-master-teams-coolchip-technologies-advance-heatsink-rotating.html

 

[Micrornd]

I wonder how they plan to keep the fan in place when the unit is mounted vertically. Every demonstration has been in a horizontal position. A conventional hsf can operate in either position so I'm curious to see the implementation they use for it.
(snip)
Will the fan operate in any orientation? A conventional fan doesn't care if your case is upright or on its side and I run my pc both ways especially if I'm working on it.

^ This is the problem I see ^

Sandia said in their white paper and in interviews for articles, that this design would function in any orientation.
While that may apply to the heat transfer aspect (depending how the shaft is retrained in a non upright condition), it doesn't translate to air bearings.
AFAIK, Sandia nor any of those coming to market have demonstrated this either long or short term.

I have experience in both air and magnetic bearings, from smaller rotating masses than are involved here, to rotating masses thousands of times heavier.
I can safely say from experience, you don't cantilever weight horizontally (support the shaft by a single bearing).
Gravity makes bad things happen in proportion to weight and time when you cantilever a rotating mass in that fashion.
Air bearings can be run horizontally, but not as simply as these cooler prototypes and CES models are made.

I don't see anyone addressing the startup or shutdown of these coolers either.
Physical contact is a given unless an outside air source (as Sandia used) is initiated during start up and shutdown.
Contact between the 2 metal surfaces produces metal particles (especially on shutdown) and that doesn't go well with such tiny air gaps as are used here.
Commercially, air bearings of this type use an external source to elevate the rotating mass before rotation starts and the gap to the bearing touchdown surface is much greater as the idea is to not transfer heat in those applications.

The concept makes perfect sense and I see no problem if the rotating mass is in a vertical position.
I just can't see it working properly if the cooler were mounted horizontally as in a tower case.

I also think Sandia's thermal resistance and heat transfer numbers are skewed optimistically because of their test setup, but I won't debate that, as it will easily be checked in the next few months when production units are actually tested by websites