Sandia-designed CPU cooler lacks fan, rotates heatsink instead

[onethreehill]

CPU coolers aren't rocket science, but the Sandia National Laboratories has produced one anyway. As one might expect, it's rather special. The patent-pending "Air Bearing Heat Exchanger" promises to substantially reduce the amount of energy required to cool microprocessors while simultaneously cutting noise levels. Fans are the only source of energy consumption for common air coolers, and they happen to be the sole generators of noise. Sandia's cooler doesn't have one. Instead, it rotates the heatsink itself. Here's a diagram pulled from the official press release:

http://techreport.com/discussions.x/21262

[Janus67]

Sounds like it could be a bit dangerous with a thing spinning around in there like that, presumably keeping the base still though I hope otherwise it could start wearing the IHS down too.

[Jamesrt2004]

Well it says stationary base plate!


He'll I'd get one of these, looks cool. Expensive tho.


Mmm thinking. The reasons fans make noise is 'generally/mostly' the actual motor rotating right? Surely this will be a more 'heavy duty' motor seeming it's moving metal ? Or does it not have to spin as fast or something

[Computurd]

dude i think its ingenious. ive read over the whitepaper already and its pretty impressive. i want one. well four actually

[VulgarHandle]

it's the compression/drag of air that creates most of the sound isn't it?

also, will there be a grease port? lol

[eXa]

Meh, the limit on todays cooler aint the fans, its getting the heat up and away from the cpu/ihs. This heatsink is not going to improve that part. Also, radial fans (wich this is) seemes to move less air per db than normal fans(tho with higher head pressure), that and the bigger engine needed to move this amount of weight is going to make this quite noisy.

Tho, it is very cool :p

[zalbard]

How can rotating a heavy heatsink consume less power than rotating light fan blades?

[bamtan2]

yeah if it spins faster than a cpu fan and is heavier too, it'll have to use more power...

but the design is neat and the release seems serious. this guy is probably a lockheed martin employee and also an extreme computer nerd, which is pretty fun.

so this could become reality...eventually. the technology readiness level (TRL) is 4, which is far from 9, so it's not reality yet!

[DeltZ]

How can rotating a heavy heatsink consume less power than rotating light fan blades?

once it's finished accelerating the "heavy" heatsink shouldn't matter unless it adds a lot of bearing friction. It doesn't consume less power per say...but it should be more efficient cooling per db/rpm/power due to the reasons given in the long research paper thingy. Also that prototype is going to have significantly less air resistance than a deep finned fan whose purpose is to push air at high CFM, which is much less needed for this design (apparently)

[STEvil]

Too unreliable.

DeltZ - there must be contact force between the stationary plate and rotating assembly. This will cause much more friction and resistance to overcome than spinning a fan motor has.

[Panther_Seraphi]

Dont see how it could work as you would need contact between the fins and the part of the HS in contact with the processor.

Smoke and mirrors IMO.

[turbox997]

If these prove to be durable enough I could totally see them in laptops.

[Particle]

Too unreliable.

DeltZ - there must be contact force between the stationary plate and rotating assembly. This will cause much more friction and resistance to overcome than spinning a fan motor has.

We as a species have been known to build many reliable, heavy, rotating objects. You don't really have any basis for your claim without knowing more about the design.

[tool_462]

We as a species have been known to build many reliable, heavy, rotating objects. You don't really have any basis for your claim without knowing more about the design.

Psh. Without even reading the whitepaper you can know facts about something that top Department of Energy research scientists obviously don't know :p /sarcasm

[STEvil]

We as a species have been known to build many reliable, heavy, rotating objects. You don't really have any basis for your claim without knowing more about the design.

They tend to have low friction surfaces with poor thermal transfer properties.

The only way to create a low friction heat transfer surface would be to use a vapor chamber like design. This would require the use of a seal (huge mechanical liability vs a fan) to keep the chamber below or above ambient air pressure depending on the required characteristics of the liquid/vapor used internally. Wicking structures would have to be internally separated which reduces contact surface area slightly.

Did I mention a vapor chamber is going to be required for any kind of actual thermal efficiency beyond a handful of watts?

EDIT

I should mention that use of copper ball bearings is a possibility, but they will also break down and have limited energy transfer capabilities as well.

[Hornet331]

Interesting concept, but introducing a insulating layer, even if it is only 25 µm is what i see as the weakest part of that design. Conduction is always better in solid metal then in air. While I can see the noisereduction is very true, I doubt there is much energy saving going on. A fanblade as what 20g? For the heatsink you could easily say it has 200g.

Edit:
I have read a bit now in the paper and when it comes to performance that cooler isn't that "hot". The prototyp has a °C/W of 0.2, while this is better then lets say intels stock heatsinks (which are around 0.3) a cheapo tower cooler like the CM hyper TX3 already offers 0.14-0.16. He suggestes that a modified version (higher fins, smaller baseplat) could score around 0.1 which is basically current highend cpu cooler region (like the Megahalems or NH C-14). I dont know when he gathred his data, but the selection of cpu coolers he used to compare this results to, is lets say, quite outdated..

What I am really disapointed in, is power consumption... 8-10W for his configuration? Tz.. a typical 120mm fan uses between 1 and 3W... And it seems that the second prototype will consome more, because of the bigger rotator.

I don't know if this concept will be that good as cpu cooler but for the other suggested application area in the HVAC sector this might be a winner, espeical for smaller systems.

[PatRaceTin]

not sound good

[Manicdan]

i think its interesting the guy didnt try a design with much thinner fins on the impeller. it looks like his goal was to use a solid block for best thermal transfer, but i question if that would have been as good as using fins soldered down and gaining 3-5x the surface area.

also for him to achieve <.1 C/W, he expect this thing to spin at 10000 RPMs, and the motor consuming 30W

[zalbard]

also for him to achieve <.1 C/W, he expect this thing to spin at 10000 RPMs, and the motor consuming 30W

There go the noise and power efficiency claims, lol.

[[XC] Oj101]

Well think about it, the air flowing over a fan isn't nearly the speed of the air flowing past the blades. The fan might only need to do 100RPM.

[Hornet331]

i think its interesting the guy didnt try a design with much thinner fins on the impeller. it looks like his goal was to use a solid block for best thermal transfer, but i question if that would have been as good as using fins soldered down and gaining 3-5x the surface area.

Maybe balance issues, imho the rotator needs to balanced quite well and with soldering you can easily introduce weight distribution issues resulting in a unbalanced heatsink. Thought maybe sintering can work, but i guess that will have negative impact on the heat conduction capabilities.

[SubZero.it]

How can rotating a heavy heatsink consume less power than rotating light fan blades?

+1000

[Solus Corvus]

How can rotating a heavy heatsink consume less power than rotating light fan blades?

Because of newton's first law.

It would only consume extra power while accelerating the heatsink to speed, and that is a very short period of time. Once in motion the motor only has to counteract the fluid drag on the blades and base.

[Serra]

Because of newton's first law.

It would only consume extra power while accelerating the heatsink to speed, and that is a very short period of time. Once in motion the motor only has to counteract the fluid drag on the blades and base.

No, that's a logical fallacy. "It will only consume extra power" does not go hand-in-hand with "consume less power". At best you are going for a long-term near-equality. Also, as pointed out previously, the HS will at the very least have more friction and must therefore require more energy. There is simply no avoiding it. Finally, rotating at a few thousand RPM means that their goal is to use at least 2k rpm... that's a lot more RPM than most fans have, a ridiculous amount more if they're looking at 3k+.


As well, I see that picture includes an AIR GAP between the thin conductive base and the actual heatsinks. For very good reasons good coolers do NOT have air gaps of any size, but because microscopic air gaps do exist we use thermal paste. You want to see the absolute best-case scenario for this design? Go scrape the thermal paste off your heatsink and compare its performance. Except this will perform worse because the gap has to be large enough to allow a sufficiently large gap for rotation and wobble.

Oh, plus now if you're working on a server that's heating up because a motor went out you can't just replace a fan while leaving the thing on, you need to take the whole server down so you can manually and annoyingly replace the heatsink. And these motors will be more likely to fail because they're going much faster with a heavier load...


No, no... this whole thing is a horrible idea. A 100% bad idea.

[DeltZ]

No, that's a logical fallacy. "It will only consume extra power" does not go hand-in-hand with "consume less power". At best you are going for a long-term near-equality. Also, as pointed out previously, the HS will at the very least have more friction and must therefore require more energy.

In context of theory under perfect situations he's right. You're only going to require more energy to start it up, then the power would much less after it has finished accelerating.
Second point, coefficient of friction of a bearing may be signficantly small such that the extra mass/weight of the heatsink (Which isn't as much as a conventional heatsink) will not contribute as much losses as the reduction/comparison in/to aerodynamic drag that a standard fan experiences in comparison to the spinning heatsink.

but yeah...definitely be skeptical the performance may not scale that well.
RPM needed to achieve said performance seems pretty sketchy.

Likely wont work well in practise and no one will license the idea :P