Spinning heatsink

The forum for non-component-related silent pc discussions.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

bobo5195
Posts: 54
Joined: Thu Apr 15, 2004 2:45 pm

Spinning heatsink

Post by bobo5195 » Fri Jul 08, 2011 5:26 am

Any thoughts on this
http://prod.sandia.gov/techlib/access-c ... 100258.pdf
From here http://nextbigfuture.com/2011/07/sandia ... ffers.html

Uses a spinning heatsink by transferring heat across an air bearing (40 nanometer disk) they seem to show that there is minimal thermal resistance from the bearing ~0.05 C/W and of course a spinning bit of aluminium helps with the heat transfer.

I think the numbers look a little high if anything compared to the current best of class on the PC side. It is likely they can’t beat heatpipes for getting a large surface area.

Half tempted to make my own as it looks easy enough. Machining on that air bearing maybe a little hard though. Sure you can make a quiter centrifugal arrangement with some different techniquies.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Re: Spinning heatsink

Post by NeilBlanchard » Fri Jul 08, 2011 8:26 am

I wonder how well the heat is transferred to the rotating impeller? It is hard enough if the contact between the base of a heatsink and the CPU is a little off, how can there be good transfer by contact with a spinning part?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Spinning heatsink

Post by CA_Steve » Sat Jul 09, 2011 8:34 am

The white paper is an interesting read...and 18 months old. I wonder if the 2nd gen prototype is under evaluation, yet?

1st gen shortfalls:
- 10W+ to run it at optimum. Kinda painful.
- 0.4" impeller height was deemed to short. 2nd gen will be 1" to reach 0.1C/W. That will drive motor power up more.

General questions and comments"
- will the impeller exhaust provide some airflow/cooling to the surrounding VRM circuitry?
- cost: I wonder how this would compare to a decent traditional tower ($30-50)?
- the fan motor is mounted to the baseplate.....so, it'll be in close proximity to 60C+ temps. As opposed to traditional coolers where the fan motor is in proximity to the cooling fins temps (~30-45C maybe).

deimos3428
Posts: 10
Joined: Fri Dec 17, 2010 8:07 am

Re: Spinning heatsink

Post by deimos3428 » Tue Jul 12, 2011 11:15 am

I want to see somebody combine this idea with the Dyson bladeless fan. Bladeless spinning heatsink for the win.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Re: Spinning heatsink

Post by NeilBlanchard » Wed Jul 13, 2011 3:19 am

The blades on the Dyson are inside the base.

deimos3428
Posts: 10
Joined: Fri Dec 17, 2010 8:07 am

Re: Spinning heatsink

Post by deimos3428 » Wed Jul 13, 2011 6:10 am

NeilBlanchard wrote:The blades on the Dyson are inside the base.
Yes, of course. But we can do away with them too, I think, by spinning the ring itself in the opposite direction at a high enough RPM.

What I'm imagining is something like a standard heatplate/heatpipe assembly going up to a static metal heat dissipation ring. (You could put some fins on one or both sides of it if you wanted, like a Zalman cooler.) Inside (or outside) of that ring there's a spinning ring-shaped airfoil, similar to the Dyson "bladeless" design. Due to the airfoil's shape and motion it should induce a Coanda effect, significantly magnifying the airflow through the heatsink.

But I'm no engineer. It probably won't work.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Wed Jul 13, 2011 7:07 am

Maybe I just don't get it... but this seems goofy.

What it appears to do is to is improve the conduction of heat through air... if I understand it right.

How can it do that better than heatpipes transferring heat to a large fin assembly... with air blowing through the fins? It solves for a heat conduction problem that doesn't exist in PCs.

Is this BS or am I missing something here?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Spinning heatsink

Post by CA_Steve » Wed Jul 13, 2011 10:49 am

The pdf white paper does a pretty good job of going through the appropriate math and following up with results.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Wed Jul 13, 2011 9:09 pm

CA_Steve wrote:The pdf white paper does a pretty good job of going through the appropriate math and following up with results.
It no doubt solves a problem. My question is whether or not the problem it solves is, or is not, applicable to PC cooling issues. Can you speak to that question?

Myth!
*Lifetime Patron*
Posts: 151
Joined: Sat Oct 01, 2005 3:30 am
Location: Beds, UK
Contact:

Re: Spinning heatsink

Post by Myth! » Thu Jul 14, 2011 7:17 am

"In conventional FFHS devices, the difference in temperature between the base of the finned heat sink and ambient air is almost entirely accounted for by the temperature drop across the boundary layer. The exception to this rule is lap top computers, where available electrical power is extremely limited. In this special case, CPU clock speeds and fan rotor speeds are reduced to conserve power, albeit at the expense of CPU performance. At these low fan speeds the residence time of air in the heat exchanger is greatly extended, resulting in much higher exhaust air temperatures."

Already they are getting mixed up with temperature and heat. Who cares the temperature of the air? If it is moving very slow it removes less heat. If the air (or water) is moving fast but only rises by 1c or less it is still removing alot of heat.

"The high level of audible noise generated by the larger, more powerful fans used in high-capacity CPU coolers has also proved a deterrent to further scaling up of such devices."

Just isn't true when pertaining to CPU coolers

"commercially available CPU cooler typical of those used in many desktop computers. For example, the calculation below is for a Bitspower model NP15S CPU cooler with thermal resistance rating of 0.92 K W-1, fan diameter of 60 mm, and fan speed: 4800 rpm."

Again, just isn't true. This HSF was on sale in 2001. It's only 26mm high (fan plus heatsink). Most desktop fans are 25mm high on their own. 4800rpm? lol. 60mm? not these days. 0.92 K/w? very poor compared to average nowadays.
Nexus low-7000 acheives: 0.29 c/w at 7v 1200rpm 18dba
0.23 c/w at 12v 1900rpm 31dba

"Accordingly, the temperature of the air discharged by such a CPU cooler is only slightly greater than the temperature of the surrounding ambient air, even if the CPU is running very hot."

Again mixing up heat and temperature.

"figure 5: the pressure-flow curve for a 4800rpm 60mm axial fan typical of those used in cpu cooling applications"

These havent been typical for over a decade.

"Our version 1 prototype device on the other hand has a measured thermal resistance of 0.2 C/W
For example, state-of-the-art CPU coolers providing thermal resistances as low as 0.2 C/W are not only very large devices unsuitable for mass market applications, but in addition are prohibitively expensive."

yet the low-7000 acheives 0.23 c/w at a cost of around £40 in shops, so much less at trade prices.

"As a final example, we note that the much lower noise levels generated by this new air cooling architecture has significant implications for scaling up air flow (and therefore cooling), as discussed in Section 1. While we have yet to record comparative laboratory data pertaining to sound levels, consistent with the discussion of fan noise in Section 1, it is clear that low noise operation will ultimately be another important advantage in real-world applications."

Utter guesswork.
All HSFs mentioned in the white paper were manufactured 2002 or earlier.

At "several thousand rpm" their cooler uses 3-4 watts, I think most of ours work at 1.5 ish.

"No quantitative acoustic measurements were conducted due to limited resources, either. Nonetheless, it was obvious during initial testing that the broadband acoustic spectrum generated by conventional small high-speed fans was largely absent in the version 1 prototype device."

ie they can hear it in their lab, plus small high-speed fans are no longer conventional. The version 1 prototype was a 100mm diameter item, so should be compared with 92mm and 120mm fans. A quick look at a nexus fan of that size would show 18-21 db (from memory, probably innaccurate). I doubt their lab has an ambient noise level of that or less.

"With regard to the effect of PWM frequency, the Motortron ESC has only two settings, 8 KHz and 16 kHz. Nonetheless, recording of data at both PWM frequencies was important in that it showed that operation 16 KHz provided an improvement in motor efficiency, significant reduction in audible electro-acoustic noise associated with magnetostriction of the brushless motor stator poles."

ie, they can hear the change in noise in their lab.

"The FY09 Tier 1 LDRD and Sandia Royalty funds invested in this proof-of-concept study have resulted in a major breakthrough in air-cooling technology. The arguments put forth in the original proposal in support of the boundary layer thinning effect, negligible air gap thermal resistance, low electrical power consumption, low-noise operation, and immunity to fouling have all been shown to be correct."

havent shown lower noise, consumes more power, might be cheaper to make but will it be cheaper to buy?, immunity to fouling has advantages with certain markets, but not the overclocking crowd who are the target for >3ghz clock speeds. The rest of the world has happily moved on to parallel cpus at lower clock speeds. They haven't shown much lower thermal resistance than what is already available to the consumer at reasonable prices.

Just my 2 cents ;-)

Reachable
Friend of SPCR
Posts: 396
Joined: Wed Nov 19, 2003 9:55 am
Location: Western Mass.

Re: Spinning heatsink

Post by Reachable » Fri Jul 15, 2011 1:26 pm

Here's a very brief article that intros the topic, for those -- like me -- who don't want to read the whole long technical .pdf.

http://www.techspot.com/news/44636-sand ... -fins.html

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Sat Jul 16, 2011 7:54 am

Myth! wrote:Just my 2 cents ;-)
So I take it that you agree that whatever problem they have solved, it isn't particularly relevant to the cooling of PC cpus.

Reachable
Friend of SPCR
Posts: 396
Joined: Wed Nov 19, 2003 9:55 am
Location: Western Mass.

Re: Spinning heatsink

Post by Reachable » Sat Jul 16, 2011 9:59 am

If this can be applied to compressors in air conditioners, etc., then it can be a great boon.

Assuming that it's used to make quieter units, rather than more powerful units with the same level of noise.

Myth!
*Lifetime Patron*
Posts: 151
Joined: Sat Oct 01, 2005 3:30 am
Location: Beds, UK
Contact:

Re: Spinning heatsink

Post by Myth! » Sat Jul 16, 2011 12:29 pm

ces wrote:
Myth! wrote:Just my 2 cents ;-)
So I take it that you agree that whatever problem they have solved, it isn't particularly relevant to the cooling of PC cpus.
yes, I don't think the problems they say they solve actually exist. It just seems odd that all the cpu coolers they compare it to were designed a decade earlier. I still think it could be an interesting idea but needs proper noise testing, as it spins at 3k rpm and is a very heavy thing (relatively) that is spinning. So I also wonder about vibration, they suggest it works better without a top and they are gonna make it taller (ie heavier and higher centre of gravity), but i guess if it touches nothing then no vibes are transmitted.

Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Re: Spinning heatsink

Post by Tzupy » Tue Jul 19, 2011 4:50 am

Inventor answers questions about the spinning heatsink here:
http://www.extremetech.com/computing/90 ... e-inventor

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Tue Jul 19, 2011 9:03 am

Tzupy wrote:Inventor answers questions about the spinning heatsink here:
http://www.extremetech.com/computing/90 ... e-inventor
The inventor, Jeff Koplow ([email protected]) still seems to unintentionally dance around the main issues relating to its practical application to PCs.

The weakest area of current technology is likely the thermal grease. Is this thin film of shearing air a better conductor of heat than thermal grease. If it isn't... then it isn't useful in our application. If he doesn't speak to that issue it is hard to assess the benefits.

Even if it is, then the next question is how does it compare with a water block... as the cost of water cooling systems has dropped to equal the cost of current high performance air cooling systems.

Another part of the problem that the author doesn't seem to be aware that most CPUs operate most of the time at idle. He assumes that they spend most of the time at full load.

His assumptions are "I suspect eventually we’ll end up at about 0.05 C/W for a 10 cm diameter device that’s of order 3 cm high, operates at three to four thousand rpm, is inaudible, and consumes about 5 watts of electrical power."

Modern Sandy Bridge CPUs use up about 4 watts at idle and not all that much more under typical use (perhaps 10 watts). The vast majority of the time they are running at idle. And most cpu heat sink fans running at 1000 rpm or less don't use more than about 1 watt:
http://www.xbitlabs.com/misc/picture/?s ... ig.png&1=1

EDIT: 3 cm by 10 cm is certainly a nice size factor. That certainly solves a compelling problem for many of us.

ame
Posts: 488
Joined: Tue Jul 10, 2007 10:35 pm
Location: Israel

Re: Spinning heatsink

Post by ame » Tue Jul 19, 2011 11:01 am

This seems like its still in early stages of development and proof of concept rather than ready for production design. I like the out of the box thinking here.

To me it makes more sence in air conditioning or other possible industrial/motor/car enviorment rather than in a PC. If it can realy make an AC unit improve its efficiently and be quiet doing so it may prove to be very useful and welcome development.

Posted to keep track of this thread.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Tue Jul 19, 2011 5:33 pm

I exchanged some email with the inventor. I told him I would post his answers for him.

The weakest area of current technology is likely the thermal grease. Is this thin film of shearing air a better conductor of heat than thermal grease. If it isn't... then it isn't useful in our application. If he doesn't speak to that issue it is hard to assess the benefits.
Jeff Koplow wrote: I respectfully disagree with the statement that the thermal interface material (TIM) is the bottleneck to heat transfer, although to some extent that may reflect my perspective with respect to real world applications. I work for the Department of Energy, and our highest priority is to try to develop technology solutions that can be widely applied in mass market applications. In some cases energy savings is not the primary goal. Much of the technology we develop is geared towards questions such as how can we make our IT infrastructure provide the tax payer with better service (e.g by allowing clock speeds in mass marketed CPUs, GPU, etc. to be increased). So when I think of end-use applications, I think in terms of the "vanilla ice cream" ~3 GHz CPUs that are used in enormous volumes throughout the residential, commercial, and industrial sectors.

With regard to such vanilla ice cream (VIC) CPUs, the thermal interface material (when properly implemented), while very important, is not the bottleneck to heat transfer. Otherwise it would not be possible to overclock a VIC CPU with a big, noisy, expensive, high-end air cooler (or by using water, for that matter), because you'd be barking up the wrong tree. Having said that, I certainly share your opinion that thremal interface material technology leaves a lot to be desired. Forexample, one big unresolved problem is that a properly installed heat sink often exhibits a reduction in cooling performance over time because the paste/grease-filled thermal joint eventually loosens up, delaminates, etc. over time (e.g., because of repeated expansion and contraction associated with thermal cycling). This is a liability for sure.

Were we to make great strides in the performance of air cooled heat exchangers, eventually the thermal interface material (even when properly implemented) would start to become the dominant limitation. A great deal of work has been done in the area of better thermal interface materials by DARPA in last couple of years. It's a really important problem. At the end of this email I cut and paste a passage from a patent application I filed last year that discusses how to solve the problem of minimizing thermal resistance at the thermal interface to the CPU, GPU, eliminate the problem of thermal joint degradation over time, and provide very efficient heat spreading so as to facilitate the use of an air bearing heat exchanger. The proposed solution will definitely work. It will be a matter of minimizing manufacturing costs.
Even if it is, then the next question is how does it compare with a water block... as the cost of water cooling systems has dropped to equal the cost of current high performance air cooling systems.
Jeff Koplow wrote: I would be the first to agree that if the overall value proposition of a water-based cooler system is better than that of an air-cooled system, you shouldn't use an air cooler. Opinions vary widely on the subject. Again, because I'm mostly thinking in terms of mass marketed CPUs, GPUs, etc. I'm working under the assumption that the logistical advantages of air cooling are very hard to pass up. One other point is that eventually, the heat collected by a circulating fluid loop has to go somewhere. Unless you're not recirculating the water (e.g. discharging it down the drain), the function that a fluid-loop heat exchanger is much like that of a heat pipe.

1) It allows very high rates of heat transfer per unit area. With highly concentrated heat loads, this is a critical consideration.
2) It allows the heat to be relayed to a remote location where it can be dealt with by whatever means necessary.
3) It can be used to reformat the heat load to a much larger area, thereby placing far less of a demand on whatever thermal interface material is used in the remotely located heat exchanger.

So while fluid-loops and heat pipes are indispensible for transmitting heat through choke points (from the standpoint of available cross-sectional area at the thermal interface), relaying heat to a different location where more real estate is available, and/or reformatting highly concentrated and/or unevenly distributed heat loads, it important to consider the fact that neither technology performs the next step of heat rejection to the surrouding air.
Another part of the problem that the author doesn't seem to be aware that most CPUs operate most of the time at idle. He assumes that they spend most of the time at full load.
Jeff Koplow wrote: Agreed, but that's not where you run into thermal management problems. Any thermal management system that can handle a thermal load of 150 W will also function well when confronted with a 5 W thermal load.
His assumptions are "I suspect eventually we’ll end up at about 0.05 C/W for a 10 cm diameter device that’s of order 3 cm high, operates at three to four thousand rpm, is inaudible, and consumes about 5 watts of electrical power."
Jeff Koplow wrote: I'm not assuming this. Rather, I'm trying to give people an idea of where the technology is most likely headed if we are successful.
Modern Sandy Bridge CPUs use up about 4 watts at idle and not all that much more under typical use (perhaps 10 watts). The vast majority of the time they are running at idle. And most cpu heat sink fans running at 1000 rpm or less don't use more than about 1 watt"
Jeff Koplow wrote: I would be the first to agree that where a high-perfomance CPU cooler is not needed, it should not be purchased and installed. It all depends on what is likely to be the worst-case thermal load scenario. That's where inadquate heat sinking can be fatal, even over a relatively short period of time. If the CPU is only generating 5 W of waste heat, then let's have the OS turn off the motor to the heat-sink-impeller, or slow it down.

With regard to the question of power consumption, if 5 W represents an increase in power consumption, that doesn't automatically mean that it's a deal breaker. Part of the question is whether the difference between using 1 W for cooling and 5 W for cooling is significant enough to worry about. In the VIC PC I'm using to type this e-mail, 5 W only represents a tiny fraction of overall power consumption. I guess you can always argue that less power consumption is better, but my opinion is that in engineering you always need to consider the question of dimishing returns.

jkoplow
Posts: 3
Joined: Tue Jul 19, 2011 5:42 pm

Re: Spinning heatsink

Post by jkoplow » Tue Jul 19, 2011 6:36 pm

JKOPLOW: Let me also take this opportunity to answer questions raised by Myth, because I think he/she brings up several important issues:

MYTH: "In conventional FFHS devices, the difference in temperature between the base of the finned heat sink and ambient air is almost entirely accounted for by the temperature drop across the boundary layer. The exception to this rule is lap top computers, where available electrical power is extremely limited. In this special case, CPU clock speeds and fan rotor speeds are reduced to conserve power, albeit at the expense of CPU performance. At these low fan speeds the residence time of air in the heat exchanger is greatly extended, resulting in much higher exhaust air temperatures."

Already they are getting mixed up with temperature and heat. Who cares the temperature of the air? If it is moving very slow it removes less heat. If the air (or water) is moving fast but only rises by 1 c or less it is still removing alot of heat.

JKOPLOW: The significance of air temperature is simply that the rate of heat transfer per unit time, per unit area heat sink, is proportional to the difference in tempertaure between the surrounding air and the heat sink fins. I would add that ensuring adequate air flow is important, but there always is a point of diminishing returns. The point of dimishing returns is in fact defined by the issue of air temperature in the vicinity of the fins. If the boundary layer is the bottleneck to heat transfer, then my goal should be to move as much air as necessary to keep the temperature of the air immediately outside the thermal boundary layer pegged to near room temperature (that's the best I can do), provided I can afford the electrical power, real estate, and hardware required to establish such an air flow rate. In a laptop computer, where space is at an absolute premium and battery life is paramount, your options are very limited.

MYTH: "The high level of audible noise generated by the larger, more powerful fans used in high-capacity CPU coolers has also proved a deterrent to further scaling up of such devices."

Just isn't true when pertaining to CPU coolers

JKOPLOW: I don't agree with that assessment. CPU cooler noise is the reason that sites such as Tom's Hardware list dBA ratings as well as thermal performance data. The OEM cooler used in my office PC that a I bought a few months ago is loud enough that I no longer use speaker phone. I guess it's possible that you are less sensitive to fan noise than the average person. I would assert that CPU fan noise is something many people care about. That for example is why CPU coolers now often quote one C/W figure for maximum fan speed, and a separate figure for operation in "quiet mode". The same is true of the fans used in air conditioners, which many people choose to operate on "quiet mode". This forum is entitled Silent PC Review because many people do find the noise of high-performance CPU coolers to be objectionable.

MYTH: "commercially available CPU cooler typical of those used in many desktop computers. For example, the calculation below is for a Bitspower model NP15S CPU cooler with thermal resistance rating of 0.92 K W-1, fan diameter of 60 mm, and fan speed: 4800 rpm."

Again, just isn't true. This HSF was on sale in 2001. It's only 26mm high (fan plus heatsink). Most desktop fans are 25mm high on their own. 4800rpm? lol. 60mm? not these days. 0.92 K/w? very poor compared to average nowadays.
Nexus low-7000 acheives: 0.29 c/w at 7v 1200rpm 18dba
0.23 c/w at 12v 1900rpm 31dba

JKOPLOW: Referring back to the white paper, I was attempting to point out a little known but critically important fact, that small high-speed fans are surprisingly terrible from the standpoint of aerodynamic efficiency (a few %). I would be the first to agree that a 100-mm fan operating at 1200 rpm is less surprisingly terrible, but it's still terrible. The question of whether your expending 97% of the mechanical work supplied electric motor to do unproductive work on the surrounding air, or only 90% of the mechanical work supplied electric motor to do unproductive work on the surrounding air doesn't change the conclusion of the white paper, which is that such numbers suggest that perhaps much could be gained by reconsidering our strategy for encouraging interaction between the finned heat sink and the surrounding air to. This is one of those cases in which you don't need state-of-the-art hardware to make a valid and important observation. I've yet to find a fan manufacturer that quotes values for aerodynamic efficiency (probably because the numbers are so horrible). Thus I think it's a useful and important observation when asking the question "How might we be able to do better"? Basically, that's my job.

MYTH: "Accordingly, the temperature of the air discharged by such a CPU cooler is only slightly greater than the temperature of the surrounding ambient air, even if the CPU is running very hot."

Again mixing up heat and temperature.

JKOPLOW: Please see above.

"figure 5: the pressure-flow curve for a 4800rpm 60mm axial fan typical of those used in cpu cooling applications"

These havent been typical for over a decade.

JKOPLOW: To the extent that this is true of contemporary OEM CPU coolers standard on vanilla ice cream PCs, such as those used throughout the residential, commercial, and private sector, I apologize for suggesting otherwise. The conclusion I came to when I looked into this issue back in 2009 was that the vast majority of mass-marketed PCs usually ship with inexpensive CPU coolers that entail the use of a relatively small heat sink (lower material costs), and a fan of comparable diameter, wherein high rpm operation is used to partly make up for the fact that a smaller, cheaper fan was used (e.g., compared to a high-end CPU cooler). Having said that, what really matters is the observation that there is enormous potential for improvement with regard to improving the efficiency of the air-heat-sink interaction, which in turn was an important clue with regard to indentifiying a new avenue for reduced fan noise.

MYTH: "Our version 1 prototype device on the other hand has a measured thermal resistance of 0.2 C/W
For example, state-of-the-art CPU coolers providing thermal resistances as low as 0.2 C/W are not only very large devices unsuitable for mass market applications, but in addition are prohibitively expensive."

JKOPLOW: I apologize for not clarifying that I was referring to mass-marketed PCs.

MYTH: yet the low-7000 acheives 0.23 c/w at a cost of around £40 in shops, so much less at trade prices.

"As a final example, we note that the much lower noise levels generated by this new air cooling architecture has significant implications for scaling up air flow (and therefore cooling), as discussed in Section 1. While we have yet to record comparative laboratory data pertaining to sound levels, consistent with the discussion of fan noise in Section 1, it is clear that low noise operation will ultimately be another important advantage in real-world applications."

Utter guesswork.

JKOPLOW: I guess to some extent you could say it was guess work. We ran our device side-by-side in the lab with a state-of-the-art ultra-quiet CPU cooler (out in the open, not inside a PC chassis), and both were inaudible at a distance of 1 meter. It's guess work from the standpoint of quantitative dBA measurements, but if the a CPU cooler is inaudible, I would assert that it's not of great importance what the actual dBA figure is. Having said that we do plan to make measurements, but it's not a high priority because the unit is so quiet.

MYTH: All HSFs mentioned in the white paper were manufactured 2002 or earlier.

At "several thousand rpm" their cooler uses 3-4 watts, I think most of ours work at 1.5 ish.

JKOPLOW: As explained above, neither of these points turn out to be important in the context of the point I was attempting to make regarding the poor aerodynamic efficiency of CPU cooler fans.

MYTH: "No quantitative acoustic measurements were conducted due to limited resources, either. Nonetheless, it was obvious during initial testing that the broadband acoustic spectrum generated by conventional small high-speed fans was largely absent in the version 1 prototype device."

ie they can hear it in their lab, plus small high-speed fans are no longer conventional. The version 1 prototype was a 100mm diameter item, so should be compared with 92mm and 120mm fans. A quick look at a nexus fan of that size would show 18-21 db (from memory, probably innaccurate). I doubt their lab has an ambient noise level of that or less.

JKOPLOW: Please see above. By the way, some of your comments thus far suggest that you are not concerned with fan noise, and some of your comments suggest you are concerned about fan noise. Since most people are, I'll assume this is the case in any discussion that follows.

MYTH: "With regard to the effect of PWM frequency, the Motortron ESC has only two settings, 8 KHz and 16 kHz. Nonetheless, recording of data at both PWM frequencies was important in that it showed that operation 16 KHz provided an improvement in motor efficiency, significant reduction in audible electro-acoustic noise associated with magnetostriction of the brushless motor stator poles."

ie, they can hear the change in noise in their lab.

JKOPLOW: There are two points here. The first point is that to hear the motor you have to get very close; the device as a whole is inaudible at 1 meter unless you crank the motor up past ~5000 rpm. The second point is that the problem of fan motor noise is insignificant in commercially available CPU coolers unelss they are very poorly implemented. Accordingly, we are not particularly concerned that the brushless motor used in the version 1 prototype generated a small amount of audible noise. Ostensibly, when it comes time to roll out a commercial product based on the air bearing heat exchanger, the manufacturer could avail themselves of the same quiet brushless motor technology.

MYTH: "The FY09 Tier 1 LDRD and Sandia Royalty funds invested in this proof-of-concept study have resulted in a major breakthrough in air-cooling technology. The arguments put forth in the original proposal in support of the boundary layer thinning effect, negligible air gap thermal resistance, low electrical power consumption, low-noise operation, and immunity to fouling have all been shown to be correct."

havent shown lower noise, consumes more power, might be cheaper to make but will it be cheaper to buy?, immunity to fouling has advantages with certain markets, but not the overclocking crowd who are the target for >3ghz clock speeds. The rest of the world has happily moved on to parallel cpus at lower clock speeds. They haven't shown much lower thermal resistance than what is already available to the consumer at reasonable prices.

JKOPLOW: I would assert that you're are not fully appreciating the scientific and engineering implications of the data taken with this highly unoptimized version 1 device. In my opinion this research has clearly identified a new pathway to much better overall CPU cooler performance, and now it's a matter of ironing out the details, doing optimization, and ensuring low per-unit manufacturing cost. No matter how successful air bearing heat exchanger technology proves to be, there will always be people who opt not use it, and I certainly don't have a problem with that. We would like to make it available where ever it will useful, however.

-Jeff

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Tue Jul 19, 2011 10:55 pm

The issue that we have to solve with PCs is removing heat from the CPU and transferring to air outside of the PC case.

We can't do anything about moving heat of the semiconductor to the baseplate of the CPU. But once it gets there we have to remove it to the air outside the case.

Water cooling moves it to radiators that move it to outside air.

Traditional cpu heat sinks move it to fins that transfer the heat to the air and then the heated air is removed from the case.

It seems to me that current technology does a good job of removing heat and transferring it to the air. The heatpipes work fairly well. The large fin arrays work fairly well.

It seems to me that all this invention can do better is improve on the efficiency at which thermal paste can transfer heat from the base plate of the CPU.

Does anyone know how the efficiency of thermal paste compares with the efficiency of this invention?

ame
Posts: 488
Joined: Tue Jul 10, 2007 10:35 pm
Location: Israel

Re: Spinning heatsink

Post by ame » Wed Jul 20, 2011 4:15 am

TBH it seem to me like this device has noting to do with termal paste.

Reachable
Friend of SPCR
Posts: 396
Joined: Wed Nov 19, 2003 9:55 am
Location: Western Mass.

Re: Spinning heatsink

Post by Reachable » Wed Jul 20, 2011 9:16 am

If these spinning heatsinks do make it to PCs, one would think they could only be used in embedded processors on motherboards or customer buys whole system arrangements. How on earth could an aftermarket heatsink mounting system be rigged up when you have to work with an invariable .0001" gap?

That might be OK as far as I'm concerned. Installing the HSF, especially with the thermal paste, is the most nerve-wracking and imprecise and error possible part of assembling a PC.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Wed Jul 20, 2011 9:50 am

ame wrote:TBH it seem to me like this device has noting to do with termal paste.
It replaces the thermal paste with a thin layer (.0001") of of constantly sheared air. If you use this invention, you dispense with the need for thermal paste.

ame
Posts: 488
Joined: Tue Jul 10, 2007 10:35 pm
Location: Israel

Re: Spinning heatsink

Post by ame » Wed Jul 20, 2011 10:07 am

ces wrote:
ame wrote:TBH it seem to me like this device has noting to do with termal paste.
It replaces the thermal paste with a thin layer (.0001") of of constantly sheared air. If you use this invention, you dispense with the need for thermal paste.
I don't think that is true. As far as I can tell the air bearing (0.00~") is between the circular base plate and the impeller, not the heat source. I'd imagine on a CPU they would still use termal paste and possibly even heat pipes.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Wed Jul 20, 2011 10:58 am

ame wrote:
ces wrote:I don't think that is true. As far as I can tell the air bearing (0.00~") is between the circular base plate and the impeller, not the heat source. I'd imagine on a CPU they would still use termal paste and possibly even heat pipes.
Then why go through all the effort of running the heat through one more thermal barrier (the thin layer of 0.0001" of constantly sheared air). Why not just skip that step?

If it's big advance is a rapid current of cool air running past heated fins... there are a lot simpler and less expensive ways of doing that.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: Spinning heatsink

Post by MikeC » Wed Jul 20, 2011 12:17 pm

Finally found a few minutes to go through the documentation on this. Several salient points:

1. It's not interesting for SPCR old hands because we have already achieved lower noise and better cooling of CPUs in our PCs using 120mm fans at low speed on very large coolers. 0.2 C/W is no big deal. (Edit: except for the fact that this device is tiny.)
2. It's of greater interest for small computers where space is at a premium, and also for much higher thermal loads (like HVAC).
3. The summary info states, "The potential implications in the U.S. energy sector (air conditioners, heat pumps, and refrigeration equipment) amount to a ~5% reduction (future optimized devices could get almost 30% improvement) in electrical power consumption, significantly increased grid operating margin, and significant reduction in heat-wave generated load spikes. " Desktop PCs cannot be part of this equation, since >90% of them are at idle or just above that >90% of the time that they are powered up. Given the increasing idle efficiency of CPUs (way lower than 10W at idle on avg these days), and the widespread success of sleep mode, the savings in heat & energy for PCs are trivial.
4. The use of data from outdated, long surpassed HSFs makes the comparative analysis (in performance, cost, etc) look very weak and biased.
5. I wonder, too, about the energy efficiency of moving a whole heatsink vs. plastic fan blades. Certainly the cost of bearings, dynamic balancing, etc would have to be higher than for a fan.

jkoplow
Posts: 3
Joined: Tue Jul 19, 2011 5:42 pm

Re: Spinning heatsink

Post by jkoplow » Wed Jul 20, 2011 4:22 pm

Mr. Chin, I have to say, I expect a lot more from an editor. Unlike your readers, presumably you have a responsibility to refrain from posting "off-the-cuff" analyses of technical subject matter. Let's go through each through each one of your “salient points”:

1. It's not interesting for SPCR old hands because we have already achieved lower noise and better cooling of CPUs in our PCs using 120mm fans at low speed on very large coolers. 0.2 C/W is no big deal.

Was that the main result of the white paper, that 0.20 C/W was achieved? May I suggest you read through all of the posts on this site, the posts at ExtremeTech, and take more than “a few minutes” to review the white paper. Comment #1 above completely misses the point. Do you think the Wall Street Journal would have gone to print with an article on the subject if all we had accomplished was built a 0.2 C/W heat sink? As an editor, can you muster the time and energy to figure out what you overlooked in your cursory analysis?

2. It's of greater interest for small computers where space is at a premium, and also for much higher thermal loads (like HVAC).

Can you explain how you reached this summary judgment?

3. The summary info states, "The potential implications in the U.S. energy sector (air conditioners, heat pumps, and refrigeration equipment) amount to a ~5% reduction (future optimized devices could get almost 30% improvement) in electrical power consumption, significantly increased grid operating margin, and significant reduction in heat-wave generated load spikes. " 5% is not worth pursuing for a PC, since >90% of them are at idle or just above that >90% of the time that they are powered up. Given the increasing idle efficiency of CPUs (way lower than 10W at idle on avg these days), and the widespread success of sleep mode, the savings in heat & energy are trivial.

Did you surmise that the purported potential for reducing electricity consumption alluded to in the white paper was associated with making CPUs consume less electricity? Seriously, how much time did you spending reading the white paper?

4. The use of data from outdated, long surpassed HSFs makes the comparative analysis (in performance, cost, etc) look very weak and biased.

Did you consider what these data were actually used for in the analysis? For example, do you understand what specific conclusions were drawn using data from the outdated BitsPower heat sink with a 4800 rpm, 60 mm fan? Do you understand why these conclusions are still valid and incisive despite the fact that state-of-the-art numbers were not used in the calculation? Was the BitsPower heat sink actually used by the author as a basis of comparison with regard to CPU cooler performance (e.g. C/W)? Does the use of objective, well-defined performance metrics such as C/W strike you as biased? I would assert that the use of objective performance data such as C/W invites comparison to any other CPU cooler on the market, past, present, or future.

5. I wonder, too, about the energy efficiency of moving a whole heatsink vs. plastic fan blades. Certainly the cost of bearings, dynamic balancing, etc would have to be higher than for a fan.

That you wonder about the first issue you raise above is a clear indication that you didn’t read (or didn’t understand) the white paper. The “direct drive advantage” as it is referred to in the white paper assumes central importance in the operation of the air bearing heat exchanger, and is discussed at length in the white paper.

What about the cost of such a system? The device consists of base plate, a finned heat sink, and a brushless motor. The finned heat sink is fabricated in a one-step cold forging process, and emerges from the die as a rotationally balanced part. Die fabrication costs are of order $15,000 for a single die, and much lower when purchased in quantity. One die can be used to stamp out several hundred thousand parts before it must be replaced?

I would submit that your comments regarding cost amount to superficial speculation. Again, as an editor, I would think you would hold yourself to far higher standards. I also fail to see what value you feel you provide to your readers by furnishing an analysis of such low rigor. Many of your readers have shown far greater intellectual curiosity regarding what might come of research being undertaken in air bearing heat exchanger technology, and far more of a willingness to do their homework.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: Spinning heatsink

Post by ces » Wed Jul 20, 2011 8:44 pm

MikeC wrote:It's of greater interest for small computers where space is at a premium,
I predict space is becoming more and more of an issue. If this invention can do more cooling in less space, it does solve a problem that is going to become more and more important in the next few years.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: Spinning heatsink

Post by MikeC » Thu Jul 21, 2011 9:24 am

jkoplow --

My off-the-cuff comments were not intended to insult, but it appears your responses are meant to do just that. But I will ignore that for the sake of communication.

I'd venture that you have little idea what SPCR really does, and what kind of performance SPCR "old hands" achieve with existing technologies. We built our own 10 dBA anechoic chamber for acoustics testing, and run measurement gear accurate down to that level. Many of our Win7 lab PCs run at <15 dBA/1m using what you consider inefficient, conventional heatsink/fans. This holds true for every type of load and application we routinely throw at these PCs in the lab. We have several that are always inaudible from a meter away, and I know that many of our active forum members have achieved similar results.

W/o substantive acoustic performance data/info, your invention is simply one of a dozen or more that have splashed on the public tech stage over the past decade. Interesting, often unique, very clever, with reams of scientific R&D, & perhaps potentially interesting to those who value silence in their PCs... and yet ultimately, they've disappeared w/ nary a trace.

We know from long experience and well-established research that the perception of noise is far more than a simple SPL numeric, that a sound can often be "noisier" despite lower amplitude, or conversely, quieter despite higher SPL. The details of a fan's blade geometry, material composition, motor, rpm, etc all come into play -- but all you can say is that your "unoptimized" first working prototype is quiet compared to mainstream mass market stock coolers, in what appears to be a noisy environment. That tells us nothing at all and certainly cannot sway anyone here that your product has any inherent acoustic advantage over a good DC axial fan and conventional heatsink.

You refer many times to the high TDP of CPUs -- 150W is the number you cited in the extremetech interview -- but fail to acknowledge the point I made (along with others in this thread): Desktop CPUs simply do not run at 100% load except for brief bursts some of the time. Idle or close to idle is the norm. Several power consumption studies of desktop PCs in the past decade have shown that >90% of the time, 90% of PCs are running at idle or extremely low load. On top of this, power management in microprocessors and PCs has improved dramatically in the last few years. (As Myth! mentioned earlier, 4W at idle is typical for a CPU.) The problem you address -- cooling 150W CPUs -- is essentially non-existent in the desktop space. The most power-hungry overclocked i7-1366 on the market runs on SPCR's heatsink test platform, and iirc, its actual measured power draw is barely 100W, and that is not accounting for motherboard VRM losses.

Now, if you were trying to cooling that same CPU in a mini-ITX PC the size of a phone book, yes, your invention becomes much more relevant and interesting. (But then again, a high power CPU is simply not needed or that useful in 90% of home desktop apps.) It also has relevance on high end GPUs that actually run closer to their full TDP (as high as ~200W) more often than CPUs (mainly in gaming). Ditto in servers where 50~60% constant load is a target that many server farm managers try to achieve.

Finally, a decade of experience assessing heatsink performance convinced us some years ago that C/W is NOT an absolute objective metric for comparing heatsinks across different test platforms. Our position today is that it is best to limit comparisons of heatsinks to those done on the same platform. Fine details/differences in test platforms and environments affect C/W, and a HSF can have substantially different C/W results on different test platforms. The use of this numeric does not automatically raise your study to a higher level.

Having said all that, I still applaud your efforts to make a better mousetrap, and look forward to more optimized implementations of the technology. I'd invite you to submit any samples for us to at least measure acoustically in our anechoic chamber -- assuming you trust our ability to produce objective results. And if/when you develop the hardware that would allow one of your prototypes to actually be mounted on a real CPU, we'd be more than happy to conduct comparison cooling tests that could be used to compare noise/cooling/power vs more conventional HSFs.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Spinning heatsink

Post by CA_Steve » Thu Jul 21, 2011 1:09 pm

+1 to Mike's post.

Seems odd that a principal investigator at Sandia opts to be rude in a public forum. Then again, a username here doesn't provide provenance. The poster could just be "some dude". If he is the researcher, then perhaps he should look through this site and keep reading until he realizes our baseline is "inaudible" in a quiet environment.

Post Reply