Re: Spinning heatsink
Posted: Thu Jul 21, 2011 5:30 pm
Mike,
I apologize if you thought my last post was unfair. It’s just that it’s clear from your five salient points that you hadn’t read the material, but nonetheless felt inclined to be broadly dismissive. Your skepticism is understandable in light of what you said regarding the long string of disappointing product introductions over the past 10 years. But my opinion is that posting comments such as “0.2 C/W is no big deal” without understanding the technical context is at a minimum counterproductive. The significance of the 0.2 C/W result is that if you divide through by the surface area of our heat exchanger, we are transporting 30 times more heat per unit surface area than a conventional fan-plus-finned-heat sink arrangement. Apparently something very interesting is going on here. The implication is that if we make the heat-sink-impeller larger, with more fins, we can bring the thermal resistance MUCH lower. Saying “0.2 C/W is no big deal” is missing the entire point. In my opinion, an editor has a responsibility to make sure that he/she is fully informed before passing judgment. This is because the vast majority of people reading your post are likely to assume that everything you say is accurate and complete. If I were an editor this would be one of my greatest concerns.
With regard to the objective of CPU cooler improvement and optimization, it may be that for the most part we are talking cross purposes. For the most part, we are targeting demanding applications that can’t be addressed with today’s air cooling technology. The ability to overclock an entire server farm, for example, strikes many people as a compelling value proposition; from the comments in your 2nd post, I surmise we’re in agreement on that. Check out http://www.tomshardware.com/reviews/rad ... 97-18.html to see the power draw (idle and max), and noise levels for several contemporary Nvidia graphics cards. There are many people concerned with reducing fan noise who operate at very high power levels. Again, based on what you wrote in your last e-mail, we concur on this point as well. If the maximum amount of heat you will ever be called upon to remove is of order 5 W, then I would be the first to agree that that you will have no problem achieving what most people consider very low noise levels.
I would also agree that without quantitative acoustic data it’s hard to know just how good a given device will turn out to be in practice from the standpoint of noise (or perceived noise). On the other hand, the fact that none of the waterfall-like white noise one would expect is present when operating this device at 5000 rpm is suggestive that there is something different about this device topology from the standpoint of noise generation (and worth mentioning rather than not mentioning). I think that was part of the reason some of your readers found air bearing heat exchanger interesting. Basically what it boils down to is that unlike a conventional fan, we have a great deal of design flexibility with regard to the shape of the heat-sink-impeller blades. Thus we are free to design the fin geometry around the goal of cleanly parting, and then cleanly rejoining, the flow field at the intake and exit of the impeller channels. This decoupling of design constraints is very valuable from the standpoint of reducing noise. Having said that, if your heat sinking requirements are limited to 5 W, extremely low noise performance is readily at hand using conventional technology.
I also agree that C/W is a highly imperfect proxy for actual performance on different platforms. We are in the process of setting up a test bed for A/B comparison heat sink testing on two identical Core i7 CPUs. In addition, we are setting up a “synthetic temperature test platform” such as those used by FrostyTech for precision measurements. The data collected on such test beds (e.g. 23.2 C temperature rise at 150 W power dissipation) is widely used for inter-comparison of CPU cooler performance, and is well correlated enough with real-world performance to considered valuable.
Lastly, I appreciate your offer to assist with measurements. My understanding is that Sandia has an alternating wedge anechoic chamber down at our New Mexico facility which we can get access to during periods of time that they’re not conducting measurements for other projects. Having said that, I would be the first to agree that when a technology is mature enough to productize (ours is certainly not), it’s time to have measurements of cooling performance and noise conducted by a disinterested third party who has established widespread credibility in the community.
In summary, I think we have cleared the air. I suspect our utilitarian outlook with respect to acoustic performance (if it’s unnoticeable, it’s not of great concern) is somewhat incongruous with your priorities. If you think I’m holding you to unreasonably high standards as an editor with regard to the five salient points from your first post, maybe you’re right. At any rate, to the extent that I conveyed it in an insulting way, I apologize. I have a lot of respect for the “old hands” in areas such as this because real-world experience and know how is critically important to doing work that is well informed, credible, and reported in terms of meaningful objective metrics. One the other hand, there is sometimes a tendency to fixate on certain ways of looking at things, and when that happens, technology improvements are often agonizingly incremental over extended periods of time. What many people found interesting about the work conducted at Sandia is that we attacked the noise problem from a completely new (and highly plausible) angle, much as we did the thermal boundary layer problem, and the dust fouling problem, and that it all three goals are being approached in a manner that is mutually compatible. Air bearing heat exchanger technology will succeed or fail on the basis of its merits in real-world applications. For now, I would advocate withholding judgment in favor of exploring the potential merits of what our research team came up with by starting with a clean sheet of paper.
-Jeff
I apologize if you thought my last post was unfair. It’s just that it’s clear from your five salient points that you hadn’t read the material, but nonetheless felt inclined to be broadly dismissive. Your skepticism is understandable in light of what you said regarding the long string of disappointing product introductions over the past 10 years. But my opinion is that posting comments such as “0.2 C/W is no big deal” without understanding the technical context is at a minimum counterproductive. The significance of the 0.2 C/W result is that if you divide through by the surface area of our heat exchanger, we are transporting 30 times more heat per unit surface area than a conventional fan-plus-finned-heat sink arrangement. Apparently something very interesting is going on here. The implication is that if we make the heat-sink-impeller larger, with more fins, we can bring the thermal resistance MUCH lower. Saying “0.2 C/W is no big deal” is missing the entire point. In my opinion, an editor has a responsibility to make sure that he/she is fully informed before passing judgment. This is because the vast majority of people reading your post are likely to assume that everything you say is accurate and complete. If I were an editor this would be one of my greatest concerns.
With regard to the objective of CPU cooler improvement and optimization, it may be that for the most part we are talking cross purposes. For the most part, we are targeting demanding applications that can’t be addressed with today’s air cooling technology. The ability to overclock an entire server farm, for example, strikes many people as a compelling value proposition; from the comments in your 2nd post, I surmise we’re in agreement on that. Check out http://www.tomshardware.com/reviews/rad ... 97-18.html to see the power draw (idle and max), and noise levels for several contemporary Nvidia graphics cards. There are many people concerned with reducing fan noise who operate at very high power levels. Again, based on what you wrote in your last e-mail, we concur on this point as well. If the maximum amount of heat you will ever be called upon to remove is of order 5 W, then I would be the first to agree that that you will have no problem achieving what most people consider very low noise levels.
I would also agree that without quantitative acoustic data it’s hard to know just how good a given device will turn out to be in practice from the standpoint of noise (or perceived noise). On the other hand, the fact that none of the waterfall-like white noise one would expect is present when operating this device at 5000 rpm is suggestive that there is something different about this device topology from the standpoint of noise generation (and worth mentioning rather than not mentioning). I think that was part of the reason some of your readers found air bearing heat exchanger interesting. Basically what it boils down to is that unlike a conventional fan, we have a great deal of design flexibility with regard to the shape of the heat-sink-impeller blades. Thus we are free to design the fin geometry around the goal of cleanly parting, and then cleanly rejoining, the flow field at the intake and exit of the impeller channels. This decoupling of design constraints is very valuable from the standpoint of reducing noise. Having said that, if your heat sinking requirements are limited to 5 W, extremely low noise performance is readily at hand using conventional technology.
I also agree that C/W is a highly imperfect proxy for actual performance on different platforms. We are in the process of setting up a test bed for A/B comparison heat sink testing on two identical Core i7 CPUs. In addition, we are setting up a “synthetic temperature test platform” such as those used by FrostyTech for precision measurements. The data collected on such test beds (e.g. 23.2 C temperature rise at 150 W power dissipation) is widely used for inter-comparison of CPU cooler performance, and is well correlated enough with real-world performance to considered valuable.
Lastly, I appreciate your offer to assist with measurements. My understanding is that Sandia has an alternating wedge anechoic chamber down at our New Mexico facility which we can get access to during periods of time that they’re not conducting measurements for other projects. Having said that, I would be the first to agree that when a technology is mature enough to productize (ours is certainly not), it’s time to have measurements of cooling performance and noise conducted by a disinterested third party who has established widespread credibility in the community.
In summary, I think we have cleared the air. I suspect our utilitarian outlook with respect to acoustic performance (if it’s unnoticeable, it’s not of great concern) is somewhat incongruous with your priorities. If you think I’m holding you to unreasonably high standards as an editor with regard to the five salient points from your first post, maybe you’re right. At any rate, to the extent that I conveyed it in an insulting way, I apologize. I have a lot of respect for the “old hands” in areas such as this because real-world experience and know how is critically important to doing work that is well informed, credible, and reported in terms of meaningful objective metrics. One the other hand, there is sometimes a tendency to fixate on certain ways of looking at things, and when that happens, technology improvements are often agonizingly incremental over extended periods of time. What many people found interesting about the work conducted at Sandia is that we attacked the noise problem from a completely new (and highly plausible) angle, much as we did the thermal boundary layer problem, and the dust fouling problem, and that it all three goals are being approached in a manner that is mutually compatible. Air bearing heat exchanger technology will succeed or fail on the basis of its merits in real-world applications. For now, I would advocate withholding judgment in favor of exploring the potential merits of what our research team came up with by starting with a clean sheet of paper.
-Jeff