For those who are financially active controlling one's own investing outcome and you carry belief in the demand growth and need for this technique I have public information to share as it relates and supports this article.
A few months back Eaton, a global electrical equipment supplier, announced acquisition of a company doing exactly this "direct-to-chip cooling systems" using liquid. This is not investment advice but only information to consider for future choice that reinforces this article's claims. Many here are aware of the foundational issues fast approaching while this large public company has already acted through acquisition to offer a direct liquid cooling solution to its customers. This solution comes at a higher Capex cost yet the long term Opex savings on energy expenses will more than justify that cost over time as compared to current methods using indirect cooling via air as the heat transfer medium. For the many here that have and still work and live in the heat and noise of datacenters removing one of the two will be a win, also direct liquid cooling is likely to reduce noise as well as fewer fans are needed to move air.
I also believe that in time, given the creativity of our species, that someone will devise an applied use for this captured heat in liquid just as some landfills pipe venting methane to local heavy industries for revenue instead of just wasting it via flared burn off.
Surreal times but this also brings significant opportunity of which energy savings will be top of mind for many.
> Due to a phenomenon known as Joule heating – an unavoidable consequence of how they operate at a fundamental level – chips dissipate almost exactly the amount of power they consume as heat.
I don't think you'll ever make a chip not dissipate as heat the energy you feed into it for operation. Where else would it go?
I think the main idea is that MOSFET chips don’t sink or source much current on their own (“fundamental” seeming to mean the way that current does not propagate into the gates, and most designs only really draw current during transitions)
OTOH some chips (amplifiers for example) may indeed have current flowing through them and therefore the power consumption of the “chip” would equal the sum of heat loss and output power. At least that’s my interpretation of the framing “how they operate at a fundamental level”. I could be wrong too, I’m not a working EE
LEDs are chips; some of the energy goes to light (which will be heat eventually, but not necessarily locally). Some chips also have measurable piezo effects, turning some energy into sound. And of course various data storage technologies (eg floating gate flash) have higher potential energy in one state than the other, so write operations may take more energy than is converted to heat.
I heard copper is better an transferring heat but aluminum is better at RELEASING heat via airflow. Hence you see copper tubes on cpu coolers that terminate in aluminum fins.
The capacity of an object to radiate heat has essentially nothing to do with material properties. The equation is entirely dominated by surface area and nothing else. It is purely a matter of geometry. Which is why we use aluminum: it's cheaper and its lower thermal conductivity is pretty much irrelevant.
Not really. They use aluminum fins because they are way lighter and cheaper. The copper tubes are actually heat pipes that transfer the heat through liquid/vapor phase change. And copper is used because the liquid inside is water (aluminum would corrode) and they are also easier to bend into shape (aluminum fatigues easier with bends, and pores would allow the liquid to escape).
In the energy results they are comparing their novel water block cold plate against an air cooled facility, not against a similar water-cooled facility.
For those who are financially active controlling one's own investing outcome and you carry belief in the demand growth and need for this technique I have public information to share as it relates and supports this article.
A few months back Eaton, a global electrical equipment supplier, announced acquisition of a company doing exactly this "direct-to-chip cooling systems" using liquid. This is not investment advice but only information to consider for future choice that reinforces this article's claims. Many here are aware of the foundational issues fast approaching while this large public company has already acted through acquisition to offer a direct liquid cooling solution to its customers. This solution comes at a higher Capex cost yet the long term Opex savings on energy expenses will more than justify that cost over time as compared to current methods using indirect cooling via air as the heat transfer medium. For the many here that have and still work and live in the heat and noise of datacenters removing one of the two will be a win, also direct liquid cooling is likely to reduce noise as well as fewer fans are needed to move air.
I also believe that in time, given the creativity of our species, that someone will devise an applied use for this captured heat in liquid just as some landfills pipe venting methane to local heavy industries for revenue instead of just wasting it via flared burn off.
Surreal times but this also brings significant opportunity of which energy savings will be top of mind for many.
Stay Healthy!
90% of 30% of total energy use. So, actually 27%. What a title.
Original publication - https://www.cell.com/cell-reports-physical-science/fulltext/...
> Due to a phenomenon known as Joule heating – an unavoidable consequence of how they operate at a fundamental level – chips dissipate almost exactly the amount of power they consume as heat.
I don't think you'll ever make a chip not dissipate as heat the energy you feed into it for operation. Where else would it go?
I think the main idea is that MOSFET chips don’t sink or source much current on their own (“fundamental” seeming to mean the way that current does not propagate into the gates, and most designs only really draw current during transitions)
OTOH some chips (amplifiers for example) may indeed have current flowing through them and therefore the power consumption of the “chip” would equal the sum of heat loss and output power. At least that’s my interpretation of the framing “how they operate at a fundamental level”. I could be wrong too, I’m not a working EE
It could go into the information directly. https://en.wikipedia.org/wiki/Landauer%27s_principle
LEDs are chips; some of the energy goes to light (which will be heat eventually, but not necessarily locally). Some chips also have measurable piezo effects, turning some energy into sound. And of course various data storage technologies (eg floating gate flash) have higher potential energy in one state than the other, so write operations may take more energy than is converted to heat.
We should block all stories with "could" in the headline; "could" implies conjecture - not news.
I heard copper is better an transferring heat but aluminum is better at RELEASING heat via airflow. Hence you see copper tubes on cpu coolers that terminate in aluminum fins.
The capacity of an object to radiate heat has essentially nothing to do with material properties. The equation is entirely dominated by surface area and nothing else. It is purely a matter of geometry. Which is why we use aluminum: it's cheaper and its lower thermal conductivity is pretty much irrelevant.
Not really. They use aluminum fins because they are way lighter and cheaper. The copper tubes are actually heat pipes that transfer the heat through liquid/vapor phase change. And copper is used because the liquid inside is water (aluminum would corrode) and they are also easier to bend into shape (aluminum fatigues easier with bends, and pores would allow the liquid to escape).
In the energy results they are comparing their novel water block cold plate against an air cooled facility, not against a similar water-cooled facility.
Copper is expensive, and the manufacturing process sounds finicky.
Until the micro/nano-patterned surface gets dirty at least.