Keeping Computers Cool From the Inside

IBM internal server cooling system — IBM's System x iDataPlex dx360 M4 internal server cooling system.

The issue of how to keep computers cool generally calls to mind the techniques used in enterprise data centers; that is where energy has traditionally been concentrated, so to speak. Yet, improved cooling from the inside out is crucial to reducing the carbon footprint of computers—otherwise, the need to consume increasing amounts of energy to offset the heat generated by computers will severely hamper the advancement of computing, experts in the field say.

Cooling the data center is no longer enough. In a moderate climate, for example, air-cooling a data center constitutes about 30% of the total energy it consumes and is a substantial contributor to its carbon footprint, according to a 2010 IBM report Direct Waste Heat Utilization From Liquid-Cooled Supercomputers. Room air conditioning is the second-largest energy component of data center operations and is highly inefficient, the report’s authors state, adding, “More than 40% of the carbon footprint of an air-cooled data center is caused not by computing but by powering the cooling systems needed to keep the microelectronic components from overheating.”

Advances in micro-miniaturization of electronic components required rapid growth in transistor density and a rise of clock frequency of integrated electronic circuits; this, in turn, led to increased heat generation rates in electronic chips. While chips are currently cooled primarily by forced air convection, some industry observers believe this technique will not be sufficient for next-generation electronics, which require more efficient and compact cooling solutions to maintain acceptable operating temperatures.

The computer industry has become hindered by the thermal issue, which is among the top five most pressing concerns of computer companies and users, maintains Bruno Michel, manager of Advanced Thermal Packaging at IBM Zurich Research Laboratory and one of the report’s authors. “The risk is that development will stop in just a few years after the 30 years of Moore’s Law,” he said. With data centers making up about 2% of overall U.S. electrical usage and the rapid growth of data processing translating to an increase in energy consumption, the challenge for computer manufacturers is to provide more energy-efficient products.

Fortunately, progress is being made by a number of research institutes and computer technology companies.

At Purdue University, for example, the Cooling Technologies Research Center (CTRC) has a number of projects under way focused on high-conductivity thermal heat spreaders; specifically, vapor chambers and heat pipes. These devices are small, sealed chambers in the same form factor as standard metal heat spreaders, which “efficiently transport heat caused by cyclical vaporization and condensation of an internal working fluid,” according to Suresh V. Garimella, Goodson Distinguished Professor of Mechanical Engineering at Purdue. The CTRC has also worked on developing physics-based models for the devices’ performance, and to experimentally measure their thermal transport characteristics. The Center’s work covers the entire spectrum of heat dissipation from the chip, board, and building scale, he says.

Replacing traditional Al/Cu heat spreaders with appropriate, low-cost, new materials can reduce the total thermal resistance of a cooling package, according to the CTRC. Carbon nanotubes (CNTs) can play a key role in this; CNTs are synthesized on heat spreaders to improve heat transfer at interface, says Garimella.

“One of the ubiquitous thermal management challenges is the thermal resistance inherent to contact between any system components,” he explains. “Often, the performance of [even] highly optimized thermal packages is governed by these resistances. Additionally, the mechanical stresses [that] develop from thermal cycling at these interfaces often dictates the entire system’s reliability.” He says CNTs have a “high inherent thermal conductivity, which makes them a suitable material for joining components with a low thermal contact resistance.” Their mechanical properties, which allow flexibility between interfaces for expansion and contraction with thermal cycling, may also improve the reliability of these interfaces, compared to other attachment methods.

Replacing traditional Al/Cu heat spreaders with appropriate, low-cost, new materials can reduce the total thermal resistance of a cooling package, according to the CTRC.

Vapor chamber technologies are also being investigated for use in portable and handheld electronic devices, says Garimella. “There is specific interest in taking advantage of the reduced thermal resistance for heat spreading provided by these devices, but their overall size and thickness have historically been unfeasibly large,” he says. “We will study the operational limits of heat pipes both experimentally and theoretically for very thin geometries, with an ultimate goal of providing heat pipe design guidelines for these applications.” The goal of the center’s work is to come up with commercially viable, ultra-thin heat pipes within the next five years, he adds.

Water is considered a better coolant for microprocessors than air because of its higher heat capacity and thermal conductivity, according to the report Experimental Investigation of a Hot Water Cooled Heat Sink for Efficient Data Center Cooling: Towards Electronic Cooling with High Exergetic Utility. “Water cooling using compact manifold micro-channel (MMC) heat sinks is one of the most promising strategies to meet the cooling requirement of chips in next-generation data centers,” write the report’s authors. Using hot water as a coolant in state-of-the-art processors can eliminate the need for air cooling entirely, ensuring “a high exergetic utility in the system,” along with heat sinks that are designed to be robust and scalable.

High-performance microchannel heat sinks and other liquid cooling solutions are intended to dissipate high heat fluxes at the local chip level, concurs Garimella, and are among the technologies applicable to use to augment exterior cooling options. He adds that, “Typically, ultimate heat rejection occurs to ambient air, and there are a number of low-power-consumption air cooling technologies such as piezoelectric fans, synthetic jets, and mist cooling strategies that address this need.”

At giant chipmaker Intel Corp., much work has been done on not just cooling chips, but also in reducing the load of what needs to be cooled, says principal engineer Michael K. Patterson, who is focused on thermal, power, and energy efficiency. “We’ve done a lot to reduce the idle power on computers. In the past there was very little difference between idle power and peak power; if the computer was doing nothing, it used pretty much the same amount of power as when it was very active, and at a high utilization rate.”

One area of concentration for Intel has been on changing the speed of a chip, and creating the ability to slow it down if it does not need its full clock speed, while keeping just enough performance, he says. “If it’s waiting for network traffic … it will slow itself down and use less energy, which obviously helps … anywhere between idle and peak power.”

When the chip is not running as fast and not generating as much heat, internal fans slow down, reducing the energy needed to run them. “As we reduce the silicon load, the support energies get reduced too, and that’s where we can make a difference,” Patterson says.

With each generation of Intel chips, Patterson says, the goal is to try to reduce power consumption and improve performance at the same time, thereby increasing performance per watt. The impact of this could range from more battery life for your phone, tablet, or notebook computer at the portable product level, to getting more science done for your energy investment in high-performance computing at the enterprise or supercomputer level, he says.

Commercial Viability

Besides more energy-efficient chips, other products are also becoming commercialized that are aimed at reducing the heat load of electronics.

Ionic wind cooling and piezoelectric fans are available, for example, although piezo fans are a relatively new technology that has not yet found its way into a wide variety of applications. While offering benefits to large computers/servers, the bulk of the focus for piezo fans appears to be in smaller consumer electronics, according to Nick Gilligan, production manager at Mide Technology Corp., an engineering company that makes the fans.

Piezo fans operate on the principle of piezoelectricity (an inherent material property in which electricity is created as a result of mechanical stress), and they do not suffer the same pitfalls as traditional fans, Gilligan says. Piezoelectric fans are typically configured as a cantilever beam, with a tip that can achieve “significant displacement to create airflow,” he says. “Piezoelectric fans offer extremely high reliability and do not suffer from the failure modes inherent in traditional fan designs. There are no moving parts to break, they are nearly silent, and have practically infinite lifespans under known driving conditions.”

GE is using “Dual Piezoelectric Cooling Jets” (DCJ) that are each 1mm thick in a prototype Core i7-powered laptop. Based on a technology GE researchers first developed for commercial jet engines, they “behave as a micro-fluidic bellows that provide high-velocity jets of air to cool electronic components,” according to GE. These jets consume a fraction of the power of a traditional fan, and so the cooling technology could add an extra 30 minutes of battery life to a laptop, GE believes. Additionally, the company says, the dual piezo cooling jets can move the same amount of air as a cooling fan twice its size.

Piezoelectric fans have “no moving parts to break, they are nearly silent, and have practically infinite lifespans under known driving conditions.”

Other cooling technologies under development include high-heat flux vapor chamber heat spreaders for military applications, and dielectric fluid microchannel heat sinks for automotive power electronics cooling.

To explore the use of vapor chambers (passive heat transport devices that utilize capillary forces to circulate a fluid to capture waste heat in electronic devices and transfer it to a heat sink or cold plate for dissipation) and heat pipes in military applications, a group of CTRC faculty was awarded a $2.7-million U.S. Defense Advanced Research Projects Agency (DARPA) grant in partnership with Raytheon Company from 20082011.

Purdue aided in the development of a “Radio Frequency Thermal Ground Plane” (RFTGP) that provides state-of-the-art commercial heat spreaders with what Garimella calls significantly improved performance. As part of this work, Raytheon was provided an experimentally validated model for heat transport in vapor chambers, which Garimella says has improved the readiness of the device, and has led to additional reliability testing. “The thermal resistance reduction benefit provided by the developed RFTGP will have direct impact by decreasing the operational temperature and increasing the reliability of current solid-state devices,” he explains. Raytheon expects the RFTGP technology to have a significant impact on both defense and commercial electronics packaging technology, he notes.

Converting Heat into Electricity

Thermoelectric generators (TEGs), also known as thermogenerators, are devices that convert heat directly into electricity. Garimella says it is possible that TEGs will be able to recharge portable electronics with their own waste heat or, at least, to diminish the power consumption of office or portable electronics.

Energy efficiency is arguably a major challenge of the 21st century, claim the authors of IBM’s liquid-cooled supercomputers report.

“Unlike larger-scale thermodynamic waste heat recovery cycles, such as the organic Rankine cycles currently being researched in the Center, thermoelectric devices scale favorably to small sizes,” he says. That makes it feasible—from a size standpoint—to embed thermoelectric modules for waste heat recovery in device-scale systems, he says. However, he warns, the efficiency of these thermoelectrics is “still too low for economic feasibility at these scales.”

Energy efficiency is arguably a major challenge of the 21st century, claim the authors of IBM’s liquid-cooled supercomputers report. Computer systems can no longer be designed solely on the criteria of computational performance, they maintain, stating “The new target must be high performance and low net power consumption (and, concomitantly, low net carbon footprint).” This requires a continued focus on more energy-efficient processors which, in turn, means striving for improved transistor designs, as well as high-performance cooling capabilities that reduce power consumption.