Power consumption is a major factor that limits the performance of computers. We survey the “state of the art” in techniques that reduce the total power consumed by a microprocessor system over time. These techniques are applied at various levels ranging from circuits to architectures, architectures to system software, and system software to applications. They also include holistic approaches that will become more important over the next decade. We conclude that power management is a multifaceted discipline that is continually expanding with new techniques being developed at every level. These techniques may eventually allow computers to break through the “power wall” and achieve unprecedented levels of performance, versatility, and reliability. Yet it remains too early to tell which techniques will ultimately solve the power problem.

/pdf/power-reduction-techniques-for-microprocessor-systems-3dfczj5khm.pdf

Power reduction techniques for microprocessor systems

The IBM S/390Â® G4 CMOS system, first shipped in 1997, was the first high-end system to use refrigeration. The decision to employ refrigeration cooling instead of other cooling options such as high-flow air cooling or various water-cooling schemes focused on the potential system performance improvement obtainable by lowering coolant temperatures using a refrigeration system. This paper reviews the historical background of refrigeration from its use in the early 1800s to its implementation in computer systems in the early 1990s. The advantages and disadvantages of using refrigeration in the cooling of computer systems are examined. The advantages have outweighed the disadvantages, leading to the first use by IBM of refrigeration in cooling the S/390 G4 server. The design of the refrigeration system for the S/390 G4 system is described in detail, and some of the key parametric studies that contributed to the final design are described.

High-end server low-temperature cooling

A cooling system employs a cooling liquid and a cooling gas in a combined thermodynamic cycle to overcome the flow resistance of dense assemblies of heat generating components and to improve heat transfer by inducing turbulence, thereby reducing the effects of thermal hysteresis and boundary layer formation. Sensible heat gain to the cooling liquid and gas and latent heat of vaporization of the cooling liquid also occur in channels through and over the components. The flow of cooling gas propels the cooling liquid through the channels. The cooling system is advantageous for cooling electronic components such as integrated circuits which exhibit relatively high degree of energy and physical density, in supercomputers. The cooling system may also be advantageously combined with an immersion cooling system for the power supply components in the computer.

Gas-liquid forced turbulence cooling

The advantages of operating CMOS at liquid nitrogen temperature (LN) are attributed to increased carrier mobility, reduced subthreshold swing, increased conductivity, reduced leakage, an improved device and circuit reliability such as electromigration ionic migration, and latchup. The gain in performance compared to room temperature (RT), however, begins to erode as the channel length is decreased below one micrometer, where increased lateral field causes the drift velocity to approach its scattering limited value along a large fraction of the channel. In the limit when saturation velocity is reached along the entire channel, the improvement at LN does not exceed a factor of approximately=2, after all other enhancements are considered. This gain must justify the added inconvenience and cost of operating the system at LN temperatures. As the channel length is reduced to deep submicrometer, below approximately=0.15 mu m, operating at low temperature could become a necessity rather than mere improvement over RT, because of the lack of a room-temperature process and device design point. >

Low-temperature CMOS-a brief preview

Thermodynamic optimization of cooling techniques for electronic packages

The thermal aspects are discussed of an immersion-cooled computer designed to operate at cryogenic temperatures. Several of the major thermal design issues are discussed in detail. These include: (1) the selection of a working fluid; (2) the determination of the mode, or modes, of heat transfer to be used: (3) the selection, or development, of any required heat transfer correlations; and (4) the selection of a refrigeration system. The results of a preliminary analysis of a candidate thermal control system are also presented. >

Design of a candidate thermal control system for a cryogenically cooled computer

The performance and costs of three types of cryogenic cooling systems are analyzed; a once-through system in which liquid nitrogen is allowed to boil on the CPU boards and is then vented to the surroundings; a saturated pool boiling refrigerated system in which gaseous nitrogen generated from boiling on the CPU boards in condensed in the vapor space above the liquid nitrogen pool; and a subcooled pool boiling refrigerated system in which the condenser is completely submerged in a subcooled liquid nitrogen pool surrounding the CPU boards. Thermal/fluid numerical models used in conjunction with an optimization code GRG2, were used to predict the performance and cost of the cryogenic cooling systems. The uniform annual cost of the once-through system was calculated to be $28.8 K. This compares with $34.6 K for the optimal saturated pool boiling system and $34.7 K for the optimal subcooled pool boiling system. Due to the extremely high initial cost of the refrigerated systems, the once-through cryogenic cooling system is recommended if the local heat flux on the CPU board remains below 15 W/cm/sup 2/. >

Optimization of the thermal design of a cryogenically cooled computer

This work addresses the thermal aspects of an immersion-cooled computer designed to operate at cryogenic temperatures. Several of the major thermal design issues are discussed in detail. These include: (1) the selection of a working fluid, (2) the determination of the mode, or modes, of heat transfer to be employed, (3) the selection, or development, of any required heat transfer correlations, and (4) the selection of a refrigeration system. The results of a preliminary analysis of a candidate thermal control system are also presented.

http://ieeexplore.ieee.org/iel2/742/1166/00028690.pdf

J.R. Parsons

Papers

Design of a candidate thermal control system for a cryogenically cooled computer

Optimization of the thermal design of a cryogenically cooled computer

DESIGN OF A CANDIDATE THERMAL CONTROL SYSTEM FOR A CRYOGENICALLY-COOLED COMPUTER Complete manuscript reviewed