Temperature-aware microarchitecture

Skadron, Kevin; Stan, Mircea R.; Huang, Wei; Velusamy, Sivakumar; Sankaranarayanan, Karthik; Tarjan, David

doi:10.1145/859618.859620

Cited by 773 publications

(330 citation statements)

References 15 publications

Supporting

Mentioning

330

Contrasting

Order By: Relevance

“…The first full-scale chip in our study is adopted from the chip named Compaq Alpha 21364, which was used by Skadron et al [39] to introduce the HotSpot, a fast and practical thermal modeling approach for chip simulation. The floor plan of Compaq Alpha 21364 is shown in Fig.…”

Section: Whole Chip Modelingmentioning

confidence: 99%

Porous Media Modeling of Microchannel Cooled Electronic Chips with Nonuniform Heating

Yao

2015

Journal of Thermophysics and Heat Transfer

View full text Add to dashboard Cite

Microchannels are used for the cooling of electronic chips. However, the three-dimensional computational fluid dynamics modeling of the large number of channels in a full chip requires a huge number of meshes and computation time. Although porous media modeling of microchannels can significantly reduce the effort of simulation, most previous porous media models are based upon the assumption that the surface heat flux or temperature is uniform on the chip. In reality, the heat flux on the chip is usually highly nonuniform. In the present study, the porous media model considers the simultaneously developing entrance effect at the microchannel inlet and the thermally developing entrance effect due to the severe heat flux variation along the channel. Duhamel's integral is used to provide the Nusselt number distribution corresponding to the nonuniform heat flux distribution along the channel. The computing cost of this modeling method is only about 1% of the three-dimensional conjugate simulation. This porous media thermal modeling method is applied to model two full-scale electronic chips with realistic power distributions on the surfaces, and temperature maps are generated. The porous media thermal modeling offered by this study is an accurate and efficient alternative for modeling the electronic chips cooled by microchannels. Nomenclature a = wetted area per volume, m −1 c p = thermal capacity of the fluid, J∕kg · K D h = channel hydraulic diameter, m f = Fanning friction factor, τ w ∕1∕2ρV 2 H c = channel height, m H s = substrate height, m h = convective heat transfer coefficient, W∕m 2 · K K = permeability of the porous media, m 2 k 1 = thermal conductivity of solid substrate, W∕m · K k 2 = thermal conductivity of liquid coolant, W∕m · K L = channel length, m L = characteristic length scale, m L i = length of different periods in single-channel case, i 1; : : : ; 5, m Nu = Nusselt number, hL∕k 2 Po = Poiseuille number, fRe Pr = Prandtl number, c p μ∕k 2 q w = heat flux on top of the substrate, W∕m 2 q wi = heat flux of different periods in single-channel case, i 1; : : : ; 5, W∕m 2 Re = Reynolds number, ρVD h ∕μ s = fin width, m T = temperature,°C V = mean velocity of fluid flow, m∕s V D = filter velocity of porous media, m∕s W = width of modeling region, m W i = design parameters for split-flow-type design, i 1; : : : ; 3, m W inlet , W outlet = inlet and outlet width for split-flow-type design, m x, y, z = Cartesian coordinates, m Y i = design parameters for split-flow-type design, i 1; : : : ; 3, m y = dimensionless hydraulic axial distance, y∕LRe L y = dimensionless thermal axial distance, y∕LRe L Pr y th = dimensionless thermal entrance length, y th ∕LRe L Pr α = channel aspect ratio, δ∕H c < 1 Δp = pressure drop, Pa δ = channel width, m ε = porosity of the porous media κ = Hagenbach factor μ = dynamic viscosity of fluid, kg∕m · s ξ = integration variable in Duhamel integration ρ = density of fluid, kg∕m 3 Subscripts app = apparent DP = developing FD = fully developed i = local in = inlet s = solid substrate

show abstract

Section: Whole Chip Modelingmentioning

confidence: 99%

Porous Media Modeling of Microchannel Cooled Electronic Chips with Nonuniform Heating

Yao

2015

Journal of Thermophysics and Heat Transfer

View full text Add to dashboard Cite

show abstract

“…One such design tool is HotSpot (Skadron et al, 2003) used for microprocessors, which models temperature distribution using thermal resistances and capacitances derived from the layout of micro-architectural structures that have been validated using finite element simulations. One such design tool is HotSpot (Skadron et al, 2003) used for microprocessors, which models temperature distribution using thermal resistances and capacitances derived from the layout of micro-architectural structures that have been validated using finite element simulations.…”

Section: Thermal Modelling Of Serversmentioning

confidence: 99%

“…Such estimations have been used for developing temperature triggered computer systems scheduling (Weissel and Bellosa, 2004). These tools, which allow integrated performance and power/thermal studies, have been applied extensively (Brooks and Martonosi, 2001;Kim et al, 2006;Gomaa et al, 2004;Gurumurthi et al, 2003;Pinheiro and Bianchini, 2004;Shang et al, 2004;Skadron et al, 2003) in the architecture community for reducing power/temperature. These tools, which allow integrated performance and power/thermal studies, have been applied extensively (Brooks and Martonosi, 2001;Kim et al, 2006;Gomaa et al, 2004;Gurumurthi et al, 2003;Pinheiro and Bianchini, 2004;Shang et al, 2004;Skadron et al, 2003) in the architecture community for reducing power/temperature.…”

Section: Thermal Modelling Of Serversmentioning

confidence: 99%

CFD modelling of thermal distribution in industrial server centres for configuration optimisation and energy efficiency

Paradis

Ramdenee

Ilinca

et al. 2014

IJSPM

View full text Add to dashboard Cite

The use of servers for computational and communication control tasks is becoming more and more frequent in industries and institutions. Ever increasing computational power and data storage combined with reduction in chipsets size resulted in the increased heat density and need for proper configurations of the server racks to enhance cooling and energy efficiency. While different methods can be used to model and design new server centres and optimise their configuration, there is no clear guideline in the literature on the best way to design them and how to increase energy efficiency of existing server centres. This paper presents a simplified yet reliable computational fluid dynamics (CFD) model used to qualitatively evaluate different cooling solutions of a data centre and proposes guidelines to improve its energy efficiency. The influence of different parameters and configurations on the cooling load of the server room is then analysed. , he works as research manager at the Wind Energy TechnoCentre (Canada). He is actively involved in R&D activities, more specifically in the theme of wind energy application in remote areas, wind energy storage systems, compressed air energy storage, wind-diesel-compressed air hybrid system, heat and mass transfer, Fluid dynamics and de-icing methods of wind turbines operate in cold climate. This paper is a revised and expanded version of a paper entitled 'CFD modelling of thermal distribution in industrial server centres for configuration optimization and control' presented at

show abstract

“…Foe example, dynamic temperature management (DTM) is performed by the hardware. These DTM techniques include dynamic voltage and frequency scaling [1,2], fetch toggling, and fetch throttling [1,3]. Even if they are effective in cooling on-chip hot spots by stopping or slowing down the processor when the temperature exceeds the threshold, they always pay a hardware cost and lead to significant performance degradation.…”

Section: Introductionmentioning

confidence: 99%

Thermal-Aware Code Transformation across Functional Units

Chen

Chang

2011

2011 IFIP 9th International Conference on Embedded and Ubiquitous Computing

View full text Add to dashboard Cite

SUMMARYVarious functional units (FUs) have been designed in modern embedded processors to perform different functions when running an application. Traditionally, compilers did not take this characteristic into account to reduce the temperature of a processor during execution. For many applications, the occurrences of different instructions are not the same after they are compiled. As a consequence, the temperature of the processor is very high, arising from the major heating contribution of the special structure of the active functional unit, and thus, the system will suffer severe damage. Consequently, to remedy this hurdle, this paper provides a solution by shifting the loading from heavy-loading FUs to light-loading FUs. Our approach first identifies all the FUs that can exchange the loading among themselves and then presents a thermal model for these exchangeable FUs to estimate the temperature impact on shifting loading. Finally, the loading shifting has been performed by transforming code under the consideration of limited performance loss without hardware cost. The result shows that our approach can reduce the temperature to a small cost of performance degradation and code expansion.

show abstract

Temperature-aware microarchitecture

Cited by 773 publications

References 15 publications

Porous Media Modeling of Microchannel Cooled Electronic Chips with Nonuniform Heating

Porous Media Modeling of Microchannel Cooled Electronic Chips with Nonuniform Heating

CFD modelling of thermal distribution in industrial server centres for configuration optimisation and energy efficiency

Thermal-Aware Code Transformation across Functional Units

Contact Info

Product

Resources

About