Navid Borhani

Towards thermally-aware design of 3D MPSoCs with inter-tier cooling

New tendencies envisage 3D Multi-Processor System-On-Chip (MPSoC) design as a promising solution to keep increasing the performance of the next-generation high-performance computing (HPC) systems. However, as the power density of HPC systems increases with the arrival of 3D MPSoCs, supplying electrical power to the computing equipment and constantly removing the generated heat is rapidly becoming the dominant cost in any HPC facility. Thus, both power and thermal/cooling implications play a major role in the design of new HPC systems, given the energy constraints in our society. Therefore, EPFL, IBM and ETHZ have been working within the CMOSAIC Nano-Tera.ch program project in the last three years on the development of a holistic thermally-aware design. This paper presents the exploration in CMOSAIC of novel cooling technologies, as well as suitable thermal modeling and system-level design methods, which are all necessary to develop 3D MPSoCs with inter-tier liquid cooling systems. As a result, we develop energy-efficient run-time thermal control strategies to achieve energy-efficient cooling mechanisms to compress almost 1 Tera nano-sized functional units into one cubic centimeter with a 10 to 100 fold higher connectivity than otherwise possible. The proposed thermally-aware design paradigm includes exploring the synergies of hardware-, software- and mechanical-based thermal control techniques as a fundamental step to design 3D MPSoCs for HPC systems. More precisely, we target the use of inter-tier coolants ranging from liquid water and two-phase refrigerants to novel engineered environmentally friendly nano-fluids, as well as using specifically designed micro-channel arrangements, in combination with the use of dynamic thermal management at system-level to tune the flow rate of the coolant in each micro-channel to achieve thermally-balanced 3D-ICs. Our management strategy prevents the system from surpassing the given threshold temperature while achieving up to 67% reduction in cooling energy and up to 30% reduction in system-level energy in comparison to setting the flow rate at the maximum value to handle the worst-case temperature.

Boiling on a Tube Bundle: Part II—Heat Transfer and Pressure Drop

An extensive experimental study was undertaken to measure nucleate pool boiling heat transfer coefficients, local bundle boiling heat transfer coefficients, and two-phase bundle pressure drops for R134a and R236fa on one plain tube bundle configuration. The experimental database allowed the refinement of frictional pressure drop models previously developed at the Laboratory of Heat and Mass Transfer. Together with the new onset of dryout prediction method presented in part I (preceding article in this issue), this constitutes a significant improvement in such prediction methods. The local bundle boiling heat transfer data highlighted the dependency of the heat transfer coefficient on the heat flux as expected for the present conditions. The new method was proposed and worked well versus the present database and was also validated against additional refrigerants from independent studies. It was proven to also work reasonably well for falling film evaporation data, proving the new prediction method is applicable for a wide range of operating conditions.

3D ALE Finite-Element Method for Two-Phase Flows With Phase Change

We seek to study numerically two-phase flow phenomena with phase change through the finite-element method (FEM) and the arbitrary Lagrangian–Eulerian (ALE) framework. This method is based on the so-called “one-fluid” formulation; thus, only one set of equations is used to describe the flow field at the vapor and liquid phases. The equations are discretized on an unstructured tetrahedron mesh and the interface between the phases is defined by a triangular surface, which is a subset of the three-dimensional mesh. The Navier–Stokes equation is used to model the fluid flow with the inclusion of a source term to compute the interfacial forces that arise from two-phase flows. The continuity and energy equations are slightly modified to take into account the heat and mass transport between the different phases. Such a methodology can be employed to study accurately many problems, such as oil extraction and refinement in the petroleum area, design of refrigeration systems, modeling of biological systems, and efficient cooling of electronics for computational purposes, which is the aim of this research. A comparison of the obtained numerical results to the classical literature is performed and presented in this paper, thus proving the capability of the proposed new methodology as a platform for the study of diabatic two-phase flows.

A 3D moving mesh Finite Element Method for two-phase flows

A 3D ALE Finite Element Method is developed to study two-phase flow phenomena using a new discretization method to compute the surface tension forces. The computational method is based on the Arbitrary Lagrangian-Eulerian formulation (ALE) and the Finite Element Method (FEM), creating a two-phase method with an improved model for the liquid-gas interface. An adaptive mesh update procedure is also proposed for effective management of the mesh to remove, add and repair elements, since the computational mesh nodes move according to the flow. The ALE description explicitly defines the two-phase interface position by a set of interconnected nodes which ensures a sharp representation of the boundary, including the role of the surface tension. The proposed methodology for computing the curvature leads to accurate results with moderate programming effort and computational cost. Static and dynamic tests have been carried out to validate the method and the results have compared well to analytical solutions and experimental results found in the literature, demonstrating that the new proposed methodology provides good accuracy to describe the interfacial forces and bubble dynamics. This paper focuses on the description of the proposed methodology, with particular emphasis on the discretization of the surface tension force, the new remeshing technique, and the validation results. Additionally, a microchannel simulation in complex geometry is presented for two elongated bubbles

Two-phase flow boiling in a single layer of future high-performance 3D stacked computer chips

The present study focuses on an experimental investigation of two-phase flow boiling in a silicon multi-microchannel evaporator, which emulates a single layer of a 3D stacked computer chip. The micro-evaporator is comprised of 67 parallel channels, each having a 100 × 100 μm2 cross-section area, and separated by 50 μm-wide fins. Two aluminium micro-heaters were sputtered onto the backside of the test section to provide two 0.5 cm2 heated areas in order to simulate the power dissipated by active component in 3D CMOS chips. The experiments were performed with a second identical test section having 50 μm-wide, 100 μm-deep, and 100 μm-long restrictions (micro-orifices) at the inlet of each channel to stabilize the two-phase flow. The goal of this experimental campaign was to perform simultaneous high-speed flow visualization and infra-red measurements of the two-phase flow and heat transfer dynamics across the entire micro-evaporator area. Refrigerants R245fa, R236fa and R1234ze(E) were chosen as the working fluids. The micro-orifices successfully suppressed back flow, eliminated flow instabilities, provided a good flow distribution, and started the boiling process with some flashed vapor. Thermal performance was found to be uniform widthwise using these orifices.

Boiling on a Tube Bundle: Part I-Flow Visualization and Onset of Dryout

A novel visualization method was developed and implemented in a refrigerant-flooded evaporator tube bundle. Specifically, a method to visualize the two-phase flow inside plain tube bundles was developed using a dummy tube with a tubular glass section for visual access with a digital high-speed video camera and a laser light/photodiode. These were used to observe and characterize the two-phase flow structure and improve understanding of the physical phenomena taking place. In particular, a dryout line was objectively identified in a flow pattern map separating high- and low-performance heat transfer data measured locally on the bundle. The database was recorded for vertical flow of R134a and R236fa over smooth tubes at saturation temperatures of 5–15°C. The tests span low mass velocities applicable in flooded evaporators, and heat fluxes include adiabatic and diabatic tests of 7–25 kW/m2. A new prediction method for onset of dryout on smooth tube bundles is proposed.

Learning to see through multimode fibers

We use Deep Neural Networks (DNNs) to classify and reconstruct a large database of handwritten digits from the intensity of the speckle patterns that result after the images propagated through multimode fibers (MMF). Images transmitted through fibers with up to 1km length were recovered. The ability of the network to recognize the input degraded with fiber length but the performance could be enhanced if the neural networks were trained to first reconstruct the image and then classify it rather than classify it directly from the speckle intensity.

Thermal Response of Multi-Microchannel Evaporators During Flow Boiling of Refrigerants Under Transient Heat Loads With Flow Visualization

Multi-microchannel evaporators with flow boiling, used for cooling high heat flux devices, usually experience transient heat loads in practical applications. These transient processes may cause failure of devices due to a thermal excursion or poor local cooling or dryout. However, experimental studies on such transient thermal behavior of multi-microchannel evaporators during flow boiling are few. Thus, an extensive experimental study was conducted to investigate the base temperature response of multi-microchannel evaporators under transient heat loads, including cold startups and periodic step variations in heat flux using two different test sections and two coolants (R236fa and R245fa) for a wide variety of flow conditions. The effects on the base temperature behavior of the test section, heat flux magnitude, mass flux, inlet subcooling, outlet saturation temperature, and fluid were investigated. The transient base temperature response, monitored by an infrared (IR) camera, was recorded simultaneously with the flow regime acquired by a high-speed video camera. For cold startups, it was found that reducing the inlet orifice width, heat flux magnitude, inlet subcooling, and outlet saturation temperature but increasing the mass flux decreased the maximum base temperature. Meanwhile, the time required to initiate boiling increased with the inlet orifice width, mass flux, inlet subcooling, and outlet saturation temperature but decreased with the heat flux magnitude. For periodic variations in heat flux, the resulting base temperature was found to oscillate and then damp out along the flow direction. Furthermore, the effects of mass flux and heat flux pulsation period were insignificant.

Seeing through Multimode Fibers with Deep Learning

Deep neural networks were used to classify and reconstruct the distal intensity speckle patterns generated by projecting handwritten digits on the proximal facet of multimode fibers up to 1km in length.