CI-New: Collaborative Research: Modeling the Next-Generation Hybrid Cooling Systems for High-Performance Processors

未来高性能芯片的设计受到严峻的温度挑战的阻碍。例如，现有的冷却机制不能有效地冷却在百亿亿级系统中预期的极高功率密度。新兴的冷却技术可能会解决这些温度挑战，但计算机工程师并不容易进行实验。事实上，在冷却技术用于系统设计和优化之前，存在相当大的滞后。这种滞后导致设计质量被搁置。为此，该项目提出了一种软件基础设施，可以对尖端的冷却方法进行精确建模，并促进计算和冷却系统的相互定制，从而大大提高系统的能源效率。提出的基础设施是一个系统级设计自动化工具，包括新兴冷却方法的紧凑热模型（热电冷却器、带有微通道或纳米孔的两相冷却、相变材料和单相液体冷却）。项目计划包括：(1)将新的设备级模型合成为紧凑的表示；(2)通过对原型的测量来验证所提出的模型；(3)开发自动化工具，与给定的计算系统共同设计混合定制冷却子系统。该项目更广泛的影响将是通过使高功率密度系统的高效设计成为可能，帮助推动计算超越摩尔定律的限制。提出的基础设施将使设计自动化、计算机体系结构和系统设计方面的变革性研究与新兴的冷却技术，特别是在性能、可变性、能量、异质性和跨层设计方面。基础设施将作为开源软件发布给社区。该项目的跨学科性质为本科生项目和面向代表性不足的群体和K-12学生的拓展项目创造了充足的机会。

Design of future high-performance chips is hindered by severe temperature challenges. For example, existing cooling mechanisms cannot efficiently cool the extremely high power densities that are expected in exascale systems. Emerging cooling technologies, which may address these temperature challenges, are not easily accessible for experimentation to computer engineers. In fact, there is a substantial lag before a cooling technology becomes available for system design and optimization. Such lags cause design quality to be left on the table. To this end, this project proposes a software infrastructure that enables accurate modeling of cutting-edge cooling methods and that facilitates mutually customizing the computing and cooling systems to dramatically push system energy efficiency. The proposed infrastructure is a system-level design automation tool that includes compact thermal models of emerging cooling methods (thermoelectric coolers, two-phase cooling with microchannels or nanopores, phase-change materials, and single-phase liquid cooling). The project plan includes (1) synthesizing novel device-level models into compact representations; (2) using measurements on prototypes for validation of the proposed models; and (3) developing automation tooling to co-design hybrid customized cooling subsystems together with a given computing system. The broader impact of the project will be to help advance computing beyond the limitations of Moore's Law by making efficient design of high-power-density systems possible. The proposed infrastructure will enable transformative research in design automation, computer architecture, and system design with emerging cooling technologies, particularly in the dimensions of performance, variability, energy, heterogeneity, and cross-layer design. The infrastructure will be released to the community as open source software. The interdisciplinary nature of the project creates ample opportunities for undergraduate projects and outreach programs for underrepresented groups and K-12 students.