Chipmakers Race for Denser Compute
The computing world constantly demands faster performance. This demand has driven the semiconductor industry to develop smaller devices with denser circuitry pattern capabilities. But in the last decade, the scaling that let us witness a billion-fold increase in transistors on a chip over a matter of decades has been slowing [1]. As the anticipated slowdown of Moore’s Law approaches, major semiconductor companies are seeking alternative methods to increase computational density in performance computing applications.
Now, instead of designing sophisticated, monolithic chips that incorporate all the important elements on a single silicon die, chipmakers are creating products that break those larger designs into smaller pieces, called chiplets or tiles, then combining them in new ways[2]. A major challenge in developing these advanced integration strategies is the thermal management of multi-chip packaging. In this paper, we evaluate the challenges faced by chipmakers today and how microconvective liquid cooling enables the most powerful multi-die processors on the market to reach new levels of processing speed.
Inferior Cooling Threatens Next-Gen Processors
Multi-die packaging faces a multitude of thermal challenges. Chiplet technology, used in multi-die packaging, could lead to new generations of advanced computational devices for enterprise, data center, and edge applications, enabling the enormous scale of data created by humans and machines in the next decade[3]. Yet, chipmakers are facing a major challenge when developing these advanced integration strategies: thermal management. Dies of varying powers are spatially distributed in processor packages, causing localized hot spots and non-uniform temperature patterns. If not adequately addressed, local hot spots limit device operating power, lower processor efficiency, and reduce device lifetime.
For these reasons, chipmakers are searching for a powerful cooling solution that influences future capabilities while supporting next-generation product development.
Eliminate Hot Spots with Microconvective Liquid Cooling
JetCool’s patented microconvective liquid cooling technology uses arrays of fluid jets to cool the industry’s highest power devices. Unlike typical heat sinks or cold plates that pass fluid over a surface, our cooling jets route fluid directly at the surface, creating an order-of-magnitude improvement in heat transfer. Eliminating all thermal pastes and interface materials, our cooling produces a heat transfer coefficient 10x greater than competing approaches enabling devices to be more powerful, efficient, and compact.
Case Study: Testing Thermal Performance on a Multi-Die Processor
JetCool tested its cooling performance against a high-performance heat pipe on the AMD Threadripper 3960X, an unlocked commercially available multi-die processor known for its high thermal design power (TDP) and popularity in overclocking studies. JetCool designed a modular microconvective heat sink built for the purpose of targeting cooling at device hot spots. To begin, the baseline heat pipe was installed using thermal paste, and performance metrics were taken at nominal non-overclocked settings. From there, the overclocking procedure was applied with the heat pipe until the thermal limit was reached. The JetCool cooling modules were then installed and tested under the same settings to compare performance.
To learn more about how liquid cooling can improve semiconductor performance, check out our semiconductor solutions.
References
[3] https://www.forbes.com/sites/tomcoughlin/2019/05/11/chiplets-for-all/?sh=14bbba9c5e13
This is a debrief containing data from the paper, Thermal Performance of Modular Microconvective Heat Sinks for Multi-Die Processor Assemblies by authors Chris May, Jordan Mizerak, David Earley, and Bernard Malouin. Published by ASME 2021.
