In the dynamic realm of data centre cooling, the CTO of Data Centre Dynamics, Steven Warren, leads a panel discussion featuring industry experts on the Vertiv platform. This is part two of a four-part article.

There is a collective acknowledgment that the shift towards liquid cooling is both inevitable and necessary. <a href="https://www.freepik.com/free-photo/cloud-storage-banner-background_16016448.htm#query=data%20centres&position=1&from_view=search&track=ais&uuid=f934f5a4-4fc8-4120-8ee7-080c384e5e17">Image supplied by rawpixel.com</a> on Freepik

There is a collective acknowledgment that the shift towards liquid cooling is both inevitable and necessary. Image supplied by rawpixel.com on Freepik

The focus of this discourse is on an innovative approach to cooling – direct chip cooling – and whether it is emerging as the front-runner in the data centre cooling landscape.

The panel includes Ali Heydari, data centre technologist at Nvidia; Devdatta Kulkarni, principal thermo architect for Intel Corp.; Tim Shedd, engineering technologist at Dell; and Rob Curtis, a prominent figure in the industry.

… continued from part one.

Deciphering the shift: liquid cooling perspectives from industry leaders

Panellists shared insights into the evolving landscape of data centre cooling. There is a collective acknowledgment that the shift towards liquid cooling is both inevitable and necessary. The physics of cooling demands a departure from traditional methods, with liquid cooling emerging as the more viable option for managing the escalating power densities in data centres.

Curtis, a seasoned server hardware thermal designer, emphasises the ongoing escalation of power in GPU and CPU roadmaps. He highlights the challenge posed by increasing power and the need to address it, setting the stage for liquid cooling. The inefficiency of air cooling in accommodating high-density chips is underscored, making liquid cooling an essential avenue for maintaining or increasing data centre density.

Kulkarni delves into the adoption of liquid cooling from Intel’s standpoint. With a focus on CPU, GPU and accelerator thermal architecture, he outlines the industry trend of rising heat flux density, necessitating lower package thermal resistance. The imperative for sustainable solutions in the face of high-power chips and the energy efficiency of data centres becomes apparent.

Heydari from Nvidia provides a unique perspective, drawing on his four years of experience in data centre technologies. Addressing the dilemma presented by escalating power densities, he discusses the laws of physics as a driving force. The inadequacy of air cooling and the inevitable transition to liquid cooling are highlighted, with a cautionary note that liquid cooling may not suffice in the long term if power density trends persist.

Returning to Intel, Kulkarni underscores the evolving thermal challenges associated with HPC and data centre workloads. The integration of more components on the same package leads to increased resistance, demanding lower package thermal resistance. He stresses the need to balance performance and sustainability, recognising liquid cooling as a vital solution in this dynamic scenario.

The panellists anticipate future challenges and developments, setting the stage for further exploration into alternative solutions and the industry’s roadmap in the realm of data centre cooling. There’s no one-size-fits-all solution. It’s essential to understand the specific requirements and the workload characteristics. There is a need to consider factors like rack density, the specific use case, and the growth trajectory of the data centre.

For lower-density applications, air cooling might still be a viable and cost-effective solution. Liquid cooling, in this context, might be phased in where needed, especially for high-density components like GPUs and accelerators.

When it comes to the return on investment (ROI), it’s crucial to evaluate the overall cost of ownership, taking into account not just the initial investment but also the operational costs. Factors such as energy efficiency, maintenance and scalability should be carefully considered. For high-density workloads, the ROI of liquid cooling tends to be more favourable due to reduced energy consumption.

However, there are limitations of liquid cooling. If the industry continues to follow the trajectory of increasing power densities, there may be a need to explore innovative solutions beyond the current liquid cooling technologies. The industry needs to invest in research and development to stay ahead of the curve and ensure long-term sustainability.

  • In terms of adoption strategies, it’s important to have a phased approach. A start would be to identify critical areas or high-density components where liquid cooling can provide immediate benefits.
  • Evaluate the performance gains, energy savings and operational efficiencies achieved in these targeted areas.
  • As the technology matures and becomes more standardised, it can be gradually implemented across the data centre.

Continued in part three…