The latest advancements in liquid cooling solutions for data centers bring promising innovations from Hewlett Packard Enterprise (HPE) and Supermicro. Both companies have responded to the increasing cooling challenges posed by large-scale AI deployments with cutting-edge liquid cooling architectures. These technologies aim to significantly enhance energy efficiency, reduce operational costs, and help meet environmental sustainability targets. By addressing the substantial heat generated by modern computing workloads, these solutions also promise to streamline data center design and operation, contributing to more compact and efficient setups.
HPE’s Fanless Liquid Cooling Architecture
Hewlett Packard Enterprise (HPE) recently introduced a revolutionary liquid cooling design that does away with the need for fans. This advancement utilizes an 8-element cooling architecture responsible for cooling various server components like the server blade, network fabric, and coolant distribution unit (CDU). Typical liquid cooling methods blend chillers, fans, and natural heat dissipation, but HPE’s method relies entirely on direct liquid cooling without fans. This innovative approach offers a significant leap in data center cooling technology, previously deployed in HPE’s Cray EX supercomputers and now being extended to a broader range of non-supercomputing systems.
The fanless design proposed by HPE stands out for its potential to drastically reduce both the physical and operational footprint of data centers. Antonio Neri, president and CEO of HPE, highlighted the design’s potential to reduce data center floor space by 50% and decrease the cooling power required per server blade by 37%. Beyond the operational benefits, this architecture also supports organizations in their quest for generative AI technologies by improving energy efficiency and reducing costs. This is a particularly vital development for businesses aiming to balance performance with sustainability.
Supermicro’s Liquid Cooling Tailored for AI Workloads
Supermicro has introduced a liquid cooling solution specifically designed to support Nvidia’s Blackwell platform, which caters to AI and high-performance computing workloads. Their comprehensive system includes components such as CDUs, cold plates, CDMs, cooling towers, and end-to-end management software. This setup efficiently manages the immense heat output from Nvidia’s GB200 Grace Blackwell Superchip NVL72 platform, with cold plates capable of dissipating up to 1600W. This design accommodates the stringent cooling needs of Nvidia’s powerful hardware, ensuring sustained optimal performance.
Supermicro’s liquid-cooled SuperClusters are equipped to handle up to 72 Blackwell GPUs and 36 Grace CPUs connected using Nvidia’s NVLink network. This network allows for an impressive 130 terabytes per second (TB/s) of GPU communication, minimizing latency and enhancing performance. Currently being sampled by select customers, Supermicro aims to roll out full-scale production by the end of the fourth quarter, showcasing the growing importance of advanced liquid cooling systems in supporting modern AI and high-performance computing needs. These developments reflect the industry’s shift towards more robust and energy-efficient cooling technologies.
The Growing Industry Consensus on Liquid Cooling
HPE and Supermicro’s advancements underscore an industry-wide shift towards more efficient and powerful liquid cooling systems. These developments are emblematic of a broader recognition within the industry about the importance of advanced cooling technologies in managing the thermal challenges posed by modern high-density computing environments. Significant collaborations, such as that between Vertiv and Nvidia, have produced a liquid cooling reference architecture tailored for GPU-loaded AI factories, further underlining this trend.
In addition to HPE and Supermicro, other industry players like Schneider Electric have provided guidelines for implementing liquid cooling in high-density data centers. New entrants such as Accelsius, with their NeuCool solution, and ZutaCore, with HyperCool, demonstrate the expanded interest in liquid cooling technologies. These innovative solutions offer dual-phase, direct-to-chip cooling capable of deployment without extensive retrofits to existing data centers. The growing market for liquid cooling highlights its benefits in terms of cost savings, carbon emissions reduction, and meeting regulatory compliance.
Innovations in Immersion Cooling Technologies
Another promising approach in advanced cooling techniques is immersion cooling. Once regarded as a niche technology, immersion cooling has been gradually entering the mainstream. This technique involves submerging server components in a dielectric fluid, which provides superior heat transfer and can significantly reduce cooling energy consumption. Some proponents suggest it may eventually complement or even replace other cooling methods in specific scenarios.
The rising interest in immersion cooling is evident beyond traditional IT infrastructures. Supermicro’s new server, featuring a self-contained liquid cooling system, has garnered additional attention, reflecting a broader acceptance and application of advanced cooling technologies. Immersion cooling not only aligns with industry trends towards efficiency and sustainability but also provides a robust solution for managing the thermal demands of contemporary high-performance computing environments. This innovation is poised to play a critical role in the future of data center cooling.
The Importance of Efficient Cooling in Data Centers
Recent advancements in liquid cooling solutions for data centers bring exciting innovations from Hewlett Packard Enterprise (HPE) and Supermicro. Both companies are tackling the growing cooling demands brought on by large-scale AI deployments with state-of-the-art liquid cooling technologies. These new solutions aim to considerably improve energy efficiency, cut down on operational costs, and support environmental sustainability goals. By managing the intense heat generated by modern computing workloads, the technologies promise to simplify data center design and operation, making them more compact and efficient. These developments are becoming increasingly crucial as the data industry continues to expand, requiring scalable and eco-friendly cooling systems. HPE and Supermicro’s pioneering efforts set a new benchmark for the industry, demonstrating how innovative cooling solutions can significantly impact not just performance, but also the overall resource consumption and footprint of data centers. By taking these steps, they contribute to creating more sustainable IT infrastructure for the future.
