One thing was clear at the just-wrapped NVIDIA GTC event: the race to cool the next generation of HPC and AI systems is intensifying.
Let’s take a quick look at some of our favorite announcements.
Chemours, NTT DATA, AND Hibiya Engineering Team Up for Immersion Cooling Trial
In a bid to revolutionize data center cooling, Chemours, NTT DATA, and Hibiya Engineering launched a collaborative immersion cooling trial that aims to enhance data center efficiency by submerging servers in specialized cooling fluids, potentially reducing energy consumption and operational costs.
Immersion cooling is an interesting approach to escalating heat loads from advanced computing tasks, but the industry hasn’t seen broad adoption of the technology. We’ll be watching this one to see what comes out of the trial.
COOLIT Systems Unveils High-Capacity Cooling Solutions
CoolIT Systems introduced its new CHx1500, a high-capacity row-based cooling solution for AI and HPC environments that the company touts as the “world’s highest performance” row-based CDU. The solution features stainless-steel piping and a 25-micron filtration system.
Additionally, CoolIT has advanced its single-phase direct liquid cooling technology with a new 4000W coldplate, capable of efficiently dissipating heat from high-power components.
CoolIT’s innovations address the critical need for effective thermal management in high-density computing environments, ensuring optimal performance and reliability. This is a nice solution from one of the leaders in liquid cooling.
Intel’s Superfluid Cooling for NVIDIA’S Next-Gen Servers
Intel is reportedly collaborating with Taiwanese suppliers to develop its proprietary SuperFluid cooling technology for AI servers. Designed to handle thermal loads of up to 1,500W per chip, SuperFluid employs non-conductive dielectric fluids to mitigate leakage risks associated with traditional liquid cooling methods.
NVIDIA’S 600KW Rubin Ultra NVL576 Rack
NVIDIA announced plans to release the Rubin Ultra NVL576, a liquid-cooled rack system consuming approximately 600kW, slated for launch in the second half of 2027. Equipped with reticle-sized GPUs and delivering 15 exaflops of FP4 inference performance, this system represents a significant leap in computing power.
Key Takeaways
- Cooling Innovation: Collaborations and technological advancements in cooling are crucial to support the escalating demands of AI and HPC workloads.
- Strategic Positioning: Companies like Intel and Cool IT are investing in specialized cooling technologies to maintain competitiveness in the evolving data center landscape.
- Infrastructure Challenges: NVIDIA’s forthcoming high-power systems necessitate a reevaluation of data center design and energy provisioning strategies.
