ZutaCore will introduce its HyperCool technology, a waterless liquid cooling solution, at SC24 in Atlanta. Designed for direct-to-chip cooling, HyperCool leverages two-phase heat transfer to eliminate the need for facility water loops, reducing operational risks such as water leakage and erosion. This innovation addresses the extreme power densities required by modern AI accelerators while lowering energy consumption and operational costs. By achieving cooling capacities of up to 2,800 watts per chip and supporting rack power densities of 120kW, HyperCool provides a scalable solution for AI factories and data centers.
One of the key highlights of ZutaCore’s SC24 demonstration will be its integration with AMD Instinct™ MI300X GPUs, specifically engineered to handle the demands of next-generation AI workloads. The waterless cold plates are designed to manage the thermal output of these high-performance chips, ensuring consistent performance under heavy computational loads. This collaboration emphasizes the potential of ZutaCore’s technology to drive more sustainable and efficient operations for AI-focused data centers without relying on water-based cooling systems.
• Two-Phase Direct-to-Chip Cooling: ZutaCore’s HyperCool technology uses a two-phase process where a liquid refrigerant absorbs heat from the chip, evaporates into vapor, and transfers the heat away efficiently.
• Water-Free Operation: Unlike traditional liquid cooling systems, HyperCool eliminates the need for water, removing risks associated with leaks, corrosion, and water scarcity in high-density AI and HPC environments.
• High Cooling Capacity: The system’s advanced cold plates can dissipate up to 2,800 watts per CPU or GPU, supporting dense configurations with rack power densities of up to 120kW.
• Environmentally Friendly Refrigerant: HyperCool uses a non-conductive, low-global-warming-potential (GWP) refrigerant, ensuring compliance with modern environmental standards while delivering effective thermal management.
• Scalability and Versatility: The technology is optimized for advanced processors like AMD Instinct MI300X, NVIDIA H200 GPUs, and high-performance CPUs, making it suitable for AI, HPC, and data center applications.
