Flex’s OCP ORv3-inspired liquid-cooled systems are designed to support the most demanding artificial intelligence (AI) and high-performance computing (HPC) workloads, efficiently cooling up to 120kW per rack and beyond. These racks are integrated with single-phase liquid cooling and enabled for two-phase liquid cooling.
The rapid adoption of Generative AI, AI-enabled applications, and HPC workloads has accelerated the demand for advanced computing that requires high-power GPUs and AI accelerators. With this increased performance comes tremendous power consumption and heat generation.
Typical server processor Thermal Design Power (TDP) is scaling from 250W to 350W, with roadmaps to 500W+, and GPUs are on the path to scale from 800W today to 1200W and beyond. Liquid cooling has become a necessity, driving cloud service providers to re-architect the data center with the necessary IT infrastructure and cooling to support these higher thermal and power requirements.
As AI-enabled applications and other high-performance workloads push the limits of conventional cooling methods, the need for advanced solutions becomes critical. Flex has developed customizable, open standards based server and rack designs that integrate JetCool’s SmartPlate™ direct-to-chip liquid cooling products.
JetCool’s patented microconvective cooling® technology extends the value of single-phase, direct-to-chip liquid cooling deployments to over 1,500W per socket, more than enough headroom to accommodate the most advanced AI servers.
The Flex liquid-cooled system and rack solutions provide a versatile platform for hyperscalers and data center customers to design customized AI chip architectures and tailored server configurations to meet their unique performance requirements.
