HPE has announced that Start the distribution of its NVIDIA solution GB200 NVL72the First launches based on the Nvidia Blackwell family. It is a system at rack scale, designed to facilitate service suppliers and large companies the agile implementation of clusters of the very large and complex clusters, through advanced solutions of liquid cooling.
The NVIDIA GB200 NVL72 system incorporates shared memory architecture and low latency, equipped with GPU technology designed for AI models considered extremely large. That is, it is prepared for models that have more than 1,000 million parameters in a single memory space.
It also offers a fluid integration of CPUS, GPUS, computer and switching trays, NVIDIA networks and software. It therefore combines high performance to work with highly parallelizable work loads, such as training and inference of generative AI models, with NVIDIA software applications.
The NVIDIA GB200 NVK72 of HPE is part of a range of high performance and supercomputing systems designed to address all types of generative use cases, scientific discoveries and other workloads that are intensive in computing. Includes 72 GPUS NVIDIA BLACKWELL and 36 CPUS NVIDIA GRce interconnected through High Speed Nvlink Nvlink. It can carry up to 13.5 TB of HBM3 memory with a bandwidth of 576 TB per second, and has direct liquid cooling technology.
Trish Damkroger, vice -president of Senior Y Directora General de HPC & Ai Infrastructure Solutions en hperecalled that «IA service providers and business models developers face enormous pressure to offer scalability, extreme performance and rapid implementation times. As creators of the three fastest systems in the world with direct liquid cooling, in HPE we offer our customers a lower training cost by Token and a top -level performance, backed by our leading experience in services in the sector”.
By Nvidia, Bob Pette, its vice president of business platformshas stressed that «Engineers, scientists and researchers need avant -garde liquid cooling technology to respond to the growing power requirements and computing capacity. Thanks to the continuous collaboration between HPE and NVIDIA, the first shipment of the NVIDIA GB200 NVL72 by HPE will allow service providers and large companies to build, implement and efficiently climb large clusters of AI”.
