HPE has announced enhancements to its NVIDIA AI Computing by HPE product portfolio, focused on large-scale AI factoriesas well as supercomputers with which customers can scale, deploy efficiently and get useful information in less time. Full stack AI solutions with NVIDIA include compute, GPU, networking, liquid cooling, software and integrated services.
Inside its exascale-class second-generation supercomputing platform HPE Cray Supercomputing GX5000HPE makes different NVIDIA products available to organizations and researchers. Thus, it has presented a computing blade with a liquid-cooled NVIDIA Vera CPU.
Each compute blade HPE Cray Supercomputing GX240 It has up to 16 NVIDIA Vera CPUs, and is designed to support AI workloads. It has the capacity to scale up to 40 blades with 640 NVIDIA Vera CPUs and 56,320 ARM-compatible NVIDIA Olympus cores per rack. Now available with HPE Cray Supercomputing GX5000, NVIDIA Quantum-X800 InfiniBand switches offer 144 ports with 800 GB/s connectivity per port, with energy-efficient features.
HPE is also strengthening the HPE AI Factory portfolio for service providers, sovereign entities and large enterprises with the NVIDIA Vera Rubin platform and NVIDIA Blackwell architecture. In this sense, the company has presented the next generation rack-scale system NVIDIA Vera Rubin NVLK72 by HPE, designed for models that exceed one trillion parameters and for neoclouds.
It incorporates 36 NVIDIA Vera CPUs, 72 NVIDIA Rubin GPUs, NVIDIA NVLink 6th generation scalability networks, NVIDIA ConnectX-9 SuperNIC and NVIDIA BlueField-4 DPUs and HPE liquid cooling.
He HPE Compute XD700 is a new AI server inspired by the Open Compute Project and based on NVIDIA HGX Rubin NVL8. The system is designed to offer more GPU density per rack and reduce space, power and cooling costs, as well as increase AI training and inference performance. Each XD700 server rack supports up to 128 Rubin GPUs, delivering twice the density of the previous generation.
These solutions are accompanied by various software updates and services from HPE and NVIDIA that enable faster AI deployments of large-scale projects. Thus, the HPE AI Factory portfolio is supported by the NVIDIA Cloud Partner program. Through technical collaboration with NVIDIA, HPE AI factories are ready for NVIDIA CLoud Provider certification, enabling cloud service providers to streamline the validation process.
Separately, HPE has enhanced the AI Factory portfolio by supporting multi-tenancy models for virtual machines with GPU gateway and Kubernetes namespace assurance through NVIDIA Multi-Instance GPU, enabled by SUSE Virtualization and SUSE Rancher Prime Suite.
HPE AI Factory portfolio supports Red Hat Enterprise Linux and OpenShiftand integrates with NVIDIA AI Enterprise solutions for customers requiring enterprise Linux. AI Factory at scale and AI Factory sovereign will offer the NVIDIA Mission Control softwarewhich optimizes the AI factory, from workload orchestration with NVIDIA Run:ai to NVIDIA Dynamo, which handles monitoring and autonomous recovery.
