Venado, the latest supercomputer at the Los Alamos National Laboratory, was jointly developed by Hewlett-Packard Enterprise (HPE) and Nvidia through a collaboration with them. The U.S. Department of Energy, the National Nuclear Security Administration, Los Alamos, HPE, and Nvidia staff gathered on Monday, April 15, to unveil Venado at the laboratory’s Nicholas C. Metropolis Center for Modeling and Simulation.
Featuring the cutting-edge HPE Cray EX computer with Nvidia GH200 Grace Hopper Superchips, Venado boasts a processing power that accommodates 2,560 robust, liquid-cooled Grace Hopper Superchips in the exascale-class HPE Cray EX computer. Additionally, the system incorporates 920 Nvidia Grace CPU Superchips, marking it as the pioneering large-scale structure in the U.S. to leverage Nvidia Grace superchips. Each Grace CPU Superchip from Nvidia is equipped with 144 Shoulder cores that deliver an immediate performance enhancement across various HPC programs.
Venado was meticulously designed with a primary focus on overall efficiency and workflow optimization. It is interconnected through the high-speed HPE Slingshot 11 network and utilizes advanced HPE Cray supercomputer software integrated with AI capabilities to efficiently manage modeling and simulation tasks. Manufactured at HPE’s facility in Chippewa Falls, Wisconsin, Venado derives its name from one of the prominent peaks in New Mexico’s Sangre de Cristo Mountains.
Nvidia Grace Hopper Superchips are specifically engineered to handle high-performance computing and large-scale artificial intelligence applications, combining an Arm-based Nvidia central processing unit with an Nvidia Hopper architecture-based graphics processing unit. These superchips execute millions of instructions per second, surpassing previous chip technologies in performance while consuming less energy. In initial trials, Venado has showcased remarkable results in high-resolution astrophysics simulations and material science-based atomic simulations.
Los Alamos officially took possession of Venado in March, further enhancing their computational capabilities in fundamental research and national security initiatives, as highlighted by Thom Mason, the director of Los Alamos National Laboratory.
David Turk, U.S. Deputy Secretary of Energy, emphasized the pivotal role of supercomputing capabilities in addressing critical challenges faced by national laboratories. He expressed anticipation for the innovative outcomes that the Venado program at Los Alamos could generate across various domains due to its adept integration of artificial intelligence methodologies.
Ian Buck, the vice president of hyperscale and HPC at Nvidia, lauded Venado as a groundbreaking advancement in computational efficiency and energy conservation, propelling scientific discoveries forward. He underscored the collaborative efforts between Los Alamos, HPE, and Nvidia in developing Venado as a versatile tool for researchers to make significant strides in disciplines such as materials science, renewable energy, astrophysics, and more.
Trish Damkroger, the senior vice president and general manager of HPC & AI Infrastructure Solutions at HPE, highlighted the transformative potential of powerful computation with AI-driven capabilities in fostering significant breakthroughs and addressing real-world challenges. She commended Los Alamos National Laboratory’s pioneering research efforts and the enduring partnership with HPE that led to the creation of Venado and Nvidia’s innovative AI solutions, poised to accelerate discoveries in fields like material science and energy research.
Venado stands as a testament to a collaborative codesign approach that lays the groundwork for future partnerships aimed at advancing processing, storage, and software technologies. Codesign methodology leverages the collective expertise of vendors, technology architects, software developers, scientists, computer experts, and mathematicians to make informed decisions regarding hardware and software components.
