This is an article related to Nouvelle Alexandrie. Click here for more information.

National Center for High-Performance Computing

From MicrasWiki
Jump to navigationJump to search
A view of the CNCAR main building.

The National Center for High-Performance Computing (Alexandrian: Centre national de calcul à haute performance; Martino: Centro Nacional de Computación de Alto Rendimiento, CNCAR), also known as the National Supercomputer (Alexandrian: Superordinateur national; Martino: Superordenador Nacional), is a federal research center for computational science in Cárdenas, the capital of Nouvelle Alexandrie. It is the nation's primary provider of high-performance computing capacity for academia, industry, and government agencies. The CNCAR provides the computing infrastructure for federal research, complementing the Institute for Computer Science and Robotics at the Máquina Science Park, which leads federal research in computer science itself.

History

The CNCAR was established in 1695 AN to centralize access to high-performance computing for federal research. Initially conceived as a distributed network of smaller clusters, it consolidated into a single, purpose-built campus in Cárdenas by 1710 AN. Its early work focused on economic simulation, climate and disaster modeling, and automation research.

In 1738 AN, the Frontera Program allocated over NAX€4.2 billion to the expansion and upgrading of the CNCAR. The funding paid for a new generation of computing hardware and for AlexAI, the center's artificial-intelligence computing platform, which came online in 1740 AN and hosts research under the Frontera Program's IA Génesis program.

Purpose

The CNCAR provides computing capacity for scientific and economic research, including:

  • computational science in physics, chemistry, and engineering;
  • large-scale economic modeling for economic trends, market analysis, finance, and international trade;
  • mathematical simulation of complex equations in theoretical and applied mathematics;
  • artificial intelligence and machine learning research; and
  • data science and analytics for public health, urban planning, and the social sciences.

The center services over 80 million node-hours per year across its computing resources, with Helix as the primary system. In 1740 AN, computational science accounted for 39% of all allocation requests, economic modeling for 21%, mathematical simulation for 17%, data science and analytics for 15%, and artificial intelligence for 8%.

Helix

The Helix Supercomputer is the CNCAR's core facility. It uses a hybrid architecture of general-purpose processors, graphics accelerators, and tensor processors. Peak performance was measured in 1742 AN at 1.2 exaFLOPS. The system carries 1.2 PB of main memory, 48 PB of solid-state storage, and a parallel file system of over 200 PB, connected by a 200 Gbit/s low-latency interconnect.

  • 12,288 nodes with dual 64-core CPUs provide bulk processing power;
  • 49,152 graphics accelerators handle deep-learning workloads; and
  • 1,024 tensor-processing pods are dedicated to artificial-intelligence research.

CPU and GPU nodes are cooled by liquid immersion, with air cooling for auxiliary systems. Power consumption peaks at 24 MW, and residual heat is fed into the district heating network of Cárdenas.

Organization

The CNCAR is overseen by the Department of Research and Development and led by a Director appointed by the Secretary. It has three service departments:

  • Director
    • Department of Administration
    • Department of Public Engagement
    • Department of Software and Hardware Development

The center's advisory board, the Computing Advisory Council, advises on allocations to public and private users, ethical use, and responsible development of artificial intelligence.

See also