Surprisingly Useful AI Article Enhancements
PanaAI has announced a partnership with Nvidia and ASPAC Ai Computing Pty Ltd, a subsidiary of Nasdaq-listed Junee Limited, to create the most powerful Ai supercomputer in the Southern Hemisphere. The project, named PanaAI AUS AISF, is set to launch in early 2025 and will be based in Australia.
The initiative aims to provide advanced cloud computing services to public and private organisations, supporting the development of large foundational models for generative Ai and driving innovation across industries.
Building a High-Performance Computing System
The PanaAI AUS AISF system will incorporate up to 4,088 Nvidia H200 Tensor Core GPUs, interconnected through the Nvidia Quantum-2 InfiniBand platform. This configuration is expected to achieve a theoretical peak performance of approximately 16.4 exaflops in half-precision (FP8).
The supercomputer will address the growing computational demands of Ai workloads in Australia, supporting both public and private sector projects. Once operational, it is anticipated to be the largest Ai supercomputing cluster in Australia and the fastest in the Southern Hemisphere.
Leveraging Advanced GPU Technology
The system will utilise Nvidia’s Hopper architecture-based H200 Tensor Core GPUs, offering improvements in memory size, bandwidth, and processing power.
Compared to the H100, the H200 provides 1.8 times larger memory at 141 GB and 1.4 times higher bandwidth at 4.8 TB/s. This enables the system to manage larger datasets and model sizes while reducing reliance on external memory.
The H200 also delivers a 45 per cent performance improvement for workloads such as Llama and GPT, enhancing its ability to handle complex generative Ai applications.
Addressing Australia’s Ai Needs
PanaAI AUS AISF will offer a range of cloud-based services, including public cloud solutions, Ai compute power rentals, and subscriptions for large language model training.
The system is designed to prioritise security while addressing the needs of large-scale Ai projects in sectors such as healthcare, energy development, financial services, autonomous driving, and life sciences.
It is expected to rank among the top 10 most powerful Ai supercomputers globally, based on the TOP500 rankings from June 2024.
Accelerating Innovation
PanaAI highlighted the importance of the project in meeting growing global demand for generative Ai and computing resources.
“To meet the growing demand for generative Ai and computing resources, PanaAI is planning to establish multiple world-class, powerful supercomputing systems to provide unparalleled accelerated services to users worldwide,” the company said in a statement.
The partnership with Nvidia will enable PanaAI to rapidly train large foundational models using extensive datasets, including images, acoustics, and sensor information.
Last Updated on December 2, 2024 by Nick Ross
