HEAVY.AI has announced the general availability of its analytics platform optimized for the NVIDIA GH200 Grace Hopper Superchip, a cutting-edge architecture featuring an Arm-based CPU and NVIDIA Hopper GPU.

This upgrade enables users to process data faster and at a reduced cost by utilizing the ultra-fast NVIDIA NVLink-C2C interconnect, which boasts 900GB/sec of bidirectional bandwidth. This advancement allows HEAVY.AI users to query and visualize massive datasets interactively, even those exceeding GPU memory capacity.

Additionally, the platform can now leverage the NVIDIA GB200 Grace Blackwell Superchip, which delivers substantial speed improvements for real-time large language model inference.

The introduction of the HEAVY.AI platform with these advanced architectures also presents noteworthy cost savings. HEAVY.AI has provided a publicly available demo showcasing over 20 billion records of ship locations, previously requiring significant GPU resources, now efficiently running on a single Superchip—illustrating nearly 70% in hardware cost savings.

“The NVIDIA Grace Hopper Superchip changes the game in terms of being able to provide best-in-class performance over our customers’ largest datasets,” said Todd Mostak, Co-Founder and CEO of HEAVY.AI. He emphasized the platform's ability to provide both increased performance and lower costs.

Ivan Goldwasser, NVIDIA's director of data center CPUs, noted the growing demand from customers to enhance performance while reducing costs when handling large datasets. The integration enables accelerated analytics capabilities for significant data processing and visualization improvements.

Recent benchmarks comparing HEAVY.AI's HeavyDB database against popular CPU-based systems highlight these advantages, showing up to 21 times faster performance and up to 9 times lower operational costs. HEAVY.AI aims to establish GPU-accelerated analytics as the default solution for analyzing very large datasets.