GIGABYTE and NVIDIA joint views on HPC market at ISC 2023

The target of the company is to address the entire spectrum of HPC workloads, from AI, Machine Learning, virtualisation, data analytics and business intelligence, 5G and Edge Computing, and cloud computing.

What the partnership brings to the market in 2023?

NVIDIA has partnered with GIGABYTE to develop and launch GIGABYTE's latest G363 Series and the G593 Series featuring up to eight units of H100 SXM5 GPUs per system, with the 4th Generation NVLink interconnect.

The NVIDIA H100 SXM5 GPU is a powerful computing solution designed for data centres and supercomputing environments. Its massive parallel processing power enables complex simulations and computational tasks with greater speed and efficiency. Use cases in scientific research, quantum computing, deep learning, and AI training can leverage the H100 SXM5 for data analytics and modelling, enabling faster insights and informed decision-making.

Meanwhile, NVIDIA has also been working with GIGABYTE to bring NVIDIA's Grace-Grace and Grace-Hopper computing solutions to the market. GIGABYTE's expertise in dense GPU computing and optimised cooling ensures that performance and efficiency can be achieved on an optimum level.

Grace-Grace's architecture is specifically designed to address the challenges of modern data-centric workloads. Grace-Grace features a power-efficient, scalable, and coherent multi-core architecture, comprising multiple Arm-based processing units interconnected with NVIDIA's advanced communication fabric. Another highlight is the memory subsystem in Grace-Grace. A high-capacity and high-bandwidth HBM LPDDR5X memory subsystem offers fast data access and low-latency communication between the CPU cores and memory.

Built upon the features of Grace-Grace, Grace-Hopper integrates specialised hardware accelerators and optimised interconnects and seamlessly combines CPU and GPU workloads. Grace-Hopper creates a powerful hybrid computing environment that harnesses the strengths of both processing units, unlocking unprecedented levels of performance for demanding applications.

Both architectures enable seamless data movement and efficient parallel execution, essential for handling large-scale scientific simulations, AI training, and data analytics. NVIDIA and GIGABYTE are excited to see customers equipped with such technologies.

What’s next?

GIGABYTE has long been building on its success in bringing exclusive product designs to the global HPC market, and tremendous progress has been made in the company’s research and development of future-generation computing hardware.

From APU platforms, CPU-GPU hybrid platforms, direct liquid cooling and immersion cooling total solutions, high-performance storage server platforms, GIGABYTE has long-term OEM/ODM projects with global tech players and hyperscalers.

GIGABYTE will continue to innovate not just in product design but also in business models, commercial offering, and consulting, so that the company continues to support its customers in the constantly evolving HPC market around new computing architectures, flexible designs, and value creation by time-to-market leadership.