Nvidia announces GPU-based computing platform for HPC and AI workloads

Nvidia HGX-2 utilises Nvidia's GPU technology in a bid to propel the company into AI and HPC

Nvidia has unveiled Nvidia HGX-2, the first "unified computing platform" for high performance computing (HPC) and artificial intelligence, based on its GPU technology.

The cloud server platform was unveiled at the company's GTC conference in Taiwan on Tuesday.

The platform is said to boast multi-precision computing capabilities, providing unique flexibility to support the future of computing.

Using FP64 and FP32 for scientific computing and simulations, it allows high-precision calculations while also enabling FP16 and Int8 for AI training and inference, Nvidia said, adding that this will help meet the requirements of the growing number of applications that combine HPC with AI.

"The world of computing has changed," said Jensen Huang, founder and CEO of Nvidia. "CPU scaling has slowed at a time when computing demand is skyrocketing. NVIDIA's HGX-2 with Tensor Core GPUs gives the industry a powerful, versatile computing platform that fuses HPC and AI to solve the world's grand challenges."

Huang added that HGX-2-serves as a "building block" for manufacturers to create some of the most advanced systems for HPC and AI, achieving a record AI training speeds of 15,500 images per second on the ResNet-50 training benchmark, and replacing up to 300 CPU-only servers.

It incorporates such features as Nvidia's NVSwitch interconnect fabric tech, which links 16 Tesla V100 Tensor Core GPUs to work as a single, giant GPU delivering two petaflops of AI performance.

To coincide with the announcement, a number of leading computer makers shared plans to bring to market systems based on Nvidia's new HGX-2 platform. These include Lenovo, QCT, Supermicro and Wiwynn. They announced they'll be bringing their own HGX-2-based systems to market later this year.

Alognside these, there'll be four original design manufacturers comprising Foxconn, Inventec, Quanta and Wistron designing HGX-2-based systems, also expected later this year, for use in some of the world's largest cloud datacenters

HGX-2 comes a year after the launch of the original NVIDIA HGX-1, unveiled at Computex last year.