Aimed at cloud providers, the Nvidia HGX-2 supports FP64 and FP32 calculations for scientific computing and simulations, as well as FP16 and Int8 for AI training and inference. The company claims this "unprecedented versatility" addresses the needs of applications that combine HPC and AI.
"The world of computing has changed," said Nvidia founder and chief executive Jensen Huang,
"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."
It features Nvidia's NVSwitch interconnect fabric, which allows 16 Tesla V100 Tensor Core GPUs to work as one, delivering two petaflops of AI performance.
The first system built using the HGX-2 was the recently announced Nvidia DGX-2.
Lenovo, QCT, Supermicro and Wiwynn have announced plans to offer HGX-2-based servers later this year, while leading original design manufacturers Foxconn, Inventec, Quanta and Wistron expect to release HGX-2-based systems for cloud data centres in the same timeframe.
Lenovo DCG vice-president and general manager Paul Ju said: "Nvidia's HGX-2 ups the ante with a design capable of delivering two petaflops of performance for AI and HPC-intensive workloads.
"With the HGX-2 server building block, we'll be able to quickly develop new systems that can meet the growing needs of our customers who demand the highest performance at scale."
Foxconn corporate executive vice-president Ed Wu said: "Foxconn has long been dedicated to hyperscale computing solutions and successfully won customer recognition.
"We're glad to work with Nvidia for the HGX-2 project, which is the most promising solution to fulfil explosive demand from AI/DL."